The Leman Paradox

A Curtain With a Guest List

Sun, 13 Sep 2026 00:00:00 GMT

At my grandparents’ house, summer came with an escort. You could not walk from the garden to the kitchen without a delegation forming behind you, mostly flies, plus whatever the heat had woken up that week. And the house faced an impossible choice every July afternoon. Keep the bay windows shut and cook, or open them wide and accept everything the season was offering: the dust, the insects, the moths at dusk, and, some summers, a snake with opinions about the cool of a tiled floor.

Their answer, the strip curtain, hung in the kitchen doorway. Ribbons of plastic you pushed through with your face, which theoretically confused the flies and practically confused the grandchildren. It half worked, the way all its descendants half work. I have been walking through that doorway in my head for years, wondering what the finished version looks like.

The first force field I ever walked through was made of plastic ribbons and my grandmother’s optimism.

Here is the finished version, the sixth climb up this site’s science fiction ladder (after the storm, the gun, the collar, the air and the Devil’s Trill). A film across the open doorway, thin and transparent, almost not there. When someone on the guest list walks through, it turns fluid and parts around them exactly, hugging their outline, suitcase included, sealing behind their heels. For everything else that moves and breathes (the fly riding your shoulder, the mouse at the sill, the July snake) it stays politely, immovably firm. Throw a cardboard box through and it passes. It is not alive, and the door has no quarrel with it. The window stays open all summer. The house chooses its guests.

Four Ancestors, All in Stores

The strange thing about this dream is how much of it is already for sale in pieces. The strip curtain begat the magnetic screen, two panels of mesh that snap themselves shut behind you, whose product pages proudly note that cats and dogs pass freely, meaning, by construction, the neighbor’s cat and anything else that pushes. Mine got field-tested. The dog pushed through but still gave it a few paw swipes, the cat tried to tear it outright, and in both cases the mesh lost. Supermarkets solved the problem differently with the air curtain, a wall of blown air over the entrance that humans cross without noticing and flies cannot fly through. That is a selective fluid barrier, selecting by mass, and its weaknesses sketch my film precisely: it is loud, it is hungry, and it does nothing about what crawls in at floor level.

And then there is the microchip cat flap, the ancestor that already does the whole concept, recognition included, at a height of fifteen centimeters. It reads the chip implanted in your cat, unlocks for that cat alone, and leaves the neighborhood’s raiders outside. Everything this essay wants has shipped, for pets, years ago. The rest is a question of scale, and of teaching the door to see.

Eyes, Brain, Muscle

My first mental image of the mechanism was a curtain of lasers, beams filling the doorway like a moucharabieh of light, parting only for the authorized. I have to give that image up before repurposing it, because light does not push. A laser grid can see a gnat at a tenth of a millimeter but cannot stop one, and a beam strong enough to be a wall would be an oven. So the light becomes the nervous system, and matter stays the muscle.

The eyes exist, off the shelf. Industry has sold “light curtains” for decades, lattices of infrared beams that guard dangerous machines and halt the press the instant anything interrupts them. Add the depth sensors phones already carry, a thermal camera to tell the warm from the cold, and my favorite instruments in the whole dossier, wingbeat sensors. Smart mosquito traps already identify an insect’s species, even its sex, from the frequency of its wingbeats crossing a beam. This door would not see “a small moving dot.” It would optically hear “tiger mosquito, female” on approach.

The brain is ordinary 2026 video. Real-time silhouette segmentation is the same commodity model that cuts you out of your background on every video call, luggage included. The guest list runs on upgraded cat-flap logic, with a token you carry, or a face, or a gait. And the living-versus-not rule falls out of the sensor stack. The mouse is warm, so it is alive, so it waits outside. The thrown box is cold and ballistic and passes. The insect is neither warm nor chipped, but it flaps, and the wingbeat sensor catches what the thermal camera cannot. Each layer covers another’s blind spot.

The muscle comes in three grades, from real to fiction. Air comes first, with steerable jets in the frame firing where the vision points and stripping the hitchhiker off your shoulder as you cross. It is supermarket technology with aim. Then the membrane itself, which materials science has been quietly preparing. Sheets that switch between drape-soft and board-stiff on command exist in soft robotics as “jamming” layers, and fluids that harden in milliseconds under a magnetic field already ship inside car suspensions. The science fiction is not the switching but the resolution, a film whose stiffness is pixelated finely enough to melt a hole in the exact shape of you, suitcase and all, opening slightly ahead of your stride like a focus that predicts.

And the laser survives in its one honest physical role, as a sniper rather than a wall. A fence that tracked mosquitoes in flight and shot them down was demonstrated on a conference stage in 2010 and has never shipped since, eye safety being one of the reasons. My door inherits both the trick and the caution. The glowing veil of my first image survives too, as pure cosmetics. The door can display its own boundary as light, so humans know where the border is. The border is photons for the eye, air and matter for the fly.

A Bouncer, Not a Prison

Now the doctrine, because a barrier that chooses needs rules more than it needs watts. First rule: fail soft. The film yields to any human-scale force, always. My clumsy fall through it is a legal requirement, not a design flaw. No fire code on Earth allows an exit that holds you in. After a power cut, a sensor crash, or a dead battery, it degrades into what my grandparents had, a curtain. The first wall in history whose failure mode is hospitality.

Second rule, and it is the heavier one: a door that recognizes is a biometric camera aimed at your threshold, all day, forever. It knows who came home, when, with whose suitcase. The anti-fly curtain is one firmware update away from being the household’s ledger. So the doctrine borrows from the way phones guard faces. Everything stays on the chip, nothing is stored, nothing is transmitted, and the guest list lives inside the doorframe the way your face lives inside your phone’s secure enclave. The door watches, and the door forgets. That sentence belongs in the product’s constitution, not its marketing.

The Screen’s Last Form

I want to close the loop on why this is more than comfort. The humble window screen is arguably one of the cheapest public-health technologies ever deployed, and its job is getting harder. Malaria still kills roughly 600,000 people a year by the WHO’s count, dengue is expanding, and the tiger mosquito is walking its way north through Europe with the climate as an accomplice. A doorway that lets the summer in and keeps its escort out is a quality-of-life toy in my grandparents’ kitchen and a public-health instrument two latitudes south. The film is where the screen was always heading, and it arrives just when the screen’s workload is growing.

Here is one fragment you could build this season, in the spirit of this site. A doorframe bar with an insect sensor and an aimed puff of air, a fan with a guest list, no membrane required. It would not stop the snake. It would settle, cheaply, the only question that matters. How much of the summer delegation can eyes and air turn away before anyone needs the science fiction? If you build it, tell me.

So here is what I keep turning over. Once a house can choose its guests down to the species, where do we stop writing the list, and who in the household gets to hold the pen?

The Devil's Trill Problem

Mon, 31 Aug 2026 00:00:00 GMT

Some mornings I wake up having composed something extraordinary. I mean it. In the dream there was a full piece, arranged, moving, better than anything I have written with my hands, and for a few seconds after waking it is still playing. Then it does what dream music does. It thins, it slips, and by the time I have found something to hum into, there is a vague contour left and a very precise feeling of having been robbed. Not by anyone. By the distance between my ears and my hands.

The daytime version is quieter but just as maddening. As a musician (an amateur one, several instruments, none of them well), I hear ideas in my head all day, and they fail in two ways. Either they are too fleeting to grab, gone before I have found a keyboard. Or they are too complex. The moment I focus on one voice to write it down, the others collapse behind it, like turning to look at one bird and losing the flock.

The worst part is not forgetting. It is remembering that it was good.

So here is the device, the fifth climb up this site’s science fiction ladder (after the storm, the gun, the collar and the air), and the third about music, counting the volume knob. It is a recorder for the inner ear. Something that captures the music you imagine, as you imagine it, whole, without passing through your mouth or your fingers. Not a dream recorder. Dreams are another essay, with a thousand years of literature attached. Just this one narrow, ancient, ridiculous need: to hear a thing in your head and keep it.

Three Centuries of the Same Morning

Ancient is the right word, because the problem has a founding document. In 1713, the violinist Giuseppe Tartini dreamed that the devil sat at the foot of his bed and played him a sonata of unbearable beauty. He woke, grabbed his violin, and wrote down what he could. The result is the “Devil’s Trill,” one of the most demanding pieces in the repertoire, and Tartini said for the rest of his life that it was infinitely inferior to what he had heard. He had my morning, in 1713, and he had it with better material.

He has company. Paul McCartney woke up with the melody of “Yesterday” complete in his head, went to the piano to trap it, and then spent weeks asking everyone he knew whether it already belonged to someone else, because nothing that arrives that finished feels like yours. And the most instructive case is Keith Richards, in 1965, who caught the riff of “Satisfaction” half-asleep on a bedside cassette recorder. In the morning the tape held about two minutes of the riff, followed by forty minutes of snoring. The bedside recorder is the ancestor of the device I want. It worked exactly once in the history of music, and it is still the best tool we have.

The losses have their monument too. Coleridge composed “Kubla Khan” in a dream, woke, began writing it down, and was interrupted by a visitor from Porlock. When he sat back down the rest was gone, and the poem has been published ever since as a fragment. Every musician I know has a person from Porlock. Mine is usually an alarm clock.

The Bottleneck Is the Mouth

Why does humming fail? Because the inner ear does not hold music as a sequence. It holds it as a shape. The chords, the bass moving under them, the drums, and the line on top are all present at once, the way you hold a face. But the only exits available are serial. A mouth can carry one line at a time. A hand on a keyboard can carry a few, slowly. And serializing the shape destroys it. By the time you have hummed the melody, the harmony that made it worth humming has evaporated, and you are left with a tune that sounds, correctly, like a thinner version of a thing you can no longer compare it to.

Today’s tools inherit the bottleneck. Voice-to-MIDI and hum-to-notation apps exist and work well, one monophonic line at a time, which is exactly the constraint the problem is made of. A phone’s voice memo is the Richards method with better tape. None of them touch the actual fault, which is that a rich musical thought cannot survive being pushed through a hole the width of one note.

This is the deep reason the device deserves to exist rather than a better dictaphone. Reading the inner ear directly would read it in parallel, the whole flock rather than one bird at a time, and parallel is the only way the shape survives the trip.

What the Science Just Did

What I did not expect when I started checking was that the first rung of this ladder had been climbed weeks before I wrote this. A study published in 2026 in eNeuro decoded imagined melodies of familiar children’s songs from ECoG recordings in ten epilepsy patients, by representing melody as relative pitch (do, re, mi rather than absolute frequencies), and the sequence of decoded notes reconstructed the contour the patients were imagining. Single-note accuracy was modest but reliably above chance. Imagined music, read from a brain, as a contour on a screen.

The non-invasive side has been creeping up too. When you imagine music, your brain produces a response that follows the amplitude envelope of the imagined sound, and researchers have decoded that envelope from ordinary scalp EEG well enough to recognize which song a person is imagining. Another team asked trained musicians to imagine four Bach melodies and decoded them, bar by bar, from EEG alone. And on the perception side, a 2023 study reconstructed a recognizable Pink Floyd song straight from patients’ cortical recordings, while others have rebuilt heard music from brain scans by pairing them with generative audio models.

Now read the small words in those sentences: familiar, bar by bar, contour, recognize. Every one of these systems works on music that already exists. The patients imagine the continuation of a children’s song they know. The musicians imagine Bach they have learned. The decoders have a template to match against. A melody that arrived in your sleep last night has no template anywhere in the world. That is the distance left to climb.

So the device is a stack of four leaps, each one real on its own, fictional only together. The first goes from known to novel, decoding music nobody has heard yet, with no reference recording to lean on. The second goes from one voice to many. Today’s proofs of concept follow a single melodic line, but my dreams arrive arranged. The third goes from electrodes on the cortex to something you can wear. The good results come from surgery patients, scalp EEG is coarse, and a brain scanner is not a bedside object. The last goes from waking to the edge of sleep, where the best material lives and the signal is worst.

I want to be exact about which rung is science fiction, because the answer is unusually flattering to the dream. None of them is, individually. Novel-content decoding is where imagined-speech research is already heading. Polyphony is a bandwidth problem, not a physics problem. Wearable neural sensing is an entire industry’s roadmap. The fiction is only the stack, worn on one head, at 6am. Most real devices were fiction in the same way, right up until they were not.

The Edge, Not the Dream

Where would you wear it? Not, I think, all night. Dream recording is someone else’s ambition, and a heavier one. What matters for music is the edge: the hypnagogic minutes falling asleep, when ideas arrive unbidden, and the hypnopompic minutes waking up, when the dream piece is still faintly playing and already leaving. Two anchors make that edge a real place rather than a poetic one. In 2021, a team at the Paris Brain Institute recreated Thomas Edison’s famous nap trick under laboratory conditions and showed, in Science Advances, that the very first stage of sleep is a measurable hotspot for creative insight. And the MIT Media Lab built Dormio, a glove that detects the moment you drift off and gently prompts you, capturing the thoughts of the edge before they dissolve.

My device borrows the glove and gives it an ear. It knows when you are waking before you do, and it opens the capture in that window, before the eyes, before the alarm, before the person from Porlock. Nothing to press, nothing to find, no forty minutes of snoring. The Richards method with a nervous system attached.

Was It Ever Notes?

And now the question I cannot answer, which is why this is an essay and not a product. The most impressive decoders today do not read music out of a brain so much as generate plausible music conditioned on it: a model’s best guess at what would produce those signals. Point one at a dream and it will hand you something confident, competent, and possibly not yours. Would you know? You would have nothing to compare it to except a feeling.

That raises the harder possibility. Maybe the dream piece was never notes. The sleeping brain manufactures the sensation of significance on demand. Everyone has woken from a dream certain they had understood something enormous, holding nothing. Perhaps the phenomenal symphony is, at least sometimes, the feeling of music rather than music, and Tartini’s lifelong disappointment was not a failure of memory but a rare honest glimpse of the gap. The device I want could confirm my nocturnal genius. It could also, on its very first morning, break the spell for good.

I would build it anyway. Here is one fragment you could build this afternoon, in the spirit of this site. A sleep-tracking wearable already knows when you wake, and a phone can open a one-tap recorder in that second, then walk you through the flock one bird at a time (“melody first; now the bass; now the drums”) before the shape is gone. It is the cassette recorder with a nervous system, and it would rescue more than you think. If you build it, tell me, and I will send you my first recording. It will probably be forty minutes of snoring.

And if Tutti ever gets its dark stage, this is the music that fills it: the phrase caught at 6am, shaped by the same hands that failed to hum it.

So here is what I keep turning over. If you could finally hear, in daylight, the music your sleeping mind is so proud of, would you want to know how good it really was?

Reading the Air

Mon, 31 Aug 2026 00:00:00 GMT

Winter is the only season honest enough to show its work. For a few cold months a year, the air stops pretending to be nothing: your breath hangs in front of you, the manhole covers steam like the city is exhaling, every warm thing (a car hood, a vent, a dog) grows a visible plume. Condensation does the trick, cold air giving up the water it was carrying, and for a moment you see the invisible traffic that was there all along. Every January I stand on some sidewalk watching a sewer grate breathe and think that I would like this all year, for everything.

Summer asks the same question from the other side. On the worst afternoons you walk out of a doorway into what feels like a solid block of hot air, a thing with edges and a weight, and you can feel it move around you without seeing a single line of it. I am not claiming a need here. It is one of those small, stupid, recurring questions of a life, the kind that comes back on a cold platform or a hot parking lot. What is the air actually doing right now, and why am I the only one in the scene not allowed to watch?

So here is the device, in the tradition of this site’s science fiction essays (I have caged a storm, made a police gun negotiate, and hung a soundtrack around a neck, so this is round four). A pair of glasses that shows you the air. Heat rising off your own hand, the draft from the window, the wave you push when you gesture, the wind combing a field. A new layer of data over the world. Not, as I first thought, a superpower. Something closer to a repair.

The Sense We Never Had

The better framing is that we are the blind ones. Harbor seals, in experiments, follow the hydrodynamic trail a fish left behind using nothing but their whiskers, long after the fish itself is gone. Spiders and crickets carry hairs tuned to the faintest air currents. A cat’s whiskers read the displacement of air before a thing arrives. Air-sensing is not exotic in nature. It is ordinary equipment. We simply did not get it.

What we got instead are proxies, and we have become experts at them: flags, smoke, hair, a field of grass bending in a wave, a curtain lifting. We never see air, only what it pushes. Nature grants two exceptions, and everyone has witnessed both. The shimmer above hot asphalt in August is a real optical effect, light refracting through air of uneven density, the sky performing physics for free. And dust in a sunbeam through a shutter is the one time in life you actually watch air move, particles riding currents you cannot otherwise perceive. Those two accidents are the whole of human air vision. My glasses just want to make them permanent.

A Mirror in 1864

When I finally looked it up, I got a surprise. The sense I was wishing for was invented in 1864. August Toepler developed schlieren photography, an optical arrangement that makes density differences in air visible: the heat off a hand, a breath, the plume over a candle, all rendered as rippling gray currents. In 1887 Ernst Mach used it to photograph the shockwave riding ahead of a bullet, the first time anyone saw supersonic air. For a century and a half, laboratories have watched exactly what I stand on sidewalks wishing to see.

The mirror was the technique’s prison, though. Classical schlieren needs high-quality parabolic mirrors or lenses, prohibitively expensive at any real size, and optical access straight through the flow. You saw the air, but only inside a window the size of a dinner plate, in a dark room, with the setup bolted down. The world’s most beautiful view of the invisible came with a porthole.

And the porthole had its year of fame. In 2020, schlieren footage of coughs and breath went around the world, and for a few months an entire planet wanted these glasses for the worst possible reason, trying to see the shared air of a room. The dream I carry idly in January became, briefly, a public health obsession. Then the footage faded and the mirror went back to the lab.

The Mirror Escapes

Except it did not stay there. In the early 2000s, a method called background-oriented schlieren changed the rules. A camera views a textured background on the far side of the flow, and software computes the tiny shifts in that pattern caused by density gradients. No mirrors, no optical through-access, and a field of view limited only by the camera and the size of the background. NASA has since used it to image the shockwaves of one aircraft from another, with the desert floor itself as the textured background. The dinner plate became a landscape.

Then it went real-time and cheap. NASA lists real-time BOS software in its catalog and holds a 2023 patent on a reference-free version. And in 2022 a demonstration made the rounds of the internet showing a hairdryer’s airflow visualized with nothing but an Android phone’s camera and a noise pattern on a screen behind it. The crude version of my glasses is already an app. What the last twenty years did was quietly walk this sense out of the laboratory and up to the doorstep of the eye.

Three Eyes

Now the physics, and the dream deserves to hear it straight. Schlieren, in every form, sees density, not motion. Warm, cold, compressed, or chemically different air shows up. Your hand leaves a heat wake, your breath draws a plume, the window bleeds a cold curtain, and the fridge, opened in winter, pours a ghost onto the floor. But the pure wave of room-temperature air you push with a gesture, the very image that started this essay, is nearly invisible even to Toepler, because it changes nothing about the air but its speed.

So the glasses need three eyes, not one. A density eye, running background-oriented schlieren against the textured world. A tracer eye, because motion becomes visible the moment something rides it. Wind tunnels use smoke, laboratories seed particles, and the air around you is already full of dust, the sunbeam trick generalized. And a Doppler eye, borrowed from the wind lidars that exist today, mapping wind for wind farms and spotting aircraft wake vortices at airports by the frequency shift of light off aerosols. Shrinking that onto a face is the science fiction rung, and I am standing on it openly.

The second wall is the background. Schlieren needs texture behind the air, and the sky has none, which makes the wind over a field against an empty sky the hardest case of all. The field itself moves, which is precisely the proxy we already use, and a moving background corrupts the computation. Your own head moving is the one part that is already solved. Augmented reality headsets know their pose to the millimeter and can subtract it. So the sober map reads: indoors, nearby, thermal, plausible near future. A whole windy landscape against blue is the leap, assumed.

What the Air Would Show

Suppose it works. The first surprise would be the noise. Every radiator, every lamp, every body in a room is a plume, every mouth is a chimney, and a dinner table would look like a pot at a rolling boil. The invisible is not empty, it is crowded, and our blindness to it may be less a deficiency than a mercy of attention. I suspect the first firmware update would be a dimmer.

Then the social layer, and here the title of this essay comes from Japan. Kuuki wo yomu, “reading the air,” is the Japanese expression for social sensitivity, for sensing the mood of a room and when to stop talking. These glasses would do it literally. You would see someone’s breathing quicken across a table, a whisper leaving lips as a small warm cloud, the heat coming off a person who has decided not to show it. Some of that is intimacy. Some of it is surveillance with a soft name, and I do not know which side the technology would land on, only that 2020 already taught us how quickly seeing shared air turns into fearing shared rooms.

I would want to see the sky. That is the part of this I trust.

And then the sublime, which is why the dream survives the objections. The thermals that hawks ride, visible as slow columns. The sky revealed as an ocean of convection rather than a blue ceiling. The wake of a passing bus trailing behind it like a train. Artists already know how badly we want this. Ned Kahn’s building façades, thousands of small hinged panels shivering in the wind, turn entire walls into instruments for seeing air, and people stop on the sidewalk to watch. Nobody stops to watch a wall. They stop to watch the wind.

That points to the true shape of the thing. It is not a permanent new layer over the world but a mode, the way night vision is a mode. Something you switch on in January to watch the city exhale, or in August to finally see the block of heat you are walking through, and then switch off to have dinner without a boiling table.

Here is one fragment you could build this afternoon, in the spirit of this site. Tape a printed speckle pattern to a wall, point a phone at it, and run the background-oriented schlieren code that already exists. You will watch your own hand leave heat in the air of your kitchen, and you will understand the rest of this essay in about four seconds. If you build the glasses themselves, tell me. I will test them in January, on a manhole cover.

So here is what I keep turning over. If the air around us became visible tomorrow, crowded and breathing and full of everyone’s warmth, would we finally read the room better, or would we learn, very quickly, to look away?

Tutti: Conduct the Expression, Not the Notes

Mon, 31 Aug 2026 00:00:00 GMT

I have been a music software nomad for twenty years. Cubase first, in a studio, on versions that now belong in a museum. Then the strange ones: Reaktor, and Reason, where you flipped the virtual rack around and patched cables by hand as if the computer were a wall of hardware. Then years on Fruity Loops, where I made more music than on anything else. Today I live in Logic Pro. Five or six homes, two decades, and one piece of furniture followed me into every single one of them.

The automation lane. A thin line under the notes that you draw with a mouse to tell the violins how to feel. Louder here, softer there, warmer on this phrase, held back on that one. Every orchestral piece I have ever finished was finished by drawing that line, section by section, one thin curve at a time, hunched toward a screen, translating an emotion into pixels. It works. For people with an academic temperament it is even satisfying. For the visual, physical kind of musician I happen to be, it always felt like the wrong direction of service. I was the one adapting to the machine, learning its dialect, so it would let me say something simple.

Every program I have loved asked the same question with the same thin line: how loud should the violins feel here? And every one expected the answer in pixels.

Tutti is the answer I have wanted every time I put on an orchestral record and closed my eyes. A dark stage. The orchestra laid out the way it sits in a real hall, first violins to your left, cellos to your right, brass behind, percussion at the back. Each section is a pool of soft, dark color, breathing with the music as it plays. And the score, already written, flows past you like a river. Your hands do not conduct it. The notes are done, the tempo is decided. Your hands do what the mouse used to do. They shape how it feels. Raise a palm toward the cellos and their pool deepens and swells, and the phrase warms up. Close a hand slowly over the brass and they hold back. Look at the woodwinds, point, and they are yours to sculpt. A finished piece, and you walking through it, adding the emotion with your body.

The stage nobody built

The hands part exists, and it is thriving. Gesture-to-music controllers are a real, active category: webcam tools built by composers who wanted expression to feel closer to conducting than to faders, hand-tracking rigs from major instrument makers, gloves used on stage by well-known artists, a whole lineage going back a decade. I checked all of them, because I assumed someone had built my stage. Nobody has. Every one of these tools hands your gesture to one knob on one track. They are performance tools, played once, forward. There is no room. There is no orchestra in the dark. There is no going back to the phrase you just ruined and doing it again.

That gap is the product. Tutti is a place, not a controller. Spatial mixing already has such places (the tools for immersive cinema sound show you a virtual room and let you move sounds around inside it), and the color side has a hundred years of pedigree, from the abstract animators of the 1930s to Disney’s Fantasia, where an orchestra became light. Composers already read music as shapes on a screen more fluently than as dots on a staff. Tutti just gives those shapes a stage, a third dimension, and your hands.

Key Features

The stage. A dark hall with the real seating plan of an orchestra, each section a slow-breathing pool of color, brighter and deeper as the music grows, dim as it recedes. The plumbing that turns gestures into instrument data lives under the floorboards, where you never have to see it.
Look, point, shape. Choose a section with your gaze or a finger, then shape its feeling with the hand: rise and sink, tense and release, warm and cool. Broad emotional strokes, not knob-twisting.
The river. The piece flows past; grab it and pull it back. Redo a phrase, layer a second pass on top of the first, compare, keep the one that moved you. Editing, not just playing.
Listen mode. Load any piece and simply watch the stage breathe, section by section. The same room that lets you shape a score also lets you feel one.
Fits your studio. Tutti is a mode you step into for the expression pass, the way you step into a mixing room. It talks to Logic, and to anything that will take MIDI or automation data, and hands the result back as ordinary data your project already understands.

The honest tensions

Tutti does not replace the thin line, but it changes who draws it first. The hands own the phrase, the mouse keeps the surgical corrections, and that division is right, not a compromise. It has plumbing to do that nobody will find glamorous. Every maker of orchestral sounds speaks a slightly different dialect of expression, and Tutti has to translate all of them into one stage. And listen mode is honest about its limits. It works beautifully on your own projects, where it knows every section, and only roughly on a recording of a real orchestra, where separating the cellos from the violas is still research.

One objection I want to dismiss, because I have heard it about every gesture tool, is that your arms will tire. They would, if you were drawing precision curves in the air for a shift. But a piece of music lasts three to ten minutes, and what you are doing in Tutti is feeling, not precision. It is broad, slow, whole-arm honesty about how a phrase should land. Nobody gets tired feeling a symphony. They get tired clicking.

Monetization

A subscription for the app and its studio plugin, priced like the serious creative tools composers already pay for. The more interesting money is upstream. The companies that sell orchestral sound libraries live and die on expressiveness, and a stage designed to make their instruments feel alive is a partnership waiting to be signed, bundled or licensed. Conservatories and film-scoring programs are a natural third door, since a stage that makes dynamics visible is a way to teach what they are for.

Next Steps: How I’d Build It

Not a weekend, but not a laboratory either: a consumer headset with good hand tracking, one section, one pool, one axis of feeling, synced to Logic. If shaping a single string phrase with a palm produces the emotion you meant, the rest is furniture. Then the full seating plan, the river, and listen mode, which comes almost free once the stage exists. A season of work for one developer who knows 3D and one composer who is tired of the thin line. I know at least one of those.

Tutti, danced

And now the paragraph I could not resist. If a hand can shape how the cellos feel, a body can shape the whole hall. Picture the stage watching not two hands but a person, reading the entire body in real time through a camera or a suit. Now bring in a dancer. Not to illustrate the music, which is what dance on stage has always been asked to do, but to write its expression: every reach becomes a swell, every collapse a hush, every held stillness a breath the orchestra takes with her. The two oldest human arts, movement and music, exchanging data for the first time in both directions.

Dance predates every notation we have. Bodies were shaping music before anyone wrote a note down. Tutti, danced, would record that shaping as something a score can keep. A choreographer could interpret Mahler with her spine and leave behind an expression layer another orchestra could play. I do not know what that does to how we hear music. I suspect it is a lot.

Closing

Tutti is the stage I have built in my head every time I listened to an orchestra with my eyes closed, finally with the lights on. Twenty years of drawing feelings with a mouse, and a room where you can just make them with your hands. And if the Devil’s Trill recorder ever gets built, this is the room it empties into, where the music caught at 6am is shaped by the same hands that failed to hum it.

If you build this, tell me, and let me do the first expression pass. I will be the composer in the corner, palms open toward the cellos, making very small movements and very large faces.

The Gun That Asks First

Fri, 28 Aug 2026 00:00:00 GMT

I have been bingeing those police ride-along documentaries lately. Night shifts, dispatch radios, the works. And once you watch enough of them, a pattern emerges. Almost every disaster is a speed problem. A scooter that will not stop. A car that accelerates instead of pulling over. A man on a ledge. The officers have a toolbox with a talking end and a shooting end, and the gap between the two is enormous, yet it gets crossed in seconds.

The chase is the purest version of the dilemma. Pursue, and you endanger everyone on the road. Let go, and you teach every future driver that flooring it works. So officers improvise in the middle, and the middle is where people get hurt.

Watching all this, the device I kept imagining was a launcher that fires a bubble. It stops the vehicle, wraps the person, and the person keeps breathing, unhurt, annoyed, alive. My working name is the pistobulle: half pistol, half bubble, fully ridiculous. In my last essay, A Storm in a Cage, I trapped a wildfire and made it brew its own rain. I am playing the same game here. Take the science fiction object, then climb the ladder from what exists to what does not, one honest rung at a time.

The Arsenal Nobody Assembled

The strange part is that most of the pieces already exist, scattered across police catalogs. The BolaWrap is a real handheld device that fires a Kevlar tether that wraps around a person’s legs. StarChase launches a GPS tag onto a fleeing car so nobody has to chase it. The Grappler snags a rear wheel with a strap. And the British RF Safe-Stop demonstrated over a decade ago that a radio-frequency pulse can stall an engine at a distance, combustion engines included, because a modern car is a computer on wheels. The pistobulle does not invent a category. It merges one.

The most important piece, though, is the one that was abandoned. In the 1990s, Sandia National Laboratories developed an actual sticky foam weapon, and US Marines carried it to Somalia in 1995. It never became standard issue, for a simple, terrible reason: if the foam covers the face, the person suffocates.

Everything in this essay follows from that sentence. The technology worked, but the safety did not. The bubble is a direct answer to the flaw that killed the foam, not a gadget bolted onto it. It is a two-layer capsule, with an outer shell that grips and hardens, and an inner pocket that stays open and lets oxygen through, the way waterproof-breathable membranes already pass air one way and block water the other. Sticky foam with a conscience.

One Trigger, Six Machines

Stage zero of the weapon fires no projectile at all. It negotiates. Every new European car model approved since 2018 has carried a mandatory eCall emergency modem, intelligent speed assistance has been required since 2022, and manufacturers push remote updates monthly. So the first phase, the one that does not exist yet, is a digital siren, a cryptographically signed stop order that compliant vehicles are required to honor. The munition exists only for whatever refuses, or cannot answer.

Then comes the scope, which is the boring part. Computer vision identifies the vehicle model, estimates mass and speed, and computes the takedown sequence. Cars already recognize other cars all day long. The line that matters is doctrinal, borrowed from the autonomous weapons debates at the UN: the machine choreographs, the human decides. The gun is smart everywhere except the one place that matters.

The cartridge itself is a carrier round that splits in flight, each stage programmed as it fires, the way the XM25 grenade launcher programmed its airburst rounds in the chamber. The deeper ancestor is the centuries-old firework, a shell that becomes a choreography.

And the sequence is not decoration, it is survival. A car whose electronics die at 130 km/h is not stopped. It is a ton and a half of projectile with no power steering and no brake assist. That is why the doctrine echoes the storm cage. Slow first, cut second, encapsulate last. Foam to the wheels, then the pulse, then the bubble. If millisecond delegation sounds dystopian, look at your dashboard. An airbag decides in roughly thirty milliseconds to fire an explosive at your face, and civilization calls it a safety feature.

Now the sacrifice. There are no nanobots in this gun, because there are no nanobots anywhere. Real nanotechnology in 2026 is self-assembling chemistry, molecules folded into designed shapes, DNA origami rather than tiny machines. The bubble’s skin is four chemistries stacked, not a swarm. And the loveliest brick is real. There is a whole research field of CO2-responsive polymers. The inner layer stays soft and permeable wherever it senses warm, humid, CO2-rich air, and hardens everywhere else. The bubble never needs to see your face. Your breathing carves its own cave.

Admitting two limits keeps the license. The first is falls. What kills is deceleration, and fire brigade rescue cushions already save jumpers but top out at a few stories. A bubble of shear-thickening material, soft at rest and rigid on impact like motorcycle armor, adds maybe a meter of crush distance. From a bridge it changes nothing. From three floors it turns a funeral into fractures. It widens the survival window, but it does not negotiate with gravity.

The second is release, because a restraint that never opens is a coffin. So the shell borrows from resorbable sutures and stents, polymers whose death is chemically scheduled, plus cheap phone-grade sensors watching the occupant’s breathing, plus a solvent on the officer’s belt. The bubble is lethal by default, except it is designed to die first.

Who Holds the Keys

The digital siren is also the nightmare, and I will not pretend otherwise. A weapon with a projectile has a range. A weapon that is an infrastructure has a jurisdiction. One stolen master key and every car in the country can be stopped, which is why the American Clipper chip, the 1990s attempt at government key escrow, died in disgrace. The fix is to split the keys. The magistrate’s key is universal and non-negotiable, granted on judicial requisition like a wiretap, never by a lone officer at a roadside.

The second lock is stranger. It is the insurer. French police can already query the national insured-vehicles file from a plate number, and have since 2019. The pipe exists, and this essay just makes it carry cryptography. Insurance stops being an obligation you dodge and becomes armor you want, a second guardian that must also say yes before anyone touches your car. The incentive to insure arrives from above, as protection, not from below, as a stripping of rights from the poorest drivers, who are precisely the ones most often uninsured.

Somewhere in this system, an insurance adjuster holds a veto over a police officer. That is the most cyberpunk sentence I have ever typed.

Speed is the obvious objection, and the anchors answer it. On-call magistrates already exist around the clock for urgent wiretaps, and the ICANN root key ceremonies prove that multi-person cryptographic signing can be a solemn, auditable ritual. Compress the ritual and ninety seconds is enough. France already spawns specialized benches when cases get technical, a financial prosecutor’s office in 2013, an antiterrorist one in 2019, so picture the next one, a small standing bureau whose only docket is remote coercion.

One documented warning belongs taped to its door: America’s specialized surveillance court became famous for approving nearly everything it saw. A court that hears only one kind of case learns to stamp. The guardrail needs its own guardrail: short rotations, published refusal rates, outside audit.

Notice the symmetry the system settles into. The infinite-range weapon, the siren, is braked by bureaucracy. The single-shot weapon, the bubble, is braked by price. And one more admission. The technology of the siren can be copied abroad in six months, while the on-call magistrate and the split key do not export to a regime that does not want them. The cage travels worse than the animal. That one might be my next essay.

The Accountant and the Magistrate

The deepest risk is not the stolen key. It is success. The taser was sold as an alternative to the firearm and became, in documented practice, something closer to an alternative to conversation. The industry’s own vocabulary confesses the drift. Nobody serious says non-lethal anymore, only less-lethal. The safer a weapon feels, the more it gets used. A perfect pistobulle would be fired a hundred times a day, at the fleeing, at the protesting, at the merely annoying.

Restraint itself has a body count. Positional asphyxia during arrests is a documented killer, which is why the bubble monitors breathing and dissolves on schedule. The bubble has to be more careful than the officer who fired it. The manual should say it in those words.

And I have grown fond of a cynical guardrail. The price is a feature. A cartridge that costs as much as a small car will never be fired at an annoying teenager, because every shot has to survive a meeting with an accountant. The magistrate guards the siren; the accountant guards the trigger. The accountant is the ethics committee.

One fragment you could build now, in the spirit of this site, is breathable sticky foam. Resurrect Sandia’s abandoned weapon and fix its one mortal flaw with a permeable inner phase. No treaty required, no cryptography, just a materials lab and some nerve. If you start there, tell me.

So here is what I keep turning over. The day the gentlest possible weapon exists, cheap enough to use and polite enough to ask first, who counts how often it asks, and what do we stop bothering to learn to say to each other?

The Screen That Wears Your Glasses

Wed, 26 Aug 2026 00:00:00 GMT

At home, I like to exist without my glasses. Every pair I have ever owned slides, presses, and doubles as a small radiator in summer, and since my myopia only really punishes distance, home is the one place I can afford bare eyes. Everything within arm’s reach is sharp, life is lighter, and I even have one eye noticeably better than the other covering for me. There is just one casualty in this arrangement: the TV across the room, reduced to a warm colorful smear that occasionally says things.

So for years I carried a lazy little wish: forget correcting my eyes, correct the screen. An overlay matched to my prescription, sitting on the TV, so the one distant object I actually care about at home meets me halfway. I assumed this was optical nonsense. Then I finally looked it up, and it turns out I had been daydreaming, from my couch, a real research project. In 2014, teams at UC Berkeley and the MIT Media Lab built exactly this. A pinhole-punched filter laid over a display, an algorithm fed with your eyeglass prescription, each pixel’s light reshaped so the image lands sharp on your imperfect retina. They prototyped it on an iPod Touch.

The press wrote “Forget glasses” and “Goodbye, glasses.” One of the researchers described the dream better than I ever did: a screen that wears the glasses for you.

Somewhere in a lab drawer, there is an iPod Touch that already solved my evenings, twelve years ago.

And then nothing. No product, ever. The reasons are honest ones. The trick only works from a fixed viewing position, and the moment you move, the image falls apart. It serves only one pair of eyes at a time. It can switch prescriptions, but it cannot correct two viewers at once. And it competes with a rival that is very hard to disrupt: a 700-year-old product that costs twenty euros and corrects the entire room, not just one rectangle of it.

But here is where my shower disagrees with the market analysis, on two counts. First, “single viewer only” is a flaw for a family television and a specification for everyone else. Eurostat counts roughly a third of EU households as one person living alone. Those are the students, the widowed, the happily single. Every one of them owns a TV that will only ever be watched by a single prescription. The flaw is a market, and it is enormous. Second, “it breaks with two viewers” was solved by Nintendo in 2011. The 3DS used an electronically switchable barrier over its screen, with one slider to turn the light-bending on and off. Guest mode is that slider on your remote.

Someone comes over, one touch, the screen goes ordinary, you reach for the glasses on the coffee table, and nobody has suffered.

The movement problem is the real rung, so let me be precise about where it sits. Half of it already ships. Glasses-free 3D monitors exist today that eye-track you and steer their image toward your head in real time, which is the same mechanism a prescription screen would need to survive you shifting on the couch. A tracked, seated, solo viewer is plausible near-future. But the version I actually daydream about is the one where the TV stays sharp while I orbit the living room with a vacuum cleaner, across every distance and angle at once. That is the science fiction rung, and I suspect it is the quiet reason this product does not exist. Its only perfect customer is a person sitting still, alone, at a known distance, and that person’s glasses are already within arm’s reach. Physics did not kill the idea. Convenience did, by a nose.

Still. A third of households, one slider borrowed from a Nintendo handheld, and a research prototype that has been waiting in a drawer since 2014. If someone ever ships the couch edition, tell me. I will be the market of one it was designed for, sitting perfectly still, bare-eyed, watching the smear become a television again.

The Soundtrack Collar

Wed, 26 Aug 2026 00:00:00 GMT

I grew up leaving cinemas the wrong way. As a kid I started music at about the age I was falling into films, and the two fused for good. I would walk out of the old Dune (the one Toto scored), out of American Pie, out of School of Rock, and the world outside the screen felt underproduced. The scenes were all still there, the corridor walk, the small victory, the crush at the counter. Only the cues were missing. Those films knew something I have never unlearned. The music was doing half the feeling, and my life had the same scenes with the sound turned off.

The adult version of me never stopped trying to fix it with consumer electronics. I am rarely without music, and I run a small logistics operation to keep it that way: the big cans for vacuuming, the sport buds for running, the little ones for the supermarket, each with its own charge level and its own pocket. And I am my own music supervisor, permanently one scene behind, because a playlist is a guess made in advance about moments that refuse to happen on schedule. The right track for the moment you are in is always one manual skip away.

I have carried the fix in my head for about fifteen years. In A Storm in a Cage I trapped a wildfire and made it brew its own rain. In The Gun That Asks First I made the police negotiate with a getaway car. This is the third round of the same game. Take the science fiction object, then climb the ladder from what exists to what does not, one honest rung at a time.

The Ghost and the Graveyard

This dream is old enough to have a ghost. Around 2010, an app called RjDj made “reactive music” on the iPhone, with the sensors and microphone feeding a score that recomposed itself around you. They even built the official Inception app, where Hans Zimmer’s music bent to your walk, your weather, your hour of the night. It was glorious, it was fifteen years too early, and it died with the company. Spotify’s 2015 Running feature belongs in the same graveyard. It matched the tempo to your stride through the phone’s sensors, and it was quietly killed in 2018. (A 2026 reboot exists. It asks you to pick a BPM in advance and does not listen to your body.)

The survivors are instructive. Video games have shipped adaptive music for three decades. LucasArts filed its iMUSE patent in 1991, and the 2016 reboot of DOOM answered a question I care about. Its celebrated metal soundtrack rises and falls in layers with the intensity of combat, so adaptive does not have to mean orchestral. Metal bends, skate punk bends, strings and horror shrieks bend. The genre your life is scored in is a catalog question, not a physics question. And Endel exists today, a real company generating ambient soundscapes from heart rate, motion and time of day.

As for 2026, demos at this year’s CES paired real-time EEG and biometric sensing with audio that responds to your state, and trend reports place adaptive audio in cars, hearables and wellness gear. As of this writing you can buy headphones that read your brainwaves, but none that adapt your music to them. The reading has shipped, but the responding has not. Everyone is building the bricks, but nobody has assembled the object. This is the moment to describe it carefully.

The Collar and the Lens

The object is an open-eared collar, with bone conduction or near-field speakers, both on shelves today. Everything else follows from the open ear, and film theory has a word for it. A movie’s score is non-diegetic. It floats above the scene without belonging to it. Earbuds seal you out of your own scene, while the collar layers a score on top of it. That is the difference between escaping your day and scoring it.

The open ear also sets the safety doctrine, and the doctrine speaks fluent mixing desk. The world is the dialogue track, and its fader is welded above yours. For sirens, alarms or a voice addressing you, the music ducks or cuts, with no override and a hard ceiling on volume. None of this is exotic. Earbuds already duck your music for conversation, European law already caps personal audio levels, and self-driving cars already listen for sirens. A fire in view cuts the score entirely.

Now the sensor problem, because the collar alone is half deaf. A heart rate tells you how much but never what. 140 BPM could be a sprint, a job interview, or a first kiss, three scenes that need three incompatible cues.

The only sensor that points at meaning is the gaze, so the second half of the instrument is a contact lens that watches what you watch. The tracking half is mature, since VR headsets steer rendering by your pupil all day. The lens half has a mixed record. The display lens died (in 2023 a famous AR lens startup shelved its lens and pivoted to selling microdisplays). The measuring lenses split. A glucose-sensing lens was abandoned in 2018, because tear glucose turned out to be a poor proxy for blood sugar, but a pressure-sensing lens for glaucoma earned actual regulatory approval. Lenses that display have ghosts, while lenses that measure have a precedent. Ours only measures.

And the lens pays twice. Pupillometry is a decades-old laboratory tool. Your pupil dilates with arousal and cognitive load, involuntarily and continuously, so the same lens that identifies the scene also reads your reaction to it. The collar hears your day, and the lens finally knows which movie you are in.

Rooms, Corridors, and the Body’s Vote

Then the music itself, and the wall nobody talks about. Your favorite songs do not bend. They are fixed recordings, and the adaptive versions a live score needs (stems, layers, intensities) are licenses the labels have never sold. That is why Endel generates its own audio instead of adapting yours.

Two engines get around the wall. The first is the pack economy, and one profession is already trained for it. Game composers have spent thirty years writing music designed to be bent by events they cannot predict. Picture an action pack, a wedding pack, a new year pack, each one a commissioned adaptive score delivered in layers, DLC for your day, in whatever genre your life runs on. The second engine composes the gaps. It is a personal generative engine trained on the shape of your library, writing the transitions and in-betweens in your own musical language so the score never jump-cuts.

Your songs are the rooms. The engine builds the corridors.

The engine never replaces the rooms. It is connective tissue, the craft DJs have always practiced by hand and streaming apps already fake with automatic blends.

That leaves the question I struggled with longest. How does the system learn which music fit which scene, when asking is forbidden? Nobody thumbs-up mid-kiss. Your body already votes, and until now the ballots went uncounted. Musical chills are physically measurable. Goosebumps are visible, skin conductance spikes at the exact second of the frisson, and a celebrated 2011 neuroimaging study caught dopamine releasing at musical peak moments. Hearts and breath entrain to tempo.

Pupil, skin, chills and entrainment add up to a coherence score, collected passively and asked directly of the nervous system. Did this music belong to this scene? The auto mode learns from every answer and fills tomorrow’s gaps a little better. This is the science fiction hinge of the essay, and I lean on it as hard as it deserves, and no harder. Every signal in the loop is real. The loop itself, worn on a neck all day, is the leap.

It also carries the essay’s darkest tension, so look at it in the light. A system that optimizes the measured chill is a dopamine slot machine, the feed-ification of your own soundtrack. The doctrine has to be one line long: optimize coherence, not chills. A score serves the scene. A drug serves itself.

The Duet

One scene, two people, should mean one cue. So the collar has a link mode. You can offer your soundtrack to the person beside you and they can offer theirs, by explicit consent, revocable with a touch. Neither one dominates, because the engines negotiate a middle in real time. The plumbing half exists. Phones already share one stream to two pairs of ears, and streaming services already compute blended tastes for two people. The negotiation half is ours.

Couples get “our song” by accident, maybe one per lifetime, minted by a radio coincidence. The collar mints them live. The cue that played while you both watched the same storm come in now belongs to two nervous systems. And the moment either of you revokes, the duet dissolves politely back into two solitudes. Nobody’s day gets annexed.

The Right to Silence

The loop has a thermostat problem, and I will not hide it. The sensors read your state, the music changes your state, the new state changes the music. The collar is a thermostat, not a mirror, and somebody sets the target temperature. It had better be you, in settings you actually understand, and not a mood curve optimized by whoever sold the pack.

There is a precedent for the accusation this device will face. Within a few years of its 1979 launch, the Walkman triggered a moral panic. The charge was that it would privatize public space, with everyone walking around scoring their own life. The collar is the Walkman taken all the way, and the old charge deserves a straight answer. Yes, that is what it is, without the sealed ears, with the world welded on top, and with consent at every layer.

Last comes the feature I now believe matters most. Ask film composers about their most powerful cue and many will give the same answer: the cut to silence. A score that never stops is wallpaper. So the collar’s finest skill is knowing when to shut up. The same sensors can detect the scenes too large to touch, and the collar steps back so the moment stays entirely yours. The machine’s real mastery lies less in picking the perfect track than in recognizing the moments that deserve none.

Here is one fragment you could start this afternoon, in the spirit of this site. Your phone can already tell running from vacuuming from a walk to the store. A dumb DJ app that switches between three of your own playlists on that signal, and ducks hard for conversation, is a weekend of work. It would answer the first real question. Does a scored day feel like a film, or like a cage? If you build that, tell me.

So here is what I keep turning over. If every scene of your life arrived with the right cue, would you still trust what the scene made you feel on its own? And which moments would you fence off, unscored, so they stay entirely yours?

Decoy: A Patch Designed to Be Picked

Sun, 23 Aug 2026 00:00:00 GMT

I am one of those people. Give me a pimple, a cut, a scab, anywhere on my face or body, and I will religiously make it worse. Not once, not twice, but for weeks, sometimes months, until a small blemish has been promoted to a small crater.

Lockdown was the golden age of this hobby, endless hours at a desk with my face two feet from a screen and one hand on autopilot. I have never asked a doctor to put a name on it, but the literature has one anyway. Skin picking sits in a family of behaviors common enough that studies put it at somewhere between 1.4 and 5.4 percent of people, which makes it one of the most common skin-and-mind conditions there is, and one of the least treated.

The detail that always struck me is that the hand goes first. Watch anyone thinking hard: fingers on the chin, in the beard, twirling hair, tracing the same spot on a forearm. The touching is a background process, not a decision. That means any tool that wants to compete with it cannot live in a drawer. It has to live where the hand already goes.

Decoy is that tool. It is a small adhesive patch you place on the exact spot you cannot leave alone, built to be picked. Not a shield that asks your fingers to give up. A sacrifice that lets them win.

Two products exist. The gap is between them.

The market has already solved both halves of this, separately. On one side are picky pads, silicone pads full of beads and textures to dig out, a viral product category, explicitly sold as an alternative for skin pickers, with millions of TikTok views and little shops behind them. They are satisfying, and they sit on a desk. On the other side are hydrocolloid pimple patches, the K-beauty staple in every pharmacy. Medically sound, they protect the spot and absorb what needs absorbing. They work, and they are deliberately boring, a smooth dome your fingers slide off.

So the picker’s routine defeats both. The desk toy requires remembering to redirect a gesture that happens before memory is invited. And the boring patch gets picked anyway, except now picking means peeling off the protection, which is the one move worse than picking the skin. The two halves each hold one truth: the substitute must be satisfying, and the substitute must be on the wound. Nobody has merged them.

Peel me, not you

Decoy’s answer is a laminate with a job split. At the bottom sits a medical base layer that adheres and stays. It is hydrocolloid, the same cellulose-gel family as the dressings pharmacies have sold since the 1980s, because covering a wound is not a compromise. The old “let it breathe” advice is mostly folklore. Wound care flipped in 1962, when George Winter showed that wounds heal faster moist and covered than dry and open. The correct medical move and the anti-picking move are, conveniently, the same move.

On top of that base go the sacrificial layers. Thin, textured films designed to be found by a fingernail, with edges that catch just enough, each one peeling away with the exact resistance the ritual wants. The material family exists too. Soft silicone adhesives, the gentle-tack technology wound care developed for fragile skin, are sticky enough to satisfy and kind enough to re-peel without trauma. Search the edge, lift it, feel it give. The entire ceremony happens on the patch, layer after layer, while the skin underneath is left in peace to do the one thing it kept being interrupted from.

Peel me, not you.

The base layer stays put for days, hydrocolloid doing ordinary hydrocolloid work underneath everything.
Sacrificial films, a handful per patch, each one a full ritual: textured, edge-catching, re-peelable a few times before it gives up its ghost.
Formats for the real map of the habit: small dots for the face, strips for arms, and a “thinking patch” for the chin-and-beard zone that gets touched by every human who has ever pondered.
Refills as the business: the patch is a consumable by design, and that is the entire monetization section, because a product like this earns repeat purchases or it earned nothing.

The honest tensions

One tension towers over the others, and I want it in plain sight: nobody knows if this helps or rehearses. The harm-reduction reading says the ritual is satisfied, the skin is protected, everybody wins. The skeptical reading says the hand keeps learning the road to the face. Both are plausible. Decoy is a hypothesis, not a promise, and the first real batch should be tested with people who actually live this, not marketed to them.

Related and non-negotiable: this is not a treatment and will never claim to be. The serious version of this behavior has serious help, habit reversal training with actual professionals.

And the numbers say how needed that help is. In one study of 262 adults with the disorder, 87 percent had never received any treatment, while nearly nine in ten of those who had found it helpful. The honest disclaimer already exists in the market, too. Some picky pad makers state plainly on their own sites that the pads are coping tools, not a substitute for therapy or medical care. Decoy joins that tradition or it does not deserve to exist.

Next Steps: How I’d Build It

The other honesty is that this is not a weekend project, for once. A product that touches broken skin is a medical device, entry class but still, with biocompatibility testing, adhesive safety on damaged skin, and laminate manufacturing that no 3D printer in a garage will fake. So the staging writes itself. Version zero is cosmetic, a wearable picky sticker for intact skin, the thinking patch for chins and forearms, a consumer product with no regulatory mountain, shippable by one motivated person with a materials supplier. It tests the entire behavioral hypothesis, does the hand accept a decoy where the hand already goes, before a single euro is spent on the medical version. If version zero works, version one, the wound-grade Decoy, is a real company with a real lab. If it does not, we will have learned the interesting thing cheaply.

Closing

Decoy is the peace treaty between my fingers and my skin, six years of craters late. Not a cure, not a shield, a lure: the first product that does not ask the hand to stop, only to be fooled.

If you build this, tell me, and send me the chin model first. I have been prototyping the problem on my own face for years; it is time somebody else supplied the hardware.

The One at the Next Table

Sun, 23 Aug 2026 00:00:00 GMT

For most of my life, Friends was not a show I watched so much as a show that kept finding me. Waiting rooms, fast-food corners, hotel lobbies, that one kebab place with a TV bolted near the ceiling. Wherever a screen had nothing better to do, the gang was there, mid-laugh, and I was glad every single time. It has aged badly in places, I know. But for a European kid it was also a first window onto a certain America we were all discovering, and being adopted into that family, ten minutes at a time, never failed to comfort me.

They followed me everywhere, and it was always a pleasure. Then one day I did the math on my seat. Through all those years of the gang living at the next table of my life, I was never once on the couch. I was the guy at the small table, catching fragments, happy to be in the room. That, I realized, is what being a customer at Central Perk must feel like.

So here is the channel I have always wanted to watch. One fixed camera, one ordinary man, one small table by the window. Every video is a scene you already know, except you never see it. You see him. The gang is just noise from the couch, that couch, while he tries to read, or grade papers, or get over someone. A fake life in the margins of a famous one. Ten seasons of it, if we are being honest about my commitment.

He is not paranoid, the show agrees with him

The delicious part is that none of his grievances need inventing, because the series wrote them down itself. The couch, the only good seat in the house, literally carries a Reserved sign in the background of countless episodes. The café institutionalized their privilege, in writing. Fans have counted the times other customers held that couch across 236 episodes: three. Our man was, of course, all three of them. Those are the three proudest days of his life, and we will film all three.

His coffee came wrong for two straight years, because canon says Rachel was a catastrophe of a waitress. His quiet Sunday reading hour was Phoebe’s set, and even Terry, the owner of the café, is on record about what her singing made him want to do to his own brain. He has involuntary opinions about Ross and Rachel that he never asked to hold, assembled entirely from overheard fragments, the way I hold opinions about them assembled entirely from waiting rooms. He is me. He is, I suspect, most of us.

The background people of Central Perk spent ten years miming, drinking from empty cups, forbidden to make a sound. This channel gives one of them the only thing he never had: a voice.

And the production joke is that the channel is almost free. You never film the gang. You film a man, a table, a cup, and a wall of familiar noise coming from somewhere off-screen. The scenes you know by heart, playing entirely on his face. Somewhere around season four he snaps, defects to another café across the street, and comes back within a week, because the noise, it turns out, had become the furniture of his life. I have rewatched enough episodes in enough waiting rooms to know how that feels.

The same trick would work at MacLaren’s, by the way, one booth over from How I Met Your Mother, where the gang does not even stay in their corner. Barney works the whole bar. That neighbor is not weary, he is a plaintiff.

One line of realism, because this house owes you at least one. If this channel does not exist, and I could not find it anywhere, it is almost certainly because of rights. It would live at the mercy of a studio, demonetized on day one, taken down in places. That is the entire dossier, and I am closing it now, because this text is not a business plan. It is a daydream with a table number.

If you ever film it, tell me, and save me a seat. Not the couch, obviously. The couch is reserved.

Reprise: One Place for Every TV Theme

Sun, 23 Aug 2026 00:00:00 GMT

I play several instruments, none of them impressively: guitar, drums, bass, piano, and a cello keeps threatening to join. So I have a recurring evening. The one where I sit down at any of them wanting to play something that makes me grin instead of something that makes me better, and the thing that makes me grin is always a TV theme. Friends. Malcolm in the Middle. How I Met Your Mother. Thirty seconds of pure recognition, engineered by professionals to lodge in a skull.

And every single time, the same scattered pilgrimage: a tab on one site, a metal cover on some channel, a classical version on another, a lesson buried in a playlist, and no two of them within three clicks of each other. Every time, the same thought. I just want one place to go. Log in, pick the show, play, discover, done. That place does not exist, and this article is me finally writing the thought down instead of having it again.

The covers exist. The shelf does not.

None of the raw material is missing. Kerrang! has ranked the best rock covers of TV themes and notes there are entire bands dedicated to covering them. A guitarist once played 100 TV themes in a single unedited take for Reverb, in chronological order, from Looney Tunes to Game of Thrones. The channel Beyond the Guitar does gorgeous classical versions of Friends, Seinfeld and The Office. Tab sites have whole television sections. The internet has done all the work, then filed it in a hundred unlabeled drawers.

My favorite piece of evidence is small. A radio station’s blog gushed that one of those cover artists had “even transcribed the chords into tabs,” exclamation mark included. When the mere combination of two ordinary things reads as a miracle, the product is the combination. Reprise is that combination as a website: one page per show, and on it, the original theme, every cover anyone has submitted, sorted by style, and the tab or chord sheet to learn it yourself. Listen, compare, play. A WhoSampled for the songs everyone knows and nobody archives.

The playlist solves listening. It does not solve the itch.

Is the answer just a big personal playlist? I did ask myself that. But a playlist answers one question (“play me things”) and Reprise answers three: what does this theme become in another genre, who did it best, and how do I play it tonight. The join is the product. Remove any column and you are back to the hundred drawers.

And the shelf has a second reader, the cover artists themselves, and they solve the cold start. Reprise hosts nothing, so every play is a YouTube view counted on the artist’s own channel. For them, the directory is pure distribution, attribution included. The people with the strongest reason to fill the shelves are the people who made the covers, and a librarian whose suppliers volunteer is a librarian who sleeps well.

Key Features

The shelf. One page per show: original first, then covers grouped by style (metal, classical, lofi, jazz, chiptune, whatever arrives). Everything is a YouTube embed; Reprise hosts nothing and points at everything.
The learning column. Each theme links its tabs, chords, or lesson videos next to the covers, so hearing a version and learning it are the same visit.
Community submissions. Anyone can propose a cover or a tab link for a shelf; light moderation keeps it honest. The catalog of shows itself can be seeded in an afternoon from Wikipedia’s list of television theme music.
The rabbit hole, on purpose. Filters by decade, genre of show, genre of cover, instrument. The whole point of the place is leaving with three themes you did not come for.

The part where I usually list the blockers

Unlike several of my other music ideas, this one has no rights wall. It feels like a trap, but it is not. Reprise embeds rather than hosts. The copyright weight of a cover lives where it already lives, on the cover artist’s video, where Content ID claims it and routes ad revenue to the song’s publishers, the standard regime that has kept YouTube’s cover ecosystem alive for years.

Linking to tabs is linking, ideally to licensed sites like Ultimate Guitar. The directory is a librarian, and when publishers cracked down on tab sites in 2005 and 2006, they targeted the shelves, not the card catalog. Monetization, in the proud tradition of my other civic tools like Exposed and the Obscurantism Clock: barely. A tip jar, maybe an affiliate link to lessons, and the plain admission that this exists because I want it to exist.

One cultural note, because it is why the itch deserves a monument rather than a bookmark folder. These songs are a dying form. The streaming era’s Skip Intro button is quietly training a generation to never hear a theme at all.

And yet the form, at its peak, was glorious enough that Malcolm in the Middle’s theme, “Boss of Me,” was written by They Might Be Giants and won them a Grammy, and How I Met Your Mother’s theme was performed by The Solids, the band of the show’s own two creators. Friends’ theme was a real band too, The Rembrandts, dragged onto the charts by a sitcom. Reprise is a tool for my evenings, and also a small conservatory for an endangered species.

Next Steps: How I’d Build It

A weekend, really. Static site, a data file of shows seeded from the Wikipedia list, YouTube embeds, outbound tab links, a submission form. No accounts, no uploads, no backend worth the name. Launch with the ten shelves I would build for myself anyway, and let the submissions decide what the eleventh is. The hard part is not the code but the librarian’s patience of filling shelves, which is the kind of hard part one obsessed person can carry.

Closing

Reprise is the place I keep wishing existed at 10pm with a guitar on my knee, finally specified instead of wished for. The name works in both my languages. In French, a reprise is a cover, and in music, a reprise is the moment a theme returns.

If you build this, tell me, and I will be your first contributor, with the same theme submitted four times, on four instruments. The cello version will have to wait.

A Storm in a Cage

Fri, 21 Aug 2026 00:00:00 GMT

Europe burned again this summer. By mid-August, the area burned in the EU had passed half a million hectares, well above the long-term average according to Copernicus EFFIS data, and all of it in the shadow of 2025, the worst season on record at over a million hectares. In late July, France’s interior minister said the country was fighting 30 to 40 wildfires every single day, while the Gironde burned through the largest French fire since 1949. Behind the statistics, the same choreography every time: evacuations, exhausted crews, aircraft grounded by wind, and a fire that simply does not care.

What we talk about less is that fire is the cheapest weapon there is. A box of matches against a fleet of water bombers. This is not a hypothetical. During World War II, Japan launched some 9,000 incendiary balloons, the Fu-Go program, across the Pacific with the explicit goal of setting American forests alight. In a world of drones and proxy conflicts, a landscape this dry is not just a natural disaster waiting to happen. It is a target. Arson as guerrilla warfare does not need technology, only intent and a dry summer.

I wrote a while back about Asimov, Foundation, and the personal force shield, the kind of object science fiction plants in your head decades before physics catches up, if it ever does. This essay is my attempt at the same game. Watching this summer unfold, the device I kept imagining was simple to say and absurd to build: a storm you can order. Not weather control, not sunshine on demand. One storm, one place, one purpose.

What We Already Know How to Do

When I dug in, the surprise was that the impossible part is smaller than it looks. Cloud seeding has existed since 1946, when Vincent Schaefer at General Electric first coaxed precipitation out of a cloud. Disperse silver iodide into an existing cloud and you can nudge it into raining. China runs a massive weather modification program and famously used it around the Beijing 2008 Olympics. The United Arab Emirates has flown drones that zap clouds with electric charge to encourage rain. And two days after I published this essay, a California startup called Rainmaker went further, using two autonomous drones and 374 grams of silver iodide for a claimed 71,000 cubic metres of rain over Alaska in three hours. Read the caveat with the claim, though, because it is the same caveat this whole essay lives under. They did not seed one cloud and leave an identical one alone, and the clouds were already raining, so the number is an estimate of a nudge rather than a measurement of a storm. Pushing a cloud is a solved problem, more or less.

What nobody can do is create a storm from nothing. A thunderstorm needs three ingredients: moisture, atmospheric instability, and a lifting mechanism. The energy involved is beyond any machine we could plausibly build. And the cruel joke of megafires is that they strike exactly when the air is desperately dry. There is nothing to seed.

Except. A big enough wildfire builds its own lifting mechanism, a convection column so violent it can punch into the upper atmosphere. Fire crews already know the nightmare version of this, the pyrocumulonimbus, a storm cloud born from the fire itself, sometimes throwing dry lightning that starts new fires downrange. The fire already brews its own storm. Two of the three ingredients come free. The only thing missing is water.

How to Bottle a Storm

So here is the machine. A crown of pylons, dropped like darts from heavy helicopters and drone swarms, planting themselves in a wide ellipse around the fire, offset downwind, encircling not where the fire is but where it is going (which is how firefighters place containment lines today, except this line stands up). Each pylon meshes an energy field with its neighbors. The crown closes segment by segment. No roof. The cage is an open cylinder, because the fire’s chimney is the engine we need. It lifts the water we inject, we seed the top, and the fire is forced to brew its own extinguishing rain. The storm is the lid.

The Simpsons delivered their dome by helicopter to trap a town. Ours traps a disaster. Same image, opposite morality.

The wall itself is a ladder from real to speculative. At the bottom is plain physics. The pylons are lightning rods, draining the artificial storm’s strikes so they cannot start new fires. One rung up comes an industrial trick. The wall’s upper lip tilts inward like a chimney cowl and runs a curtain of charged mist, a giant wet scrubber, the same technology factories use to wash particulates out of flue gas. Hot gases pass, heavy firebrands get caught and drop back inside. The chimney keeps its engine but loses its ammunition. Higher still, the mist curtain itself, recharged by water bombers that no longer drop on the flames but on the wall.

And at the top of the ladder sits the leap I fully admit, a field that contains radiant heat and wind at kilometer scale. The germ is real. A flame is a weakly ionized plasma, DARPA showed over a decade ago that small flames can be bent and snuffed with electric fields, and Boeing even patented a laser-induced plasma shield against blast waves in 2015. But the scale-up is pure science fiction. That is the price of admission, and I am paying it openly.

Wind is the obvious killer, and the two best answers are entirely passive. The first is a skirt, an outer ring of low, porous windbreak panels, planted by the same dart logistics. Solid walls make terrible windbreaks, because they spawn violent eddies right behind them. Porous ones bleed the flow and shelter a strip many times their own height, a trick farmers and snow-fence builders have relied on for generations. The second is borrowed from the spiral strakes on industrial chimneys. The field curtain is ribbed in a helix, so a tall cylinder does not shake itself apart in the wind. Above those, a profiled windward face vaults the surviving flow up and over the cage, and a control loop shifts power between segments as sensors read the load. Not an AI, on purpose. An institution that cages storms wants decisions that are deterministic and auditable in court. The wall is dumb by design.

The order of operations is doctrine, not detail. Sweep the zone. Plant the crown. Preload: water bombers inject moisture into the column while the ring is still open. Seal. Seed. Storm. Mop-up, because rain never kills the deep embers. Human crews still walk in last.

Two rules hold it together. Water first, because sealing a dry superheated chimney builds the perfect firebrand launcher. And no seed without seal, because an unconfined artificial storm is precisely the thing that killed 19 firefighters at Yarnell Hill in 2013, when a natural storm’s outflow turned the fire on them in minutes. The cage is not an accessory to the storm. It is the only thing that makes the storm legal.

The Storm Department

Legal is the right word, because the trigger question is nastier than the physics. Your storm steals your neighbor’s rain. Weather does not recognize borders. There is already, astonishingly, a treaty for this. ENMOD, signed in 1977, bans environmental modification as a weapon. The institution almost writes itself. A dual key, borrowed from nuclear doctrine, means the host country requests and a treaty authority authorizes. The EU already pools firefighting aircraft through rescEU. Extrapolate that into a planetary Storm Authority with prepositioned regional depots, wheels up in 6 hours, crown sealed in 24-48. The inside of the cage becomes the only volume on Earth where triggering a storm is lawful.

And the human protocol is where the weight lives. The sweep is a phase of its own: thermal drones, cell network pings, mandatory evacuation, and a crown that closes progressively, keeping the final segment open as an exit corridor. The last door. Somewhere in this system there is a sealing officer whose signature closes it, and who has to live with the fact that no sweep ever reaches certainty. Max Weber defined the state by its monopoly on legitimate violence. The Storm Authority extends that to a monopoly on legitimate weather. Put another way, and this is the quiet part, a machine that makes storms is a weapon that makes storms. That is why it gets caged twice, once in fields, once in institutions.

While I was writing this, the institutional half stopped being hypothetical. Rainmaker notified the state of Alaska and skipped the town. Within days a state representative was demanding a cease-and-desist, and the borough mayor was explaining, publicly, that nobody had told him. Seventy thousand cubic metres of rain, made privately over somebody’s home, was enough to produce the crisis of legitimacy this section is about. Multiply that by a caged megafire and you see why the sealing officer needs a signature, a mandate, and a phone number the mayor already has.

One guardrail before anyone gets excited. Critics already argue that Europe pours money into fighting fires and almost none into preventing them. A working storm cage would make that moral hazard worse. Why manage forests when you can bottle the consequences? If this thing ever exists, its budget should be legally chained to a prevention budget twice its size. The cage is a tourniquet, not a cure.

Someone Else’s Essay

I know what remains broken. Helicopters still cannot fly in the winds that drive the worst fires, so the machine that tames weather still waits on weather. There is a night in every version of this story where the wind exceeds the spec and the crown stays in its depot. That problem belongs to someone else’s essay, and honestly, to someone else’s century.

Asimov’s personal shield made individual violence pointless, and Foundation spent a thousand pages exploring what that does to power. A storm in a cage would make wildfire, one of humanity’s oldest terrors and newest weapons, into something a bureaucracy schedules. This is an essay, not an idea, because nobody builds this from a garage.

In the spirit of this site, I will offer one exception. The skirt stands alone. A dumb, porous windbreak ring that fire crews could deploy tomorrow needs no plasma and no treaty. If you start there, tell me.

So here is what I keep turning over. If we ever hold the keys to one commanded storm, who do you trust to turn them, and what do we stop taking care of the day it works?

The Shield Will Come

Mon, 06 Jul 2026 00:00:00 GMT

When I was a kid, I was certain force fields were coming. Not hopeful, certain, in the same flat way I was certain about flying cars and holograms. It seemed obvious that somewhere between my childhood and my old age, a person would be able to switch on an invisible skin that nothing could cross, and walk through the world untouchable. I have carried that certainty, unspoken, for a long time.

The image came from books before it came from physics. Asimov’s characters in the Foundation stories carry personal shields as casually as we carry phones. Dune gave us the Holtzman field, the shimmering bubble that turns a knife fight into a slow, patient dance, because only a slow blade can slip through. The force field is one of science fiction’s most durable promises: protection made personal, portable, and absolute.

And then you grow up and learn why it is so hard.

Why the permanent shield is a fantasy

Picture the classic version: a field that hugs your skin and stops everything. The first problem is that “everything” includes air. A barrier impermeable to all matter is a barrier you suffocate behind. Make it selective, let oxygen through but not a blade, and you are now asking the field to sort molecules by intent, which is not a thing fields do.

Then there is heat. Your body runs at something like a hundred watts, a small lightbulb’s worth of warmth that has to go somewhere. Wrap it in perfect insulation and you cook inside your own bubble within minutes. And even setting biology aside, the energy budget is brutal. Stopping a hard stab means absorbing thousands of joules in an instant, again and again, from a battery you can wear. We have no known field that repels neutral, uncharged matter the way the fantasy needs. Magnets bend charged particles; they do nothing to a kitchen knife.

The permanent bubble, in other words, is where the dream dies. Not because a single grand law forbids it, but because it asks for the impossible on four fronts at once.

From a wall to a reflex

The escape, and I think it is a real one, is to stop imagining a wall and start imagining a reflex.

Instead of a field that is always on, picture a reservoir, a thick bracelet or a collar, holding a swarm of particles that do nothing until they are needed. A threat approaches, sensors read its path, and in a few milliseconds the swarm leaps to the point of impact, hardens into a small rigid shield exactly where the blow will land, absorbs it, and falls back to the reservoir. No permanent bubble. You breathe normally, you shed heat normally, and the energy is spent only in the instant it matters, on the few square centimeters that matter.

This is not magic, and that is the point. These are physical particles, not a mysterious field. We can already make materials that are absurdly light and absurdly strong; what we cannot yet do is command them as a fast, intelligent swarm, or detect and deploy inside the hundred milliseconds a stab gives you. Those are hard engineering problems, not violations of the universe. It is the difference between a giant umbrella held open over your whole life, and one that opens only where the rain actually falls.

The quiet good we keep forgetting

When we picture a shield we picture drama: the knife turned aside in an alley, the soldier walking through fire. Those matter. Assaults are real, and a wrist that could blunt a blade would change how safe a lot of people feel walking home at night.

But the version that would save the most lives is almost boring. The deadliest animal to human beings is not the shark or the bear. It is the mosquito, which kills more people every year than any other creature, several hundred thousand of them, most of them children, mostly through malaria. A shield that could stop something as thin as a proboscis, a stray needle in a hospital, a wasp for someone who goes into anaphylaxis, is not a weapon at all. It is public health you wear on your wrist.

That is the future worth wanting. Not cinematic invincibility, but the quiet subtraction, everywhere at once, of a whole category of small, lethal things: the bite, the prick, the fall that breaks an old hip. Imagine the anxiety that simply lifts, for the night-shift worker, the traveler, the parent raising kids where malaria still kills.

And if it actually arrived

Every good thing this size drags its shadow behind it.

The moment shields exist, weapons are designed to beat them, and we are back in Dune, inventing the slow blade that slips through. If they are expensive, they create a new and grotesque kind of inequality: a class of people who are literally untouchable, and everyone else. Policing breaks in strange ways when a suspect cannot be physically stopped. And a protection you depend on is a protection that can fail. A dead battery, a bad update, a hack, and the untouchable becomes the most vulnerable person in the room. A shield does not remove fear from the world. It relocates it.

So which is it

I still cannot shake the childhood certainty, but I have learned to hold it more carefully. The honest question is not whether we want the shield, we obviously do, but whether the universe allows it. Is there a hard wall here, like the speed of light, a law that simply says no? Or is this like every other “impossible” thing that turned out to be a matter of patience and cleverness, the way a thin modern vest that stops a bullet would have looked like sorcery to a knight?

I don’t know. But I notice that the version I can almost believe in is not the one from the films. It is a bracelet that flinches for you, faster than you can, and mostly saves you from the smallest things. Maybe that is where every big dream ends up, not as the impossible wall we asked for but as the quiet reflex we can actually build.

The Volume Knob for Reality

Mon, 06 Jul 2026 00:00:00 GMT

My mother lives with several health problems, and one of them is her hearing. What that means in practice is not what people imagine when they hear the word. It is not silence. It is noise without hierarchy. At a family dinner every voice, every fork, every chair leg arrives at the same rank, and the one voice she is trying to follow, a grandchild’s story, a question addressed to her, drowns in the middle of it. What she has lost is not so much volume as separation, the ability to choose one sound and let the rest recede, which every healthy ear performs without thinking.

So I have carried a dream for years, long before I could name the technology it would need. I dreamed of building her an ear that could choose. Point it at one precise sound, her name in a crowded room, one voice at a full table, and let everything else fall away. Not louder. Sorted. In mixing-board terms, not a volume knob at all but a solo button, for the one channel that matters.

One evening in early 2024 I dictated a version of this dream into an AI, half expecting to be told it was science fiction. The answer, roughly, was: not quite, not yet. And in the time since, “not yet” has been getting uncomfortably close to “soon.”

The mixing board is being built

Understand what the dream actually asks for, because it is not what hearing technology has historically offered. Hearing aids amplify, magnificently better every decade, but amplification raises the whole mix, the voice and the forks and the chairs together. That is exactly my mother’s problem, restated at higher volume. Noise cancelling, at the other end, is a wall. It subtracts everything, then pipes back whatever the manufacturer decides transparency should sound like. The dream is neither. The dream is a mixing board, with reality arriving as separate channels and a hand, hers, on the faders.

The pieces are appearing in public. Consumer earbuds already do adaptive transparency, softening harsh spikes while letting speech through, and can tune their passthrough to a personal hearing profile, a primitive fader in disguise. And in research labs, the real thing has been demonstrated. The prototypes, headphones their creators call semantic hearing and target speech hearing, classify the soundscape in real time (this texture is rain, that envelope is a voice) and isolate the one source you choose, in some demos simply by looking at the person you want to hear. What the dream needed was a real-time sorting algorithm for sound, and the sorting algorithm now exists in embryo.

There is no exotic physics ahead, only engineering: better models, lower latency, longer batteries. That is why the dream deserves to be taken seriously as a question rather than a gadget. The force-field essay elsewhere on this site ends by noticing that big dreams tend to arrive as modest reflexes we can actually build. This is the auditory version, and it is closer than the shield.

The mercy of it

Start with my mother, because she is the point. For her, and for the enormous population like her, this technology is not comfort but access, the difference between attending a family dinner and enduring one, between asking people to repeat themselves until she stops asking, and simply hearing the person she came to hear. Hearing loss quietly severs people from conversation first, then, too often, from company itself. Audiologists have warned for years about the isolation that follows. A selective ear does not cure anything, and this essay makes no medical claims. It restores something more social than clinical: her place at the table.

The mercy generalizes. For those with misophonia, for many autistic people, for anyone whose nervous system treats certain ordinary sounds as sirens, the world is currently take-it-or-leave-it. Endure the assault, or withdraw behind total isolation. A selective ear is a third option that has never existed, removing the one sound that makes the room unbearable instead of fleeing the room. And yes, beyond access there is ordinary relief, the version of this dream everyone recognizes. The neighbor’s drill through the wall slides to nothing, while the birds stay where they are. I want that too. But I want it the way you want dessert, and I want my mother’s solo button the way you want medicine.

The vertigo of it

Now turn the same knob the other way, and feel it, because an editable soundscape edits more than drills.

If I can mute a drill, I can mute the drilling neighbor when he says hello. The gadget joke, mute your spouse, stops being a joke the day it ships. We will be able to delete people, selectively, politely, undetectably, while nodding along. Sound is how the world interrupts us, and interruption is not always an attack. Sometimes it is the alarm, the cry from the next street, the protest passing under the window precisely because it wants to be heard by people who would rather not. An editable soundscape is also a machine for never being reachable by anything you did not subscribe to.

And someone will own the filter. The fader board is software, and software has owners, defaults, and business models. Picture sponsored sounds whitelisted into your transparency, or the store jingle your earbuds helpfully classify as ambient information. A filter you do not fully control is someone else’s editorial line running between the world and your ears. We spent two decades learning what feed algorithms did to what we read. This is the same question, moved into the skull’s last unmediated channel, and it deserves the lessons of that era built in from the start: filters legible and owned by the wearer, some sounds constitutionally unmutable (the siren, the scream, the smoke alarm), and a design bias toward the world rather than away from it.

I hold both truths at once, and the tension does not resolve. The same algorithm that could hand my mother back a conversation could hand the rest of us a deafness we chose. Most powerful accessibility technology carries this double edge. The answer has never been to withhold the mercy, but to build the mercy first and the mute list with humility.

What I actually want

I want the version that exists for her. The solo button, not the lifestyle gadget, though it will come wrapped as one. I want to sit at a loud family table and watch my mother lean toward a grandchild’s story and simply receive it, whole, first try, while the forks and the chairs and the rest of us blur into background where we belong. People who have never lost separation do not know what a luxury it is to choose a sound. She knows. I have watched her know it for years.

So I leave this one with you as a question, in the house style, but it is really two questions now. When the volume knob for reality arrives, and it will, sooner than the shield, what is on your mute list, and what would you keep at full volume out of duty to the street you live on? And the second, the one I care about: who are you building it for first, the annoyed, or the ones like my mother, who are not asking to turn the world down at all, just asking to hear one voice in it again?

Web Concierge: The Emergency Plumber for the Web

Mon, 06 Jul 2026 00:00:00 GMT

This idea was born inside agency life, watching what agencies are structurally unable to do. I have spent enough time in that world to love it and to know its caveats by heart: processes measured in weeks, briefs and tickets and validation loops, and a door that closes at six and stays closed all weekend. None of that is malice. It is how agencies are built, and for big projects it is even how they should be built. But it means that when a client’s site goes down on a Friday evening, the machine that built that site cannot help them until Monday, through a ticket.

And I kept thinking about the person on the other side of that outage. When a pipe bursts at home, everyone on earth knows the move. You call an emergency plumber. It costs more than a normal visit, someone shows up, the water stops. The web has no equivalent. Your site is down, your mailbox is dead the night before a launch, your WordPress is hacked and displaying pharmacy ads, your SSL certificate expired and every visitor sees a red warning, and there is no number to call. There is an agency that will open a ticket, a marketplace that will take days, and a cousin who knows computers.

Web Concierge is that missing number. It is not an agency and it does not fix anything itself. The concept rests on that distinction. It is a dispatch service: you describe the emergency, a human triages it, and within the hour you have one to three vetted experts proposed to you, each with a stated price and a stated timeline. You pick, they work, the concierge stays in the loop until it is resolved. You are buying time-to-qualified-expert rather than talent, with the guesswork removed at the worst possible moment to be guessing.

Market & Opportunities

The demand side is everyone who depends on a website but employs no one who understands it: small businesses, solo founders, restaurants whose bookings run through their site, marketing managers whose agency is asleep. Their emergencies are remarkably repetitive, which is excellent news for a dispatch service. Site down, email broken, hack cleanup, domain or certificate expirations, a plugin update that took everything with it, a payment page silently failing. A triage playbook can cover most of it.

Now the graveyard, because matching clients with freelancers is one of the most crowded markets on the internet. Upwork and Fiverr exist, but their profiles, proposals, and days of latency make posting an emergency there absurd. The curated platforms, Malt, Comet, Toptal, Crème de la Crème, match vetted talent to projects and missions, not to fires. Their unit of work is the week, not the hour. And the reverse-auction model, where you post a problem and collect a pile of quotes, exists too, and carries a known pathology: a race to the lowest bid, template proposals, and the pollution that makes those platforms exhausting. I flirted with that model in my first notes on this idea, and it recreates the disease it claims to cure. The gap in this saturated market is precise and real. Nobody owns urgency. Nobody sells the hour between “it broke” and “the right person is on it.”

The wedge writes itself. Be open when agencies are closed. Evenings, weekends, launch nights. That is the entire initial product, not a marketing line.

Key Features

A human dispatcher, not a form into the void. Triage is the product’s brain: a person (assisted by tooling) who can tell a DNS problem from a hack from a hosting outage in three questions, and who knows which expert on the roster eats that problem for breakfast.
An SLA on the match, not on the fix. The promise is contractual and narrow: a qualified, vetted expert proposed within the hour. No one honest can promise resolution times before diagnosis, and the ones who do are the locksmith scam of the web.
One to three offers, price and timeline upfront. Each proposed expert comes as a card: what they will do, what it costs, when it starts. The client chooses in one screen. No negotiation theater at 11pm.
A vetted on-call roster. Small by design: ten reliable, reachable experts across the core specialties beat five hundred profiles. Every expert is identity-checked, skill-verified by real incident history, and paid to be reachable, not just listed.
A credential protocol. Emergencies are when people hand their server passwords to strangers, which makes an emergency service either the answer to scams or a scam vector itself. Access goes through a defined protocol: scoped credentials, password-manager handoffs, logged access, revocation after the job. The whole brand stands on this trust.
The concierge stays in the loop. Billing runs through the platform, the fix ends with a short plain-language report (“what broke, what we did, what to do so it does not recur”), and there is a recourse if something goes wrong. That loop, not the introduction, is what the fee buys.

The honest tensions

Disintermediation is guaranteed, so plan for it instead of fighting it. After the first great fix, the client saves the expert’s number, the way a good plumber ends up in your contacts. Fine. The match fee and the urgency premium are earned on the first incident, and a client who returns does it for the guarantee, the billing, the protocol, and the fact that their favorite expert might be busy at 2am but the roster never is.

The on-call economics are the real puzzle. Freelancers are not employees; you cannot rota them for free. The workable structure is a standby pool. Experts opt into availability windows, get a small retainer or priority routing for doing so, and take a majority share of an urgency-premium rate when a job lands. If the economics of being reachable do not work for the experts, the SLA is fiction.

And the scale ceiling should be said out loud. Human triage and a curated roster do not scale like software, and this article will not pretend otherwise. This is a service business, profitable at the scale of a city, a language, a niche, growable roster by roster. It is a very good business before it is ever a platform, and it may never need to be a platform.

Monetization

The client pays a flat triage-and-match fee, visible before they commit, plus the expert’s stated rate with an urgency premium that is split in the expert’s favor. The platform takes its margin on the premium and the fee, never by squeezing the base rate. Transparency is the differentiator against the emergency-locksmith reputation, with prices stated before work starts and no surprise multipliers at invoice time.

The second layer is the retainer, and it may quietly become the core. It is a monthly “we are your emergency number” subscription for small businesses, which buys priority routing, stored access protocols (so the fix starts in minutes, not after a password hunt), and a yearly preventive check of the usual suspects, backups, certificates, updates. It is insurance pricing for insurance psychology, and unlike a maintenance agency contract, it costs little to serve until the day it earns its keep.

Next Steps: How I’d Build It

Start embarrassingly small, as one dispatcher, a phone number, a form, and ten experts you personally trust across five specialties: hosting and DNS, WordPress and CMS, email deliverability, security and hack cleanup, payments. No app, no marketplace code. A shared inbox, a triage playbook, and a rate card are the whole MVP. The first ten incidents will teach more than any spec.

The blockers are operational, not technical. Liability and insurance are real. An expert with production access can make things worse, and the concierge carries part of that responsibility, so professional insurance and a serious vetting bar are launch requirements, not later polish. Coverage is the other one. True 24/7 with ten people is fantasy, so open evenings and weekends first, the hours when the incumbents are closed, and grow the promise only as the roster earns it.

I will admit the part that appeals to me personally. The hard part of this business is a live operational puzzle, triage under pressure, routing the right human to the right fire, tuning an SLA until it holds. Some people find that stressful. I find it the best kind of challenge, which is probably the final proof that this idea was never going to leave me alone.

Closing

Web Concierge is the emergency number the web forgot to create, built on a narrow promise kept ruthlessly, a vetted human on your problem within the hour, at a price you saw before saying yes. The market is crowded everywhere except exactly here.

If you build this, tell me, and tell me your time-to-expert numbers, because that metric is the product. And if you run an agency and this article stung a little, you know precisely why this service needs to exist.

WaveArt: Limited-Edition Art From the Songs You Love

Thu, 19 Feb 2026 00:00:00 GMT

In the summer of 2023 I was deep in pivot mode on a project, and I talked myself into something I still think about. The idea was simple to say out loud: take the waveform of a song that matters to someone, and instead of running it through yet another automatic poster generator, put it in the hands of a real artist. I got far enough to write the brief. I had five talented designers around me I wanted to test, a plan to send each of them one well-known track’s waveform, and a marketing budget of exactly 300 dollars a month, which is to say, almost none.

Then it went on the shelf, for three very ordinary reasons I will get to at the end. But the idea never stopped feeling right, so here it is, whole, for whoever wants it.

WaveArt is a service that turns the waveform of a song into a limited-edition artwork. Not a generic gradient on a black background, but a piece designed by a named artist, released as a collection of three variants, available for three months, then gone. You pick the song that holds the memory, the artist’s design holds the song, and you receive a high-resolution digital file and a premium print, A3 to A5, framed if you want. When the quarter ends, a new artist and a new collection take over.

Why the twist is the whole product

Waveform art is already a saturated market. Search for it and you will find established services and countless shops that will render any audio file into a poster in seconds. They all compete on the same axis: automation, speed, price. So the output all looks the same, and the price can only go down.

WaveArt runs the other way. The bet is that in a market of infinite instant posters, the scarce things are authorship and scarcity itself. A collection signed by an artist with a name and a story, in an edition that expires, is a different object from a generated poster, even if both contain the same waveform. One is a printout, and the other is a small act of curation you were part of because you chose the song. The drop model borrowed from streetwear does double duty here. It creates urgency for buyers, and it gives a tiny company a natural publishing rhythm (a new launch, a new artist interview, a new story every quarter), which matters enormously when your entire marketing budget is 300 dollars a month. The calendar is the marketing.

Key Features

A named artist per collection. Each quarter, one artist designs the collection and signs it. Their name is on the piece and their story is on the site. They are the brand as much as the product is.
Three variants, and only three. Enough choice to feel chosen, few enough to stay a collection rather than a catalog.
Quarterly drops that actually end. When a collection closes, it leaves the storefront. A compromise softens the sting: past collections remain orderable from an archive at a higher price, unframed, so missing a drop costs something but does not cost everything.
Your song, your waveform. The buyer uploads or selects the track that matters, and its waveform is flowed into the design for their piece. Two orders of the same variant are never identical, because no two songs draw the same line.
Three tiers. Digital file only, print, or framed print. The digital tier keeps an accessible entry price; the framed tier carries the margin.
A preview you can trust. The homepage shows each variant rendered with one well-known track, so buyers see the quality they are getting. My original plan stopped there (no personalized preview before ordering) to keep production manual and simple. I now think that was a mistake. Composing a waveform into a template is the kind of step you can automate for the preview alone, while keeping the finishing by hand. Let the buyer see their own song in the design before paying. At this price point, blind ordering is the friction that kills the sale.

The honest tensions

The word “handcrafted” needs discipline here, because the production model is a template. The artist creates the design once, and each order re-flows a new waveform into it. Calling every piece hand-made would be marketing fraud on a delay. What is honestly claimable is that the artwork is an original, signed design by a named artist, and that each piece is individually finished and quality-checked by hand. The art is the template, and the personalization is mechanical. Say both plainly and the story holds. Blur them and one viral post about it ends the brand.

The second tension is the artist pipeline. The entire concept stands on finding good artists, paying them fairly (a flat commission plus a small royalty per sale is the clean structure, so a successful collection pays its creator), and refreshing that talent four times a year. That is a real recurring cost and a real curatorial job. It is also the moat. Anyone can clone the website, but the taste and the artist relationships are harder.

The rights question I never resolved

Now the section I owe you, because this is one of the reasons the idea stayed on my shelf. I was afraid the waveform itself might belong to someone. A recording is copyrighted, and a waveform is an image derived from that recording. Is selling an artwork built on the waveform of a song you do not own a derivative use that needs a license, or is it unprotectable data about the sound? I do not know, and I am not going to pretend this article settles it. A whole industry of waveform-print shops operates openly as if the answer is favorable, but “everyone does it” is a business observation, not a legal opinion.

What I can offer are the practical mitigations any builder should take before launch: have the customer upload their own audio file rather than pulling tracks from a catalog you do not license; keep song titles, lyrics, artist names, and cover art off the artwork itself unless the buyer adds them for private use; never imply endorsement by any musician; and spend one real hour with an IP lawyer in your jurisdiction before printing anything for money. That hour costs less than the first cease-and-desist.

Monetization

The pricing has to say “limited edition,” not “poster.” A digital-only tier at an accessible price feeds gifting and impulse buys. The print tier is the core product at a premium, and the framed tier is the flagship. The archive at a marked-up price monetizes regret, gently. On the cost side, on-demand production means no inventory risk, and printing can start with a quality print-on-demand partner before ever buying a printer. The artist royalty aligns everyone. The artist promotes the collection because the collection pays them.

And the constraint I set myself in 2023 still shapes the model. With 300 dollars a month, paid acquisition is a rounding error. The product has to market itself, which means every piece should be beautiful enough to be posted, every collection launch should be a small story worth telling, and the artists’ own audiences are the distribution. If the product cannot generate its own gravity, no ad budget of this size will save it.

Next Steps: How I’d Build It

Start exactly where I stopped, with the designer test. Send five artists the same well-known waveform and one page of brief, pay them for the test, and see who produces something that makes you feel the song. That single round tells you whether the concept has legs, because if five good designers cannot make a waveform moving, no funnel will.

Then build the smallest possible shop: one artist, one collection of three variants, the automated personal preview, checkout, and a print partner. The technical blockers that stopped me in 2023 (cleanly extracting waveforms from arbitrary audio, compositing them into layered templates at print resolution, automating the preview) are easier now, because the tooling for all three has improved dramatically. The hard part has not changed. It is the taste, the artists, and the discipline to let a good collection die on schedule.

Closing

WaveArt is the pivot I never made. I shelved it for the most human reasons there are: no time, an unresolved legal fear, and a few technical walls that have since gotten shorter. The brief is written, the tensions are mapped, and the trap doors are flagged. If it was ever going to be built by me, it would have been by now, which is why it belongs here. It sits in the same family as Pixel Prints, my concept for turning game captures into physical prints, and Siteprint, the same premise pointed at a website’s crawl graph. Digital memories deserve walls, and any data fingerprint someone is proud of is a candidate for the frame.

If you build this, tell me, and show me the first collection. And if you get a real answer to the waveform rights question, I want that even more than the poster.

Siteprint: Your Website's Architecture, Framed

Wed, 23 Apr 2025 00:00:00 GMT

There is a specific image that every web person I know has stopped to admire at least once: a crawl of a website rendered as a force-directed graph. The homepage as a sun, sections as constellations, thousands of pages pulling on each other with invisible springs until the whole structure settles into something that looks like a nebula, or a neuron, or a dandelion caught mid-breath. SEO tools render these as diagnostics, and we screenshot them anyway, not because the screenshot helps the audit but because they are beautiful. I have watched colleagues, serious grown professionals, pass these around like trading cards: look at the shape of mine.

That is the insight. We in the agency world spend years building things that are invisible. A site launches, it lives in a browser tab, and nothing about the physical world records that hundreds of hours of structure and craft exist. I am the geek this idea serves. I want my projects on the wall. Not screenshots of homepages, the architecture itself, the shape of the thing I built, printed properly, and, this is the part I care most about, reprinted as it grows, so the wall shows the site at launch, the site two years in, the site after the big redesign, like growth rings of a living organism.

Siteprint is the service that does this. Point it at a site, or feed it the crawl you already have, and it renders the link graph as wall-grade art (curated styles, careful typography, a discreet caption line, your domain, the date, the page count) and ships it as a print worth framing. Mapiful proved the model for city maps, personal cartography as decor. This is the same move for the places we actually built.

The buyer is the agency, and that clarifies everything

Left as a consumer product, this is a curiosity. Aimed at agencies, it becomes a machine with obvious slots to fit, and I know because I have lived every one of these moments from the inside.

The launch gift, first. A site delivery ends with an invoice and a handover call, which is a strangely cold ending for months of collaboration. An agency that hands the client a framed print of their new site’s architecture, dated, elegant, unexpectedly moving, ends the project with an object that lives in the client’s office and quietly advertises the relationship to everyone who asks what it is.

The before-and-after, second. Redesigns and SEO cleanups are notoriously hard to make visible. The client paid for months of work and sees, what, a slightly faster page? Print the crawl from before next to the crawl from after, the tangled thicket versus the clean tree, and the invoice explains itself in a diptych. No slide deck has ever communicated information architecture surgery this fast.

Conferences and fairs, third. The agency booth handing out branded site-graph prints, or offering to crawl visitors’ sites on the spot and mail the poster, has the best object in the hall.

And the wall of the builder, my favorite. The annual crawl becomes a subscription, one print a year, same style, same wall, the organism visibly growing. Growth rings for a digital garden.

Key Features

Crawl or import. Siteprint can crawl a site itself, but day one it should accept exports from the crawlers agencies already run, the industry-standard audit tools every SEO has open anyway. The render is the product; the crawl can come from wherever it already lives.
Curated styles, not a settings panel. Like the map-poster services: a handful of designed looks, nebula on ink-black, blueprint on drafting blue, botanical line-art, each tuned so any site’s graph comes out gallery-worthy. Infinite knobs produce ugly posters; taste is the moat.
The caption line. Domain, crawl date, pages, depth, maybe one number the owner is proud of. Understated, typographic, museum-label energy.
Diptych and series formats. Before-and-after as a designed pair; the annual-growth series as a matched set with consistent scale and style, built for the wall that fills over years.
Agency mode. White-label output with the agency’s mark alongside the client’s, bulk pricing, and an API so the print can be ordered from inside the delivery workflow.
Digital tier. A high-resolution file and an animated version, the graph settling into place over twenty seconds, for the case study page and the launch post.

The honest tensions

The crawl is less trivial than the poster. JavaScript-heavy sites render reluctantly, login walls hide half the structure, and a million-page site needs intelligent sampling before it becomes soup. That is why accepting existing audit exports first is the right MVP. The agencies this serves already own clean crawls, and Siteprint’s actual craft, the part nobody seems to sell yet, is the rendering.

Permission matters and is happily simple. You print sites you own or were hired to work on. Crawling a competitor’s site into a poster for your office is technically possible and legally murky, and the service should simply not care to support it; the market of proud owners is the entire market.

As for scale, this is a niche product with a gifting-and-B2B engine, not a consumer wave. That is fine, and familiar. It is the WaveArt logic transposed. In a world of infinite dashboards, the scarce thing is the object. The agency channel gives it something WaveArt lacks, though: a professional buyer with a recurring reason to purchase and a marketing budget that already spends money on far worse gifts.

Monetization

One-off prints at premium-poster pricing for individuals, framed options included. The real business is the agency side: bulk packs, white-label subscriptions priced per delivery, and the annual-crawl series as recurring revenue, a subscription whose renewal reminder is literally a beautiful object arriving in the mail. Trade-show packages, crawl-on-site plus fulfillment, round it out. Production is print-on-demand from day one; no inventory, no printer, no excuse not to start.

Next Steps: How I’d Build It

The MVP is a rendering pipeline, not a platform: parse a standard crawl export, lay the graph out with a force simulation tuned for print density rather than screen interactivity, apply one beautiful style, output print-resolution files, and hand fulfillment to a print partner. One style, one format, a landing page, and ten posters printed for the agencies where my colleagues work. Their walls are the pilot study.

The craft risk is singular and clear. The default output must be gorgeous for the median site, not just for the lucky elegant ones, and that is a problem of layout tuning and restraint, solved by iteration and a designer’s eye rather than by features. If the twentieth random site still produces a poster you would hang, build the rest.

Closing

Siteprint turns the invisible work of the web into an object, for the client who just got their site, the agency that wants its work remembered, and the geek, me, who wants to watch a project grow on the wall the way other people watch their kids’ heights climb a doorframe. It extends the family of WaveArt and Pixel Prints, digital things that earned a place in the physical world.

If you build this, tell me, and crawl this site first. I want the lake on my wall, and next year I want the lake again, bigger.

SaveThisGame: The Suggestion Box Studios Can Actually Read

Mon, 03 Feb 2025 00:00:00 GMT

I have loved a lot of games, but the ones that taught me this idea are the ones that never stop changing. Live-service titles like Call of Duty: Warzone evolve permanently. A patch lands, a weapon gets gutted, a beloved mode vanishes, and overnight a community that adored the game is furious at it. And I kept noticing where all that energy goes: nowhere. It explodes across Reddit threads, Steam discussions, and Discord rants, thousands of thoughtful suggestions mixed into an ocean of rage, unranked, unstructured, unread. The players are doing free design work at scale, and it evaporates.

I believed in the fix enough to buy the domain. The idea behind SaveThisGame is one page per game, where players post concrete improvement suggestions, the community votes, and the result is a ranked, living list of what the players who care most actually want. Then life did what it does. I had no time, and eventually no momentum. The domain sat there as a quiet reproach. But I have never stopped thinking the idea is useful, especially for independent studios, so here it is in full, flaws and roadblocks included.

Picture a public backlog, written by the players, prioritized by votes, readable by the studio at a glance. Not reviews, not scores, not “this game is trash.” One rule defines the entire site: every post must answer the question “how do we make this game better?”

The graveyard, named honestly

Pieces of this exist, and pretending otherwise would be the usual startup theater. Studios that want structured feedback can already run their own boards on tools like Canny or UserVoice, and some do. Minecraft operates its own official feedback site with voting. Subreddits and Steam forums fill the social role. So what is left to build?

The gap is everything the existing pieces exclude. The B2B boards only exist for games whose studios opted in, which is precisely not the games whose communities are loudest. The official boards belong to the publisher, who curates what is visible. And the open forums have volume without structure. There is no dedup, no ranking, no way to tell a thousand-upvote consensus from a loud afternoon. A third-party directory where players organize feedback for any game, whether or not the studio has shown up yet, does not really exist in credible form.

And there is a reason it does not exist, which took me longer to see. The structural failure mode of every third-party feedback aggregator is that without the developer in the loop, it becomes a complaint box shouting into the void. Suggestions lead nowhere, contributors burn out, the site dies. Any version of SaveThisGame worth building has to be designed around that gravity, not in denial of it.

The legal twist that changes the pitch

One detail upgrades this from “another forum” to something a studio might actually want. Many studios, on their lawyers’ advice, refuse to read unsolicited fan ideas at all. The fear is straightforward. A fan submits an idea, the studio ships something similar two years later, and an authorship claim arrives with an invoice attached. The wall between players and developers is legal exposure more than arrogance or technology.

That is a problem a platform can actually solve. SaveThisGame’s terms would make every submission an explicit, irrevocable, royalty-free license. Post a suggestion and you are, by design, giving it away to the studio, to everyone. The platform becomes the one place on the internet where reading fan ideas is legally safe. That is the real product. The votes are the interface, but the license is the value.

Readers of this site will recognize the shape of that contract, because this site runs on the same one. Ideas are offered freely, in public, for whoever wants to build them. SaveThisGame is that principle, applied to game development, with terms of service doing the handshake.

Key Features

One page per game, one ranked list. Suggestions live under the game they concern, sorted by votes, so the top of the page is always the community’s current consensus on what matters most.
A structured template, not a text box. Every suggestion states the problem, the proposed change, and its scope. Structure filters rage into design and makes duplicates detectable.
Merge and dedup as a first-class feature. Forty variants of “fix the audio” become one suggestion with forty voices behind it. Moderators and reporters feed this. It is the platform’s hardest ongoing job, and the article says so.
Verified studio presence with honest statuses. A verified developer can mark suggestions as seen, planned, or declined, and “declined, with a sentence of reasoning” must be a respected outcome, not a punishable one. A studio that says no clearly is participating, and the culture has to protect that.
Patch-aware pages for live-service games. This is the feature my Warzone frustration wants: suggestions and votes tied to game versions, so a studio can see sentiment move patch by patch, and players can see which era of the game a complaint belongs to. Feedback about a game that changes weekly is meaningless without a timestamp.
The license layer, visible. Every suggestion carries its “freely given” status on its face, so studios can read without fear and contributors know exactly what posting means.

The honest tensions

The cold-start problem is double-sided and brutal. Players will not come without games worth discussing, and studios will not come for empty pages. My 2023 notes imagined launching with the top twenty titles, Fortnite, GTA, Call of Duty, and that was backwards. Those studios drown in feedback channels they already own and have zero need for a third party. The wedge is at the opposite end, with independent studios. They want to hear their players, cannot afford a community team, and would treat a ranked, deduplicated, legally safe suggestion list as a gift. Launch with ten indie games whose developers said yes beforehand, and let the directory grow from proof, not promises.

Then there is vote quality. Gaming communities invented review-bombing, and a suggestion platform inherits that culture on day one. Brigades will push agendas, duplicates will flood pages, and “make the game good again” will arrive hourly. The defenses are unglamorous: the structured template, aggressive merging, per-account vote weight that favors sustained contributors over drive-by mobs, and human moderation as a budgeted cost rather than an afterthought.

And a word on the name. SaveThisGame is memorable precisely because it is combative, implying the game needs rescuing from its own developers. That framing recruits players and repels studios. I keep the name because it is the one I bought, but a builder should feel free to soften it, and the studio-facing pitch should lead with the license layer, not the crusade.

Monetization

Free for players, always, because charging the people donating design work would be absurd. The business is the studio side, a modest SaaS dashboard with verified presence, analytics on sentiment and trends, patch-over-patch tracking, and exportable reports for production meetings. Priced for indies first, because they are the wedge and their budgets are real but small. Two red lines keep the platform honest. No studio can pay to hide or reorder community suggestions, and verification is never paywalled into extortion (a studio can always claim its page and answer for free). The moment ranking is for sale, the whole edifice is worthless.

Next Steps: How I’d Build It

The MVP is deliberately small: pages for ten indie games with consenting developers, the structured submission form, voting, merging, and the license terms drafted by an actual lawyer, because the legal layer is the product and a template ToS from the internet does not count. No gamification at launch. Points and badges reward volume, and volume is the enemy until moderation matures.

The two real costs are not code. Moderation is a daily human expense that scales with success. And developer relations is a job in itself. Every indie studio has to be onboarded personally, every “declined” defended publicly, every early success story documented, because the site lives or dies on the credibility of the loop between a suggestion and a change in a real game.

I registered the domain and never built the thing. Time went first, then motivation, the two most ordinary killers in this whole business. The conviction stayed. If it is going to exist, it will be because someone reading this builds it.

Closing

SaveThisGame is a public, ranked, deduplicated backlog written by the people who play the game, made legally readable by the people who make it. The pieces are a weekend of CRUD, but the product is the license, the moderation, and the trust.

If you build this, tell me, and start with the indies. And if you are a small studio that would actually use this, say so publicly. Ten of you saying yes is the entire launch plan.

The Lectorum: A Spoiler-Free Terminal Into Warhammer Lore

Sun, 19 Jan 2025 00:00:00 GMT

The third site in this little constellation started, like the other two, with friends and a screen. We were admiring how Space Marine 2 renders the machinery of the Imperium. Those communication terminals, all cathedral and static, treat information retrieval as a religious act. Somebody asked the dangerous question. Could a website feel like that? Not a mockup, not a screenshot on an art station, but a real site you actually use, that stays coherent, respects the source material, and makes a fan want to come back and bring others.

At the same time we kept meeting the same person over and over: the newcomer. Warhammer 40,000 carries decades of novels, codices and games, and the answer to “where do I start?” is a minefield. Ask online and you get contradictory lists, and worse, you get spoiled on twists that took other readers years to earn.

So we built the answer to both questions at once. It is live at thelectorum.com, an unofficial, fan-made archive terminal for the Warhammer 40,000 setting. A personal reading order for starting without getting spoiled, factions, characters, a glossary, a star map, an imperial timeline, all of it presented as an in-universe machine, with easter eggs seeded along the way for the visitor who pokes at the edges.

The idea underneath: onboarding as a fan product

Here is the transferable part. Every deep franchise has a wall, the accumulated mass of canon that makes veterans rich and newcomers bounce. And every franchise’s community answers the wall with the same weak tools, wiki dumps that spoil everything and forum threads that argue about reading order. The wall is a permanent feature of big fictional universes, not a bug. So a dedicated, spoiler-disciplined entry portal is a permanent product surface, and almost nobody builds it properly.

The format is simple to state: pick a universe, define the spoiler line with editorial rigor, and build the door. Presentation in the universe’s own voice is what separates a door from a database. A newcomer who enters through an artifact that feels like the fiction has already started enjoying the franchise before reading page one.

Market & Opportunities

Franchise fandoms grow in waves. Every new game or screen adaptation throws a crowd of curious newcomers at communities that have no organized front door for them. Each wave rediscovers the same questions, and veterans burn out re-answering them.

The same shape exists far beyond Warhammer, in any long-running fantasy or science-fiction universe, in long manga runs, in decades-old comics continuities. A builder could serve one deep franchise as a labor of love, or turn the framework (spoiler-tiered content, reading paths, in-universe presentation) into a toolkit any fandom could deploy for its own wall.

Key Features

A personal reading order. The core promise in one page: where to start, what to read next, and what to postpone, with spoilers treated as a hard constraint rather than an afterthought.
Factions, characters and glossary. The reference layer for a reader mid-novel who needs to know a word without risking a plot twist in paragraph two of a wiki.
A star map and an imperial timeline. Spatial and chronological context, because orientation is half of what a newcomer is missing.
The terminal as interface. The whole archive speaks in-universe: the machine aesthetic is not a skin, it is the reason visitors share the site with other fans.
Easter eggs along the way. Rewards for the curious, threaded through the archive, so exploration itself becomes part of the experience.
Free, unofficial, and clear about it. Non-commercial fan work, plainly labeled, with a contact address staffed by, naturally, the Administratum.

The honest tensions

The commercial tension is the family trait. Games Workshop owns the universe, the site is free, non-commercial and unaffiliated, and its return is paid in community and craft rather than currency. The framework, though, is unencumbered. Applied to a universe you have rights to (or sold as tooling to franchise owners, who all face this exact onboarding problem), it is a genuine product.

The editorial tension is the spoiler line. “Spoiler-free” is a promise that gets tested by every single entry. What counts as context and what counts as a stolen twist? Holding that line demands the same discipline as the tone of the terminal itself, which brings us to the design tension, immersion versus usability. A site can look like a forty-first-millennium machine and still need readable text, working navigation and honest accessibility. The bet was making reverence and usability coexist, and that challenge was most of the fun.

Next Steps: How I’d Build It

Define the spoiler line first, in writing, before producing a single page. It is the product’s constitution and every future entry gets judged against it. Ship the reading order alone as the MVP. It answers the question newcomers actually ask, and it earns the right to grow. Add the reference layer next, then the map and timeline, and let the easter eggs accumulate the way they do in a real archive: slowly, in the corners.

Choose immersion last, not first. The in-universe presentation is the multiplier, but only on top of content that already works as plain, well-organized text. A beautiful terminal wrapped around a thin archive is a screenshot, not a site.

Closing

The Lectorum is a bet between friends that a website could feel like the machines we admired, turned into a real front door for the newcomers we all used to be. It hums at thelectorum.com. Walk in, ask it where to start, and keep an eye on the corners of the screen.

The door format is yours, like everything here. Pick the franchise whose wall you know best, draw the spoiler line, and build the terminal it deserves. If you build this, tell me. The Administratum answers faster than you would think.

Heresy Test: Twelve Questions, Five Verdicts, One Tribunal

Sat, 02 Nov 2024 00:00:00 GMT

This one also started with colleagues and a conversation that went exactly one step too far. The first question was classic fandom material. Which Warhammer character would each of us actually be? Easy, fun, twenty minutes gone. Then someone flipped it, and the flipped version would not let go. If the Inquisition audited our actual, present-day lives, our opinions, our little compromises, what would count as heresy? And how damned would each of us be, precisely, in percent?

We laughed a lot building the answer. It is live at heresytest.com, the unofficial Warhammer 40,000 heresy test. Twelve questions. The Inquisition already knows the answers; this is merely a formality, for the file.

I have shipped serious products with less joy than this quiz. Somewhere between writing the third verdict and watching a colleague dispute his percentage out loud, I remembered that fun is a legitimate spec.

The experience is played completely straight, which is where all the charm comes from. The site is styled as an Ordo Hereticus proceeding: case file, interrogation, verdict. You answer twelve questions and receive your alignment (Monodominant, Amalathian, Recongregator, Xanthite or Renegade, a spectrum lore fans will recognize immediately), a heresy percentage, and your classification from puritan to radical. No signup, no email, nothing between you and your damnation.

The part that makes it multiplayer

Solo quizzes are a solved genre. The piece we are proudest of is the tribunal, a shared room where up to forty people answer the same questions and come out ranked, publicly, from purest to most damned. That single mechanic converts a two-minute diversion into a community event. A Discord server runs a tribunal the way it runs a movie night, and the results feed an evening of accusations, appeals and screenshots.

That is the transferable idea, the one you can steal without knowing what a bolter is. Personality quizzes have always been shareable, but they have rarely been playable together. The moment results become a leaderboard inside a group that already knows each other, the quiz stops being content and becomes a party game.

The ranking does something subtle that solo results cannot. A percentage on its own is a curiosity; a percentage two points above your best friend’s is a conversation. The tribunal turns the quiz’s real product (arguing about the results) into the interface itself, and the arguments are the retention loop: the loser demands a recount, the recount needs a new room, and the new room pulls in three people who missed the first one.

Market & Opportunities

The personality quiz is one of the most reliably viral formats the web has ever produced, and it keeps working because the product is self-expression. People share the result, not the quiz. Fandoms multiply the effect, because the result doubles as a badge of belonging.

The underserved segment is groups. Communities on Discord and elsewhere are constantly hungry for low-friction shared rituals, things a moderator can launch tonight, that need no accounts, no downloads and no explanations. A themed quiz with a group verdict fits that shape, and the format transfers to any fandom or workplace culture with strong archetypes to sort people into.

Key Features

Twelve questions, played straight. Real-world situations filtered through an Inquisitorial lens, so the comedy comes from the framing, never from breaking character.
Five verdicts drawn from the lore. The alignments are the actual ideological currents of the setting, which rewards fans and stays legible to newcomers. The verdict card is the artifact people share.
A heresy percentage and a classification. Puritan to radical: a number to compare and argue about, which is what a good quiz result is for.
Tribunal rooms for up to forty. One link, everyone answers, the room is ranked from purest to most damned. The group result is a second shareable artifact, and the better one.
No signup, no email, no account. Free. The absence of friction is a feature, and for a fan work it is also good manners.

The honest tensions

The commercial one first, as always with this family of projects. Warhammer is Games Workshop’s IP, the site is explicitly free, non-commercial and unaffiliated, and it will stay that way. What you can take to market is the format: an in-character diagnostic with a group tribunal, wearing whatever universe you have the rights to, or none at all (an office “audit”, a cooking guild, a running club, because archetypes are everywhere).

The design tension is calibration. The quiz has to make people smile while respecting the canon, because the audience that will love it most is also the audience that will notice every liberty taken. Comedy that mocks the source dies with fans, while comedy that lives inside the source gets shared by them. Writing questions that map real life onto the lore without winking took more drafts than any of the code.

Next Steps: How I’d Build It

Write the verdicts before anything else. They are the product, the screenshot, the reason to share. Then work backwards to questions whose answers discriminate between them, and test the scoring on friends until the results feel unfairly accurate, because that feeling is the whole magic trick. Ship the solo quiz first. It is an afternoon of front-end once the writing exists.

The tribunal is version two, and its only hard requirement is being effortless: one shared link or room code, no accounts, results that render well as a screenshot. Resist every feature beyond that. The format is fragile precisely because it is small.

Closing

Heresy Test is a colleagues’ joke pushed to production quality, built together, and one of the most purely fun projects any of us has shipped. The tribunal awaits at heresytest.com. Bring your community, and may your file be thin.

The format is free, like everything here. Pick your universe, write five verdicts worth arguing about, and put a tribunal behind them. If you build this, tell me. My own file, for the record, is sealed.

From Collective Heroism to Individualism: Can Humanity Reclaim a Shared Vision?

Sat, 28 Sep 2024 00:00:00 GMT

Humanity’s history is short compared to the age of the universe, but it has gone through distinct phases of collective effort and of individual pursuit. There was a time when people united behind common causes: empires, explorations, great feats of architecture like the aqueducts, or knowledge advanced for the benefit of all. Whole societies achieved those things together, each person contributing to something bigger than themselves. Today, the focus has shifted sharply towards the individual.

Heroism used to be rooted in a collective spirit, in exploring unknown territory and pushing back the limits of shared knowledge. It has become something more inward-looking. Personal achievement, self-expression, and individual success now dominate our culture. Social media and the ultra-personalization of our experiences have put the spotlight on the individual rather than the collective.

Where does this leave us? Are we losing sight of the big projects that once defined our progress? There are challenges ahead that need the same kind of collective effort, especially if we think about our future survival and reaching for the stars.

Collective Heroism in History: The Power of Shared Ambitions

Earlier in human history, heroism was often tied to the collective. It wasn’t about individual glory but about being part of something larger. Ancient Greece, for instance, celebrated its heroes as much for what they represented for their cities and their people as for their personal achievements. The Olympic Games were more than competitions. They were a display of shared pride and of the strength of the city-states. In Rome, likewise, being a “citizen” carried a sense of duty and a commitment to the grandeur of the empire.

Collective heroism led to projects that seemed almost impossible, like the Roman aqueducts, massive undertakings on which the survival and well-being of entire populations depended. They are still awe-inspiring today. And yes, many of these monumental projects were built by the labor of people who had little choice but to serve.

Broad trends are useful, but history is rarely that simple. Take the Renaissance, a period celebrated for its collective cultural progress and the flourishing of art, science, and philosophy. It also saw the extraordinary work of individual geniuses like Leonardo da Vinci and Michelangelo.

These figures were models of personal excellence, and they also inspired collaborative work that pushed the boundaries of human achievement. Their talents complemented the collective spirit and showed that individual brilliance and communal progress are not mutually exclusive. The two can coexist, and society is richer for it.

People were motivated by their personal desires, but also by a sense of shared destiny. That common purpose made possible achievements no individual could have managed alone.

The Rise of Individualism: Shifting Priorities

The shift from collective heroism to individualism didn’t happen overnight. The Industrial Revolution started it, as people moved from rural, community-centric lives to urban centers where individual economic success became the focus. The trend continued as capitalism and consumer culture took hold. In that world, self-worth was increasingly tied to individual accomplishments and personal fulfillment.

The shift grew more pronounced over time, especially following the traumas of the 20th century. After World War II, society moved toward a model focused on individual fulfillment. This can be seen, in part, as a response to the failure of collective ideologies like fascism and communism, which led to immense human suffering. In the aftermath, personal freedom and happiness gained ground, and people became wary of large-scale collective movements.

The rise of neoliberalism from the 1980s onwards accelerated the trend further. The dominant narratives shifted from collective solidarity to individual success, and the emphasis from shared prosperity to personal achievement, to being able to “make it” on one’s own terms.

Technology, especially since the arrival of social media, has amplified individualism further. Algorithms cater to each person’s tastes, habits, and opinions, and build echo chambers that reinforce personal perspectives. With hyper-personalization, even shared cultural experiences now reach us filtered through individual preference.

It’s no longer about participating in a collective story, it’s about curating a personal one.

The Consequences of Individualism

The move towards individualism has brought undeniable progress. Personal freedoms have expanded, and people are, in many ways, freer to be themselves. But this focus on the self has come at a cost. We’re seeing a decline in the sense of shared responsibility, which used to be fundamental to the survival and growth of societies. With people focused mainly on their own lives, the idea of contributing to a greater collective purpose has faded.

One consequence is the weakening of social cohesion. As communities become more fragmented and the emphasis on personal success intensifies, feelings of isolation and loneliness have grown.

The research here is narrower than the slogan, and I would rather state it precisely than borrow its authority. A multinational study across 28 countries, covering 8,345 people, found that those who scored high on vertical individualism, the competitive kind that values standing out and getting ahead, reported more loneliness, while those who valued egalitarian social ties reported less. The effect is real but small, and it links values to self-reported loneliness. It does not prove that a culture makes its people depressed. That is enough for the argument I am making, and I would rather not stretch it further.

The constant comparison that social media encourages makes these problems worse, as people measure their worth against curated portrayals of others’ lives. This shift undermines social cohesion, and it erodes the support networks that mental well-being depends on.

The shift is visible in politics too. Debates are increasingly framed around individual rights and freedoms instead of common solutions, often at the expense of collective well-being. The current situation is all the more worrying because it can become fertile ground for extremist ideologies and strengthen convictions that lead to uprisings or even authoritarian regimes. History has shown what could happen.

The fragmentation extends to our very perception of reality. Digital media and personalized content create isolated bubbles where people only engage with information that fits their worldview. This has made it harder to face challenges as a unified society, because there is less common ground to begin with. The shared vision that once propelled entire nations toward greatness has given way to a multitude of smaller, conflicting goals, each tailored to individual desires.

The Future: Human Survival and Space Exploration

Technological progress is about more than convenience or innovation now. The survival of the human species depends on it. I’d argue that our future, as an inherently expansionist species, lies in the stars. To me, the conquest of space is the natural extension of our curiosity and our desire to explore new frontiers.

For centuries, humanity has dreamed of exploring space. Long before the first rockets left the ground, people would look up at the night sky and wonder what it would be like to journey among the stars. This collective dream spans hundreds, even thousands of years, yet today it seems to be fading, smothered by the growing focus on individualism. The desire to push the limits of human existence is dwindling, crowded out by immediate personal concerns and the narrow horizon of short-term politics.

Space exploration requires a unified vision and collective investment, an effort bigger than personal gain and passing interests. It’s not something a single individual or even a single country can achieve alone. It needs the kind of collaboration, funding, and shared ambition that once built empires or forged global alliances.

Few people today rank ambitions on that scale as a priority, and with enthusiasm this low, our ability to move beyond our planet is in serious danger.

A goal this monumental requires a commitment to building infrastructure, advancing shared knowledge, and providing good education and a decent quality of life. It would also require a rework of our culture, one where students are encouraged to pursue goals that contribute to national, and even global, progress, and where a meaningful career means having something to achieve for Humanity.

The International Space Station (ISS) shows what nations can accomplish when they come together with a shared vision. The orbiting laboratory is a marvel of international collaboration, with contributions from NASA, Roscosmos, ESA, JAXA, and CSA, among others. The global response to climate change is another example of positive collectivism, with countries uniting against a common threat through treaties like the Paris Agreement, even after the United States left it in January 2026. Community efforts in a crisis are a third, like the rapid mobilization of volunteers and resources after a natural disaster. They show that collective action can lead to outcomes that are both effective and compassionate.

Can We Rediscover a Collective Spirit?

The question we face now is whether we can revive that kind of collective spirit in an age increasingly focused on the individual. To reconcile individualism with shared ambitions, we might need to reshape our narratives and move from the pursuit of personal success to the pursuit of a larger, shared dream, like space exploration.

Suppose that, instead of viewing each other as competitors, we saw ourselves as partners in a grand project to take humanity beyond Earth. That shift in mindset would require major educational reforms, global cooperation, and perhaps even financial policies that invest directly in a shared future.

So the real question isn’t whether we still dream of the stars, but whether we can still do anything together big enough to reach them. Can a culture this good at celebrating the individual still rally around a project none of us could finish alone?

Corruption and the Need for Guardians of Virtue in a World Where AI Won't Save Us

Sat, 28 Sep 2024 00:00:00 GMT

Corruption and rampant selfishness seem to be everywhere. We see it when people or companies are willing to destroy natural resources, pollute rivers, or exploit others, all for their own profit. It is often about short-term gains, and the harm lands on specific groups of people, sometimes on entire regions, countries, or humanity as a whole.

We might have hoped that technology, and artificial intelligence in particular, would help us get past this. An impartial system that makes rational decisions, free from human bias, sounds ideal.

AI, however powerful it gets, is not the savior we hoped for. It is trained on our behavior and our data, and it carries the same biases and flaws we do.

It can also argue better than we do, under one condition. In a study published in Nature Human Behaviour in May 2025, GPT-4 debating strangers online came out more persuasive than the human opponent in 64.4% of the pairings where the two differed, but only when it had been handed basic demographic details about the person it was arguing with. Stripped of that personalization, it persuaded no better than we do. The machine on its own does not out-argue us. The machine that knows who you are does.

Now take a system carrying human biases and give it that advantage at scale, convincing millions of people that those biases are facts. That’s nightmare fuel for everyone not behind the algorithm.

Given these limits, the fix has to go past technology, to the root of the problem, which is human integrity.

What if we created a role in society for people with an incorruptible moral compass? Call them “guardians of virtue”, people whose integrity cannot be moved by money, power, or influence. Their job would be to oversee decisions in the industries and organizations that matter most, and to check that the big choices serve the common good.

Guardians of Virtue: A New Kind of Leadership

The idea may sound old-fashioned, like something from the days when knights swore oaths to defend the helpless and uphold justice. We may need a modern version of those knights. They would be chosen by the opposite of the usual criteria. Wealth, status, and connections would count for nothing, and moral integrity for everything. These would be people with a proven record of honesty and fairness, who do what is right whatever pressure they are under.

The guardians would oversee decisions in high-stakes fields such as environmental protection, weapons development, and social policy, where selfish decisions can do devastating damage. There they would be gatekeepers. Each major decision would be judged by what it does to humanity as a whole, and not by what it does for a few.

They would not hold absolute power. Think of them as a moral checkpoint. They would not replace CEOs, political leaders, or regulators, but they would have the authority to question a decision that threatens the collective good and, if needed, block it. Every position they took would come with a clear, detailed explanation, published for anyone to check.

The concept sounds radical, but it has precedents. Whistleblowers have shown that some people will still risk everything to uphold moral values.

Edward Snowden is often cited as an example. By exposing governmental overreach, he showed a deep commitment to transparency and accountability, and he paid a huge personal price for it. He is also deeply polarizing. Some see a hero who defended civil liberties, others a traitor who compromised national security.

The split itself shows how hard ethical calls are. Integrity is not always recognized or celebrated. We still need people who put the public good first when it costs them dearly, who can think their way through a hard moral dilemma and, when they must, stand up to the powerful.

Now imagine giving such people the tools and the authority to protect humanity’s interests where the biggest decisions are made. The guardians would live by the values we say we care about as a society: honesty, justice, and the protection of future generations.

Why AI Alone Cannot Save Us

As AI improved, many people believed it could fix the corruption and self-interest in how we make decisions. The reasoning was simple. A system that never gets tired, never craves power, and never gives in to greed would make an ideal referee for big decisions. Real AI is nowhere near that.

For now, AI systems reflect us. They are trained on data from our world, and that data carries our biases, prejudices, and mistakes. As a result, these systems often repeat those biases, and even reinforce them. We have seen AI behave in discriminatory ways, not because it “chose” to, but because it learned from biased data.

Then there is the influence of mega-corporations that have enormous power over our future. They often put profit before ethics, and their decisions can deeply affect society. AI can be very persuasive, so we have to ask who is behind the curtain. Which corporate agendas ride on AI’s recommendations? If we let technology set our direction with no ethical oversight, we may end up with a future run on self-interest.

The loop that closes on itself

A leading AI executive gathers biased data.
The data is used to train an AI.
The AI works under the influence of those biases.
Its recommendations guide decision-makers.
The decision-makers act on those recommendations.
Their actions reinforce the biases that shaped the AI.
The AI keeps learning from decisions made on its own advice, and the loop closes. Each cycle amplifies the biases and pulls society further from ethical judgment.

This is where the guardians come in. AI can process huge amounts of information, but it has no moral compass. Decisions about our society’s future need a final call from people with firm ethical convictions. The guardians would work alongside AI and add human judgment, guided by values instead of algorithms.

AI can assist. It cannot replace moral discernment, which only humans have. The guardians would keep AI a tool for good and stop it from amplifying the worst of human nature.

The Need for Change: Creating a New Framework

Our current systems have let corruption and self-interest creep in. We need a new framework.

It would start with carefully choosing people who hold themselves to the highest ethical standards. The vetting would be as rigorous as for senior posts in government or large companies, with moral integrity as the first criterion. Candidates would have to prove their commitment to ethical principles, and show they can handle the hard moral dilemmas the job would throw at them.

They would also need the tools for the job: open access to information, a platform to raise concerns, and the authority to challenge decisions that could cause harm. Part of the role would be building a culture of integrity inside the organizations they oversee, so that other people put ethics first in their own work.

I can already hear the obvious objection. An unelected group with moral authority sounds uncomfortably close to the governance models we usually warn each other about, a select elite deciding what is good for everyone. History is full of people who concentrated power in the name of virtue, and it rarely ended well. The objection is fair, and the answer has to be part of the framework itself, not a footnote.

That’s why the guardians could never hold unchecked power. Their role is a moral checkpoint, not a throne. They would be held to the standards they enforce, with regular public scrutiny, oversight by independent bodies, and a clear, evidence-based justification for every action they take. Their decisions would remain open to challenge. The point is to strengthen democratic processes, not to replace them, by adding moral integrity to decisions with far-reaching consequences.

A guardian who starts ruling instead of guarding has already failed the job description.

Last, the guardians would need public support. A society that values moral leadership can shift its norms toward accountability and transparency. We would have to celebrate the people who stand up to corruption and speak for the common good, the way we honor whistleblowers today.

A Call for Collective Reflection

Better systems are only half of it. The other half is living up to the values we claim to hold. The guardians are not a way to hand our trust to a select few. The idea behind them is that every level of society needs ethical leadership.

How do we make sure the decisions that shape our future rest on integrity and transparency? How do we, as individuals and communities, build a culture that puts the common good ahead of self-interest? These are practical questions, not just a philosophical exercise.

The answers may not be easy, but the conversation has to start now. Perhaps it is time to rethink who leads, and what we mean by leadership in the first place.

Samewave: Find Your Musical Twins

Fri, 13 Sep 2024 00:00:00 GMT

In September 2024 I typed into an AI a question that had been itching at me for a while. Is there a website where you connect your music profile and it finds the people who listen to exactly what you listen to? The answer was no, not really. That itch is the product, so let me name it. I wanted to know who else is like me, and how many. There is something thrilling about that question. Am I as unique as I secretly think, or do millions of strangers share this exact love? Either answer would be a gift. Rare is flattering, legion is comforting.

Samewave is the machine that answers it. You connect your listening history, and the first screen shows a number on a world map: how many people on Earth share your top artists, and where they are. Thirty-seven humans have your exact top three. Two of them are in your city, and the rest glow softly across a heatmap of countries. That number, your twin count, is the product. Everything else (chat rooms with your twins, country-by-country comparisons, discovering what your Brazilian doppelgänger listens to that you have not found yet) hangs off that one magic moment.

The number, and the arithmetic that threatens it

Let me get the arithmetic out of the way before it flatters anyone. A top-three artist combination is nearly a fingerprint. With millions of artists in the catalog, exact triple matches will be rare, and for listeners with eclectic taste the exact-match answer will often be zero. An app whose magic moment can display “0 people are like you” has designed its own funeral, unless zero is made to feel like a crown. You are the first of your kind, so plant a beacon and the map will alert you when your twin arrives. I like that framing a lot, but it cannot carry the whole experience.

So similarity has to come in tiers, and the tiers are where the matching actually happens. There are exact twins (same top artists), close kin (two of three shared, or the same artists in a different order), and same constellation (heavy overlap across your top twenty, the people whose library rhymes with yours even if the podium differs). Each tier has its own count and its own room. The twin count is the fireworks. The kinship tiers are where the actual humans are.

The wall I hit before I even started

Now the part regular readers of this site will recognize. My September 2024 notes cheerfully specced this on the Spotify API: OAuth login, top artists endpoint, a developer brief ready to send. Two months later, in November 2024, Spotify slammed the door on this kind of project, deprecating a raft of endpoints for new apps and, more fatally, keeping new integrations in a development mode capped at a couple dozen whitelisted users, with extended access now granted about as often as sainthood. A consumer social app that requires Spotify login is, for an independent builder today, structurally unlaunchable. I wrote about the two laws of building on platforms in my EmoTune article, and I will not re-derive them here. This idea got hit by both before a single line of code existed.

This time, though, the pivot improves on the original plan.

The pivot: build it where listening is already public

The scrobbling world (Last.fm and the open ecosystem around it) is the anti-Spotify in every way that matters here. Listening histories are public by default and two decades deep. The community already has the culture this product needs, people who track and share their stats for the love of it. And nobody can revoke the ground it stands on.

One caveat, and it fills the entire first sprint. Last.fm’s API answers questions about a user you can already name. No endpoint hands you the crowd, and scraping your way to one runs straight into the terms of use. The corpus has to come from somewhere else, and it can. ListenBrainz publishes its listening data in bulk, openly licensed, so that people can build on it. Samewave starts from the ListenBrainz dumps and lets anyone connect a Last.fm or ListenBrainz profile to place themselves on the map. The map is not empty on day one, but filling it is the first real piece of work.

And there is a delicious irony to reclaim. Last.fm itself once had a feature called musical neighbours, the people whose taste sat closest to yours. It was, for many of us, the best social feature the music web ever shipped, and it was quietly buried years ago. Samewave robs that grave with pride. The neighbours come back, this time with a world map, a twin count, and rooms to actually talk in.

Homegrown maps where musicians come from. Samewave maps where their listeners are. Same open data, opposite question, and I would happily see one person build both.

The graveyard beyond that is instructive too. Tastebuds matched people by taste and died as a dating app. Spotify’s Blend compares you with friends you already have, which is the reverse of this idea, since Samewave exists to find the strangers. Stats apps show you your own numbers in a mirror, and this one turns the mirror into a window. Two neighbours come closer than anyone else. ListenBrainz already computes similar users from this same data, which is this article’s kinship tier without the map, the count, or the rooms. And Obscurify already tells you how obscure your taste is, which is the uniqueness card without the other humans. Samewave joins two features that I think were never meant to live apart.

Key Features

The twin counter. One number, front and center: humans on Earth who share your top artists, with the map showing where. Updated live as new people join.
Kinship tiers. Exact twins, close kin, same constellation, each with its count, its map layer, and its room.
The uniqueness card. A shareable answer to the founding question: how rare is your taste, as a percentile and a count. It is also the growth engine. Every card shared is an invitation to be counted.
Twin rooms. Small chats scoped by tier and artist constellation, rather than giant genre forums. What makes them work is the eerie intimacy of strangers who share your exact library.
Country lenses. Zoom into any country and see its top artists next to yours, and how many of your kin live there. Travel planning for the ears.
Privacy at country grain. Location is never finer than the country level, visibility on the map is opt-in, and lurking anonymously with just your counts is a fully supported way to use it.

The honest tensions

The pivot trades one problem for a smaller one. The scrobbling community is devoted, but it is a fraction of Spotify’s audience, so Samewave starts deep rather than wide. I think that is the right trade. A live map of a passionate million beats an unlaunchable app for half a billion, and manual import paths can bridge the gap for anyone willing to carry their data over. But the ceiling is real and should be said out loud.

Chat needs moderation from day one, light but present, because “strangers matched by intimacy” is a lovely product and a spam vector. And the twin counter must resist the temptation to inflate. The entire emotional payload of this product is that the number is true, so the moment the numbers feel massaged to avoid zeros, the magic dies.

Monetization

Modest, and it should stay that way. Counting and being counted stays free forever. A small supporter tier unlocks map history and extra lenses. There are no ads, because an app about belonging that sells your attention to advertisers has misread its own soul. This is closer to a beloved community instrument than a venture rocket, and it should be built by someone at peace with that.

Next Steps: How I’d Build It

The MVP is almost embarrassingly direct, and it does not even need accounts. It is a landing page where you type your three favorite artists and get the count, computed from public scrobble data, with the map lighting up behind it. That single page answers the founding question and shares itself. Accounts, tiers, and rooms come only after it has proven the itch is shared.

Then connect the open scrobble APIs properly, compute similarity across top-twenty libraries and not only podiums, and seed the first twin rooms by hand around a few artist constellations with famously devoted listeners. The hard parts are keeping the numbers honest, the rooms warm, and the map alive as it grows. None of that is a technical problem.

Closing

Samewave is an instrument for the question I could not stop asking. Am I one of a kind, or one of millions? Both answers are wonderful, and for the first time a machine could tell you which one is true, with a map, a number, and a room where your twins are already talking. It joins the other music ideas on this site, and this one takes the song as a handshake.

If you build this, tell me, and tell me my number. I have been wondering for years.

Peel: Dating One Layer at a Time

Sat, 22 Jun 2024 00:00:00 GMT

This idea comes from years of the same conversation, had over drinks with single friends and lived a little myself. The swipe is broken, and everyone in it knows it is broken. A face gets half a second. The photo decides everything before a single word is exchanged. Matches pile up that nobody ever messages, a drawer of maybes kept for a rainy ego. And the friends who complain the loudest about being judged on their photos will admit, two drinks later, that they judge the same way, because the interface gives them nothing else to judge with.

One evening in early 2024 I started sketching what the opposite would look like, and the answer was not “an app with better photos.” It was an app with no photos at all, where finding out who someone is works the way it does in real life: gradually, mutually, and with permission.

Peel is a dating app built as a stack of layers. A profile starts nearly closed: age range, city, and a first thin slice of who the person is. Everything else (work, passions, values, lifestyle, sense of humor) sits in sections underneath. To see someone’s next layer, you request it. They see your request, and your own visible layers, and they choose to reveal it to you, or not. And the mechanism I care most about is that, at every layer, the interface forces your hand. You request the next section, or you let this person go. No hoarding, no limbo, no drawer of maybes. Curiosity has to be mutual and it has to keep moving, or it ends, cleanly.

Why this might actually work

Three design consequences fall out of the layer mechanic, and together they are the product. First, attention becomes expensive, in the good way. Every reveal is a small mutual yes, so a conversation that reaches layer five means ten explicit acts of interest have already happened, which is ten more than a swipe ever contains. Second, refusal is information. Someone who declines to show you their passions layer has told you something useful, at zero cost to either of you, and you move on without resentment. Third, the superficial signal is simply absent. You cannot judge the photo first, because there is no photo, and by the time faces are exchanged you already know whether the person underneath is interesting.

The other half of the concept attacks the ratio, a structural disease the swipe economy pretends not to have. Most dating platforms are heavily male-skewed, which buries women in low-effort attention and drives them off the platform, which worsens the ratio, which is the death spiral. Peel admits users on strict parity within each pool. In the straight pool, where the imbalance lives, one place for a man opens when a woman joins, and vice versa, with a live waitlist and your real position on it. Other pools get their own admission logic. Yes, that means men will wait. I will not apologize for it. The wait is the mechanism that makes the inside worth reaching, and the waitlist number is honest scarcity where boosts and paid visibility are manufactured noise.

The graveyard, and the uncomfortable lesson in it

Photo-less and photo-blurred dating has been tried: apps that unblur a face as you chat, blind-date formats, puzzle-covered portraits, progressive-reveal experiments. Most stayed niche or died. The lesson is not that the ideas were bad, but that people’s stated preferences and revealed preferences diverge. We say we hate superficial swiping, then we open the app with the fastest dopamine. Any anti-superficial app is betting against its users’ own reflexes, and that bet must be priced in, not wished away.

Peel’s answer to that lesson is structural rather than moral. The parity gate creates scarcity, and scarcity changes behavior. An app you waited for is an app you do not treat as a slot machine. The forced choice at every layer keeps the loop moving, so slow dating never means stalled dating. And the layer requests themselves are the dopamine. A knock on the door is more exciting than a hundred faces, because it had to be asked for. Maybe that is not enough and the reflexes win anyway. That is the risk, stated plainly.

Key Features

Layered profiles. A curated sequence of sections, from light (passions, lifestyle) to deep (values, what you are actually looking for). Order matters and is part of the design: the layers should read like a good first conversation.
Consent per layer. Every section is revealed person-by-person, on request, by its owner. Nothing about you is browsable by default; everything is granted.
The forced fork. At each layer: request the next, or release this person. The interface has no “keep for later.”
Parity admission with a live waitlist. One-for-one balance within each pool, a real-time position counter, and no way to buy your way past it.
Private verification, public anonymity. Identity and age are verified privately at signup, precisely so that the public side can stay anonymous without becoming a catfish farm. An app of hidden faces only works if the platform knows exactly who everyone is.
The endgame, designed. The final layer is the mutual reveal: photo and first name, exchanged simultaneously or not at all, followed by the actual point of the entire machine, a date. Peel should measure its success in conversations that leave the app, not in time spent inside it.

The two versions I killed

My original notes contained two more features, and this article exists partly to bury them in public. The first was a community rating system that sorted users into pools by score. Written down plainly, that is a caste system, the Black Mirror episode everyone has seen, operated for real. The charming get gilded pools, and the awkward get sorted into a basement they can never see the walls of. Even my 2024 self flagged it as “probably the biggest problem.” It was. Moderation against abuse, absolutely. A number attached to a human being’s desirability, never.

The second was net worth as a profile field, and it deserves the same burial, because it exposes a contradiction I did not see at first. I was removing photos to fight superficiality while installing a money signal in their place. That is not less superficial, just uglier, and it points the whole app toward transactional dynamics. Worse, hidden identities plus wealth signals is the exact recipe of romance scams. Profession and field, yes, they are part of who someone is. Bank balances, no.

Data, law, and the tensions that remain

Dating data is sensitive by nature (sexual orientation is inferable from the product itself), so Peel has to be built on data minimalism as an architecture, not a settings page. That means no religion or politics fields at all (my early notes wanted them, but European law and basic prudence say otherwise), income never stored, location coarse by design, city-level and no finer, and layers encrypted so that “revealed to one person” never quietly means “readable by the platform’s ad partner.” An app whose entire premise is consent cannot have a privacy policy that laughs at it.

Monetization has one red line and a simple model. The red line is never selling the imbalance back. No paid waitlist jumps, no boosts, no purchasable extra requests, because every one of those monetizes the pathology the app exists to cure. The model is a flat membership once admitted, same price for everyone, with the free tier generous enough to prove the experience. Scarcity apps have made membership work before, and this one has the advantage that the thing being paid for, a balanced and intentional room, is structurally real.

Next Steps: How I’d Build It

Start with one city and one gender-balanced waitlist, not a global launch. Five layers, the forced fork, private verification, and no photos anywhere. The first hundred admitted pairs tell you whether people who say they want depth will keep requesting the next layer, or bounce the moment a face is not on offer.

The hard parts are not the stack. They are the waitlist honesty (a real position, never a fake queue), the verification vendor that does not become a data lake, and the discipline to keep the red lines: no boosts, no paid skips, no scores. If those three hold after three months in one city, widen. If they do not, the bet is lost and the article above already said so.

Closing

Peel rebuilds dating as a mutual, consenting unveiling instead of a photo auction, with parity at the door, a forced choice at every layer, and a face saved for the moment it belongs to, the end of the beginning. The pieces that had to die are buried above, in public, where dead features belong.

If you build this, tell me. I want to know whether the bet holds, whether people who say they want depth will choose it when the interface finally offers it. And if you find out they will not, tell me that too. A disproven bet on human nature is worth an article of its own.

Want2Hear: Fund the Human Voice for Books That Deserve One

Thu, 16 May 2024 00:00:00 GMT

This one started on my bookshelf. I keep a mental list of books I love that deserve to be far better known than they are, the ones you press into a friend’s hands saying trust me. And I kept noticing where readers actually are now. They are in their ears. Audiobooks are how a huge share of people consume books today, on commutes, on runs, while cooking. For a book I love, not existing in audio increasingly means not existing at all for a whole population of potential fans. In 2023 that observation turned into a project called Want2Hear, a community platform where readers would vote for the books they wanted turned into audiobooks. I liked it enough to buy the domain, want2hear.io.

I no longer own it. I let it expire, which tells you how far I got. But the idea refused to expire with it, mostly because the world changed around it in a way that makes the corrected version better than the original.

The premise that dissolved

The 2023 version rested on scarcity: narration was expensive, audio catalogs were a bottleneck, and a big enough pile of votes might convince a publisher to greenlight a recording. Then AI narration arrived like a tide. Apple shipped AI-narrated books, Audible opened virtual-voice narration to self-published authors and synthetic titles flooded in by the tens of thousands, and the cost of turning text into audio collapsed toward zero. The question my platform wanted to influence, “will this book get an audiobook?”, is quietly becoming “yes, a robotic one, by default.”

A voting wishlist to trigger conversions makes little sense when conversion is free. If I had built Want2Hear as designed, the tide would have drowned it. So the right move is the one this site exists for: publish the idea with its flaw named, and with the pivot that survives it.

What actually remains scarce

Two bottlenecks survived the tide, and they are the real product. The first is rights. An enormous number of books have no audio edition not because nobody wants one, but because no audio deal was ever signed. Demand that is visible and aggregated can unlock decisions that individual emails never will. The second is human narration, the thing the tide itself made precious. A great narrator is a performance (casting, pacing, character), and precisely because synthetic voices became abundant, the human version became the premium object. The scarcity did not disappear. It moved.

And there is a structural flaw in voting itself that my 2023 notes ignored. Votes are free, and free signals are ignorable. No publisher finances a recording because a webpage counted upvotes. The fix is as old as commerce. Do not ask people what they want to hear. Ask them what they will pay to hear.

The pivot: pledges, not votes

Want2Hear, corrected, is a crowdfunding platform for human-narrated audiobooks. Instead of upvoting a book, you pledge to pre-buy its audiobook, with the money escrowed and charged only if the campaign succeeds. When a title crosses its threshold, the platform arrives at the rights holder’s door not with a petition but with a firm order book: a thousand confirmed buyers, funding secured, narrator shortlist attached. Demand stops being a survey and becomes an argument no spreadsheet can wave away.

Around that core, the community layer my original notes wanted finally earns its place. Readers champion the overlooked books they love, which was my founding itch. And they fan-cast the narrator, “who should read this book?”, which is both a delicious community game and a useful casting signal. The positioning writes itself in one line: in a world where every book gets a robot voice for free, the thing worth funding is the human one.

The graveyard, briefly

Free-signal versions of this idea already exist and already prove the point by failing at it: suggestion boards, endless Goodreads lists, request threads on forums, all shouting into the void, because none of them attach money to the wish. Meanwhile, individual authors have repeatedly crowdfunded audiobook editions of their own books, campaigns that worked precisely because backers paid rather than voted. The model is validated at the level of single campaigns, but nobody has generalized it into a standing platform where the community, not the author, starts the campaign. That gap is the whole opportunity.

Key Features

Pledge campaigns with escrow. Pre-orders, not donations: pay only if the audiobook actually gets made, receive the audiobook when it ships. Thresholds are public and honest about what professional narration costs.
Two-stage campaigns. Stage one is a free demand signal anyone can start for any book. Stage two, where money moves, opens only when the rights holder opts in, because collecting pledges for something you have no right to deliver is how platforms die in court. The stage-one heat map is itself the tool that recruits rights holders.
Narrator fan-casting. Communities propose and vote on narrators; narrators can claim a profile and signal interest. Casting becomes part of the campaign’s story, and a beloved narrator attached to a campaign is rocket fuel for pledges.
A home for the overlooked. Ranking and discovery deliberately favor the underexposed: the cult favorite, the untranslated gem, the out-of-print classic. Bestsellers do not need this platform; that is the point of it.
Rights-holder dashboard. For publishers and indie authors: see aggregated demand for your catalog, open a campaign in a click, keep the relationship with your audience direct.

The honest tensions

Rights are the hard part, and the two-stage design only manages the problem without dissolving it. A rights holder can always say no, or worse, say nothing, and a passionate stage-one campaign that never gets its stage two will frustrate the exact community the platform depends on. Expectation-setting has to be built into the interface, because a pledge page is a request, not a promise.

The economics are unforgiving. Professional narration of a full-length book costs real money, so thresholds will sit in the thousands, not the hundreds, and many campaigns will fail. That is fine (Kickstarter’s whole history says a platform survives failed campaigns), but only if failure is cheap and transparent for backers.

And the AI question deserves a straight answer. Plenty of listeners are perfectly happy with a synthetic voice, and this platform is not for them, which is fine. It serves the subset who hear the difference and will pay for it, the same population that buys vinyl, hardcovers, and letterpress. Every abundance creates its own premium, and Want2Hear just gives this one a checkout page.

Next Steps: How I’d Build It

Start where the rights bottleneck does not exist, with independent and self-published authors, who own their audio rights outright and can opt in with a signature. One genre community with strong audio culture (fantasy and science fiction are the obvious candidates), ten hand-picked books whose authors said yes before launch, and the first campaign run half-manually with an existing narrator marketplace handling production. The MVP is one funded audiobook, delivered, with everyone paid. That single proof recruits the next hundred authors better than any pitch deck.

Then comes the flywheel. Stage-one demand pages for beloved rights-locked books are used as evidence to bring publishers in, and the narrator community is grown deliberately, because in the long run the platform’s moat is not its code but being the place where readers, narrators, and rights holders already meet.

Closing

Want2Hear is my 2023 voting site, turned by the tide that would have sunk it into a machine that converts love for overlooked books into funded human voices. I let the domain lapse. The idea is yours now, and the name is optional too.

If you build this, tell me. I will arrive with my list of books that deserve more readers, pledges ready. And if you are a narrator or an indie author reading this, you are the supply side of something readers demonstrably want, and one of you should start it.

Emperor's Light: The Only Social Network of the 41st Millennium

Sun, 14 Apr 2024 00:00:00 GMT

This one was born the way the best bad ideas are born, between friends, on a night when the conversation drifted somewhere it should not have. Someone asked what a social network would look like in the year 40,000. We took the question seriously for about nine seconds. Then the obvious answer landed, and it was better than anything a product team could have designed. There is no social network in the 41st millennium. There is exactly one, and its entire purpose is to glorify the Emperor.

No profiles. No follower counts. No brand accounts doing brand things. One feed, one topic, one approved sentiment, forever. The joke stayed between us for a long time, resurfacing whenever one of us needed it. And then, because some jokes deserve better than a group chat, we built it. It is live at emperorslight.com, a shrine of the Emperor of Mankind, where pilgrims submit their vows and the faithful browse each other’s devotions.

The vows are the whole product. A visitor writes a pledge or a confession, claims an allegiance, and joins the wall of the faithful. The tone is the trick. The site speaks the ornate, reverent language of the Imperium and applies it, dead serious, to the mundane frustrations of the hobby. Half devotion, half confession, entirely in character. Two thin coats. Always two thin coats. The Emperor demands proper paint consistency.

The best product spec we had was a sentence from the original conversation: in the grim darkness of the far future, there is only one feed. Everything the site refuses to have is in that sentence.

The idea underneath: one feed, one purpose

Strip the power armor off and there is a transferable product idea here, the part you can take even if you have never rolled a die in your life. Social platforms compete by adding: more formats, more surfaces, more reasons to stay. A devotional feed competes by removing. One post type. One shared register everyone writes in. No metrics arms race, because the point is belonging rather than reach. The constraint is not a limitation of the product; the constraint is the product.

Every fandom, and for that matter every community with a shared obsession, has a devotional act its members already perform in scattered places: the gym log, the sourdough crumb shot, the finished miniature, the marathon medal. A single-purpose shrine gives that act a home and a liturgy. Big platforms cannot manufacture a room where everyone is there for the same reason. This format produces one, and it costs almost nothing to run.

Market & Opportunities

Warhammer 40,000 is one of the largest tabletop and transmedia franchises in the world, with a fanbase that keeps growing with every video game and adaptation. What interests a builder more is that it is a fandom with a strong culture of in-character writing, and its communities live scattered across forums, subreddits and Discord servers that all lack a shared public ritual.

The wider opportunity is the same one, generalized. Niche, single-purpose community sites keep proving that a small surface with a strong identity can hold attention that giant platforms lose. The audience is any community whose members already perform a repeated, shareable act of devotion, and whose inside jokes deserve better production values than a chat thread.

Key Features

Vows as the single post type. A short pledge or confession, written in the register of the Imperium. One format keeps the wall coherent and makes every entry legible in one glance.
Allegiance on every vow. Each pilgrim signs under a chapter or faction, which turns a text feed into a mosaic of identities without needing profiles at all.
Ways to browse, nothing algorithmic. Recent, trending and most blessed, and a filter by chapter if you want one. No algorithm to argue with, no infinite personalization, and the best vows rise the way they should: because the faithful liked them.
A tone that never breaks. Every label, button and error message speaks Imperial. The immersion is the moderation: content that breaks character is content that does not belong.
A constellation, not a silo. The shrine links to its allied fan works (a heresy examination, a lore archive), so the joke opens doors instead of closing them.

The honest tensions

The first is the one you already guessed. None of this can ever be a business. Warhammer belongs to Games Workshop, the site says so plainly, and it is free, non-commercial and unaffiliated by design. So why does it sit in a catalog of business ideas? Because the format underneath is yours to monetize in your own universe, and because this site has always been honest that some projects pay in something other than currency. This one pays in community, in craft, and in the rare pleasure of a joke sustained properly, at production quality, with friends.

The second tension is the tone itself. Reverent parody is a knife’s edge. Wink too hard and it collapses into memes, play it too straight and it reads like a real cult. The discipline is editorial, not technical, and it is permanent. Respect for the source material is what earns the laughter of the people who love it.

Next Steps: How You’d Build Yours

Start by finding your community’s devotional act, the thing members already do and already photograph or write about. Then design the liturgy: the single post format, the register everyone writes in, the identity marker that replaces the profile. Build the smallest possible wall (submit, browse recent, browse loved) and launch it into one community you actually belong to, because the first fifty entries set the tone for everything after. Moderation is the real engineering. A single shared register is also your content policy, and it needs hands on it early.

Technically this is a weekend-friendly stack, with a static front, a tiny backend for submissions, and most of your budget spent where it belongs, on words and art direction.

Closing

Emperor’s Light is a joke between friends that refused to stay one, now collecting vows at emperorslight.com, built together with the same friends who started the conversation. Go pledge something. Two thin coats.

And there is a part of the idea we left on the table, free like everything else on this site. It is the dark mirror, the other feed, the one that whispers instead of glorifying. Same mechanics, opposite liturgy, and every fan knows exactly which four flavors it comes in. We may build it one day. You may beat us to it. If you build this, tell me.

Pact: If We Are Both Single at 40

Sun, 18 Feb 2024 00:00:00 GMT

Let me open with a confession that doubles as market research. Over the years, I have made at least five of these pacts. You know the one. A late dinner, a certain kind of laugh, and someone says it, half joke, half not, while looking slightly past you: if we are both still single at forty, we marry each other. Deal. Everyone laughs, nobody writes anything down, and the pact evaporates by morning, surviving only as a warm little clause in the back of two heads.

Five different people. Five pacts. I am, apparently, heavily over-subscribed on a contract that nobody has ever drafted. When I dug those memories up, I realized I would like to reactivate them, and that I have no mechanism to do it. Which phone number? Whose memory? Who brings it up first, and how, without it being weird? The pact is one of the most widely signed agreements in human culture, and it has a one hundred percent enforcement failure rate.

Pact is the app that takes the joke seriously. Two people seal their pact for real: names, birthdates, the trigger date, both signatures, today. Then it sleeps. Years. On the day, each of you gets the same quiet question, are you single, and are you still in? Only a double yes opens a private room and lets the conversation begin. Any other combination, a no, a silence, a pact defused years earlier, produces nothing at all, forever, for both of you.

The silence is the product

Every design decision follows from one rule. Nobody ever learns what the other answered, unless both said yes. That double-blind is the entire emotional architecture. It is what makes it safe to say yes at forty (you risk nothing, a lonely yes is never exposed) and safe to say no (you break nothing, your no is indistinguishable from the service never asking). Silence is deliberately ambiguous. Maybe they said no, maybe they dissolved the pact in 2031, maybe they never saw the message. Ambiguity is mercy.

The same rule shapes the rest. The pact is mutual at creation, always. Both parties sign at sealing, so nobody is ever secretly named, and nobody discovers at forty that they starred in someone’s plan for twelve years without knowing. And either party can defuse the pact at any moment, silently, with no notification and no guilt trip. The trigger day simply never comes. People marry, drift, change. The exit has to be as gentle as the entry is playful.

The prior art makes the positioning easy. Stanford’s Marriage Pact, a real and clever company, runs the inverse concept. A questionnaire assigns you an algorithmically optimal backup stranger on your campus. Pact holds no algorithm and offers no strangers. It contains exactly one person, chosen by you, already loved in whatever complicated way produced the joke in the first place. It is the anti-dating app. Zero discovery, pure commitment infrastructure.

Key Features

The sealing. A small ceremony by design: both confirm identity and birthdate, read the terms aloud (the app insists), pick the date, sign. A pact deserves more gravity than a checkbox. Several pacts are allowed (I am not judging; I am the market); the app staggers their trigger dates and enforces one pact per pair. If two of yours fire in the same season, that is not a bug, that is your third act.
The sleep, with a heartbeat. Every two years, a gentle re-confirmation: still in, still at this address? It refreshes contacts, revalidates consent, and proves the service is still alive.
The trigger. On the day: one question to each, separately. Double yes opens the room. Anything else is eternal, indistinguishable silence.
Silent defusal. One tap, any time, no notification to the other side. The kindest exit in software.
The certificate. At sealing, both receive a beautiful document, print or PDF, that contains the entire pact: full names, birthdates, date, terms. Not a pointer to the app; the protocol itself, executable by two humans with no company in between. And a lucid note on what it carries: contact details rot in twenty years, so the durable payload is identity and, above all, the mandate, we signed, you have the right to come looking for me. Each heartbeat reissues it with fresh coordinates.
The witness. Pacts are born in groups, so the sealing invites one: a lifelong friend, a sibling, the person laughing at the table when the joke was made. Ritually, they countersign. Practically, they are the disaster-recovery layer no technology matches, a human who can still find both of you in fifteen years, long after every phone number on the certificate has died.
The meeting clause (optional). The cinematic tier: at sealing, a pair can also fix a place and an hour. Not a café, cafés close; a landmark, a fountain, a bridge, the kind of spot cities keep for centuries. If everything else has rotted, the pact still executes the oldest way there is: on the date, at noon, you simply go, and presence is the yes. One warning comes with it, printed on the certificate: this clause trades the double-blind for romance, because standing there alone is a reveal. Some pacts will want the discretion. Some were always going to want the fountain.

The honest core: promising something twenty years out

The real engineering problem has nothing to do with code volume. Pact’s flagship feature fires ten to twenty years after purchase. Companies do not reliably live that long. Every piece of standard infrastructure (servers, credit cards, app stores, domains) assumes a company that keeps paying attention. My first instinct was a dormant server with a cron job, prepaid for two decades, and it does not survive contact with reality. Nobody sells twenty reliable years of anything.

So the architecture inverts the problem instead of fighting it. First, never make a twenty-year jump. Make ten two-year hops. The biennial heartbeat converts one heroic delivery into a chain of trivial ones, each within the competence of ordinary email. Second, plant the date in infrastructure more durable than any startup. At sealing, Pact drops the trigger day into both parties’ calendars, and people carry their calendars across decades and platforms the way they carry their names.

Third, the disaster plan is honest about what actually survives two decades. Phone numbers, emails, and addresses all rot. What survives is identity (a full name and a birthdate remain findable in a connected world), the witness (a human anchor who knows both parties), and the mandate the certificate confers (the signed permission to go looking without it being strange). A degraded pact, service dead, contacts stale, is still executable by a determined forty-year-old with a name, a witness, and the right to search. And for the pairs that chose the meeting clause, the final fallback needs no search at all: a place, an hour, and whoever shows up. Fourth, the money matches the physics. A single sealing fee is priced knowing that a sleeping pact costs pennies per decade, and the delivery daemon is published as open source so that a dead company is an inconvenience rather than an annulment.

The one job that cannot be decentralized is the double-blind itself, and it is the moat. Paper reveals unilaterally, and a calendar reminder tells you nothing about the other side. Only a trusted intermediary can hold two sealed answers and speak solely when they match. Pact’s server is, in the end, one small sacred function, the discreet friend who only talks if you both said yes. Everything else is redundancy. (And no, not a blockchain. A promise this tender deserves better than being immortalized next to monkey JPEGs.)

One more admission, about metrics. Most pacts should never fire, and that is success. People marry, happily, elsewhere. The pact’s job was to be a warm clause in the back of the head, and the app’s job is to hold it without ever becoming a threat to the lives people actually build. A product whose best outcome is its own silence is a strange business, and the sealing-fee model exists so that silence never starves it.

Monetization

One-time sealing fee per pact, positioned as what it is, a small ceremony with a decades-long tail, not a subscription (subscribing to something designed to sleep for twelve years would be absurd, and worse, would make the pact die with a lapsed card). Premium tier for the printed certificate, letterpress-grade, because this is the WaveArt instinct again. A promise this analog deserves an object. Anniversary touches, optional and tasteful, a yearly “your pact is sleeping soundly” postcard, can exist as gentle upsells for the sentimental. That is all. A product built on trust and silence should have the least clever pricing imaginable.

Next Steps: How I’d Build It

The MVP is deliberately tiny: sealing flow, the pact record, the biennial heartbeat, the calendar seed, the certificate generator, and the trigger-day double-blind exchange. A weekend of code and a decade of patience. The hard work is all wording, in the sealing ceremony, the heartbeat emails, the trigger question, and above all the silence. Every message has to be drafted knowing it may land in the middle of someone’s marriage, someone’s grief, someone’s second thoughts. This is a product where the copywriting is the engineering.

Start with the people who already have a pact, which, if my own inventory of five is any indication, is a bigger market than it looks. The launch motion writes itself: an article like this one, shared with the exact person you once made the joke with, no message attached. They will know.

Closing

Pact finally gives the most-signed unwritten contract in the world a vault, a heartbeat, a trigger, and the discretion it always needed. Sealed as a joke, held like it matters, silent unless you both mean it.

If you build this, tell me. I have five reactivations pending, and I am curious which of them says yes.

Stackpage: Your Toolbox, Made Visitable

Thu, 21 Dec 2023 00:00:00 GMT

Time for a confession with a row count. Somewhere in my Google Drive lives a spreadsheet I have been maintaining for more than ten years. It has over two hundred and fifty rows, one per tool, and four columns that never change: what it does, which category it belongs to, the rating I gave it, and the link to go there. It is one of the most valuable documents I own, I consult it weekly, and it is, objectively, a horror. No previews, no search worth the name, no way to show a client without apologizing for it first. A friend of mine does the same thing in a PDF he regenerates by hand. Everyone I know in this trade does it somewhere, in browser bookmarks collapsing under their own weight, a Notion page abandoned at forty entries, or a text file called tools-NEW-v3.

The need is so real that an entire artisanal culture grew around its public half. Developers hand-build “/uses” pages, personal pages listing the tools they swear by, and uses.tech collects hundreds of them, every single one coded by hand. Think about what that means. The demand is so strong that people write HTML for it. The directory is a list, not a product. Nobody has turned the private spreadsheet into that public page automatically. My spreadsheet is the private half of the same instinct, bricolage on the other side of the same missing product.

Stackpage is that product. A private library for the tools of your trade, structured the way my spreadsheet always wanted to be. Capture a URL and the entry builds itself (name, logo, description fetched), then you add the category, your rating, and the one line that actually matters, why I use it. Search that works. And then, the second half, one click turns your library, or the slice of it you choose, into a public page. Clean, fast, categorized, yours, stackpage.com/you or your own domain. The private workbench generates the public storefront.

Library first, page second: the lesson of a dead startup

The graveyard around this idea is instructive, and one grave more than the others. Kit.co built exactly the public half, beautiful recommendation pages for creators’ gear, and shut down in May 2026. I take one lesson from it. A page that exists only to recommend is a chore you update twice and abandon, because it serves your audience, not you. The order has to be reversed. Stackpage is a tool you keep open for yourself, the place you check on Monday for that invoicing app you starred in March, and the public page is a free byproduct of a library you were maintaining anyway.

Self-interest keeps the data fresh; the audience gets the overflow.

The rest of the graveyard confirms the gap from the other side. Raindrop and its public collections are the closest living neighbor, and they share a folder, not an identity. No ratings, no voice, no “here is my stack,” just links in a box. Start.me and Toby organize private dashboards and never grow a public face worth linking. StackShare does public stacks, but for companies and engineering org charts, cold as an RFP. And the /uses pages themselves are lovely and frozen, hand-coded, rarely updated, unsearchable as a set. Between the folder, the dashboard, the org chart, and the artisanal HTML, there is a person-shaped hole: the professional whose taste in tools is part of their professional identity.

Key Features

Capture in five seconds. Paste a URL, get a filled entry: name, favicon, description, and a preview image if the site offers one. You add category, rating, and the one-liner. Anything slower than my spreadsheet loses to my spreadsheet.
Import that eats your history. Browser bookmarks, a CSV, a Google Sheet with ten years of seniority: day one must ingest the bricolage this product replaces. I have two hundred and fifty rows that demand it.
The why-line. The single highest-value field in the product. “Fast invoicing, French tax rules handled” tells a visitor more than any star rating. Ratings sort; reasons convince.
Private by default, public by selection. The library is a workbench, not a performance. You choose which categories and which entries surface on the page, and the page auto-updates as the library lives.
A page worth being your link. Typographically clean, categorized, searchable, with your name and one paragraph of you. The kind of page you put in a proposal footer, a conference bio, an email signature.
Rot detection. Tools die constantly. Stackpage pings your links, flags the dead and the acquired, and nudges you to prune, which solves the moment every hand-made /uses page silently gave up.
The directory. Every public stackpage can opt into the shared index: browse stacks by trade, find what other freelance designers or SEO consultants actually pay for. The hand-built culture proved people want this; the directory is where the network effect lives.

The honest tensions

The Notion question first, because it is the killer objection. Could this not be a template? For the private half, yes, roughly, minus the capture speed, the previews, and the rot detection. What a template cannot produce is the public format, a standard, recognizable, beautiful stack page at a stable URL, or the directory that makes those pages discoverable as a set. The hand-coded /uses culture is the proof. Those people all have Notion, and they wrote HTML anyway, because they wanted a page, not a document. The product is the format plus the network; the library is how you earn the right to generate it.

The adjacency is crowded and the differentiation is thin on any single axis. It only holds as the specific bundle, workbench plus identity plus directory, aimed at the specific person, the web professional whose toolbox is part of their reputation. That is a niche, not a landgrab, and the economics below are sized accordingly.

And affiliation, the obvious revenue, is also the obvious corruption risk. The moment ratings feel bought, every page on the platform is worthless. The line has to be structural. Affiliate links are auto-applied by the platform where programs exist, clearly labeled, revenue shared with the page owner, and never, ever, connected to ranking, rating, or visibility. You recommend what you recommend; if money follows, it follows disclosed and downstream.

Monetization

Free tier generous enough to live in: full private library, public page on the shared domain. Pro for the professionals the product serves: custom domain, branding, page analytics, multiple stacks (one per service line, one per client type). And the affiliation layer as described, platform-brokered, labeled, revenue-shared, which has the elegant property of paying the people who maintain honest, current, useful pages, since dead links earn nothing.

Next Steps: How I’d Build It

The MVP has a name and it is sitting in my Drive. The day Stackpage can import my ten-year spreadsheet and render it as a page I would genuinely send to a client, the concept is validated by its harshest available critic. Capture, import, categories, ratings, the why-line, one clean public template. No directory yet, no affiliation yet.

Then fifty hand-picked web professionals, the kind who already keep a sheet or a dead /uses page, invited personally, their imports assisted if needed, because fifty real stacks make the directory worth opening and the format worth copying. The affiliation layer comes last, once there are pages worth monetizing and a reputation worth protecting.

Closing

Stackpage is the spreadsheet every web professional secretly maintains, given capture, previews, a search, a public face, and a neighborhood. Mine has two hundred and fifty rows and a decade of seniority, and it deserves better than cell A1.

If you build this, tell me, and I will be your first import: ten years of tools, rated, categorized, reasons included. Consider that both an offer and a warning about my standards for the CSV parser.

GPT Stat Central: How you actually talk to a machine

Mon, 06 Nov 2023 00:00:00 GMT

I thank GPT after almost every answer. I know, I know. Does GPT even care? Probably not.

But it makes me wonder about everyone else. Are you a polite AI conversationalist, or do you keep things transactional? Some people, like me, treat it as a digital companion. Others take the direct approach, a tool that does not need the niceties. I have no idea which one I would rather be, and I would very much like to see the number.

GPT Stat Central is the Wrapped I want for that habit. You hand over your official chat export and get a snapshot: how often you thank the model, how long you stay, whether you write novels or one-liners, whether you come back to the same question as if you were checking reality. Casual or power user. Creative prompts or strictly business. Love and family, or geopolitics and code.

It is a playful dashboard, not a personality lab, and not an institute. I want to know my thank-you rate. I would like to laugh at it with a friend. That is enough of a product.

I built it, later, under a cleaner name. It is live at anonwrap.com. It turns your ChatGPT or Claude history into a personal Wrapped of ten animated cards, computed entirely in the browser. Nothing uploaded.

I am not trying to diagnose anyone. I am trying to make a private joke measurable. If the page comes back sounding like a report card, rewrite the copy before you add another chart.

Market & Opportunities

People already screenshot a funny chat. They do not already have a page that counts the year. The gap is the same shape as Spotify Wrapped: a file you already own, turned into a story you might share.

Who shows up first? People who have been talking to a model long enough to be a little embarrassed about it. Who pays? Anyone who wants last year’s picture, and later a team that wants a boring aggregate (“support drafts a lot, sales barely opened it”) and nothing else.

I will not hang a global AI-market figure on this, and I will not call it a Myers-Briggs for prompts. The relevant market is “I export my data once a year and I want a poster.” Duolingo and Strava proved people will wear a streak. They did not prove the world needs a new social network. I would steal the shareable card and leave the feed.

Key Features

A dashboard that starts with the thank-yous. Session counts, average length, please and thank you, creative versus technical. A snapshot, not a cockpit.
You against the pile. Optional comparison to anonymized totals from other exports. Are you alone in writing novel-length prompts? Fine. Are you in the middle? Also fine.
Badges you can turn off. “Top Thanker,” “Long-haul,” a score if you want one. A public card if you want one. Private by default.
A category split. Productivity, writing, code, research, casual. A trend line across imports, so you can see a month of code and a month of something else.
A monthly or yearly Wrapped. Peak hours, longest thread, whether you are a morning user. One image you might send to a friend. That image is the product.

I would not have the model write you a note about “balancing your usage.” That is a wellness product wearing a chart. Count the words. Leave the advice.

How the file actually gets in

There is no API that streams your ChatGPT usage to a third party. The data has to come from you. The workable path is the official ChatGPT data export. You download the archive and hand it to the app, which parses it.

This is batch analysis of a file a user gives you, not a live tap on OpenAI.

Less dramatic, actually buildable, and far easier to promise privacy on. The work is then ordinary and still hard: keep up when the export format moves, run as much as you can in the browser so raw threads never have to leave the machine, survive a dozen chats and survive five years of daily use, and accept other providers later, because GPT is not the only player in town.

If parsing happens on the user’s side, the service only needs the derived stats: counts, categories, trends. Never the conversations. That is still enough to tell you how many times you said “please” this year.

Monetization

A free pass that gives you the thank-you count and one Wrapped. A premium around ten dollars a month (or a cheaper yearly) for history, extra breakdowns, and a nicer card. I would not invent a Pro tier just to print an API.

There is a team version of this, but it has to stay on the right side of a bright line. Aggregated and anonymized, never individual scorecards. A company could see that its support team leans heavily on AI for drafting, or that adoption is uneven across departments. Useful, boring, aggregate stuff.

What it should never get is a per-employee “AI efficiency score” feeding into performance reviews. That is surveillance dressed up as analytics, and the fastest way to make people resent both the tool and the AI. If the team product cannot be built without individual rankings, it should not be built.

I would not open a marketplace of third-party “insight modules.” That is how a Wrapped becomes a junk drawer of other people’s psychology.

Bonus: what this file contains

People put real things into these chats. The app should keep derived stats, let the user delete them, and say plainly that it will not sell the pile or hand it to an employer’s HR dashboard. Categories that look like emotional support or private writing should be opt-in, or not computed at all.

I would not add a feature that watches for “unhealthy” usage. That is a health claim I cannot stand behind. If someone imports every day, they get a chart. They do not get a lecture.

Next Steps: How I’d Build It

The first version is already at anonwrap.com. What follows is still the honest path, for a replica or a cousin in another domain.

A weekend: accept one official export, count thank-yous, session length, and a crude category split, render a single share card. No badges, no team, no comparison. If ten people will import a real archive to see that card, the idea has a pulse.

Do the parse in the browser first, even if it is slow. The moment raw threads hit your server “just for a minute,” you have become a different company. Say that on the upload page. People who will not upload after reading it were never going to be happy customers.

Then local parsing, then the comparison pile, then the yearly image. The wall is not the chart library but getting anyone to upload a year of chats they were not already going to screenshot. Start with people who already joke about thanking the model. I would be in that line.

Closing

GPT Stat Central, shipped as AnonWrap, is a Wrapped for the way you talk to a machine, built from a file you chose to hand over, with a hard no on turning it into a score for your boss.

Go get yours at anonwrap.com. If you build the writers’ version, or the team version that stays on the right side of that line, tell me. You owe me nothing. I still want to know your thank-you rate.

Overhead: The Social Layer Floating Above Our Heads

Wed, 25 Oct 2023 00:00:00 GMT

For as long as I can remember, I have carried the same picture of the future: you look at the people around you, and floating above their heads there are little icons. Status metadata, like in a video game. Single or taken. Open to talking or leave me alone. A tiny flag for the thing they are proud of this week. In my head you could always toggle it, show what you want, hide the rest, switch the whole layer off when you are not in the mood. I did not get this from one specific film. It just always seemed obvious that this is where things were going.

Then I started thinking seriously about how you would actually build it, and I hit the wall everyone hits. To see icons floating over someone’s head, something has to draw them between the world and your eyes. Smart glasses, in other words. And I already wear glasses, the corrective kind. I am not stacking a second pair on top just for this, and neither is anyone else. It is too expensive, too conspicuous, too soon. That is why this future keeps not arriving, even though the AR demos have existed for a decade.

So the honest version of the idea is to drop the floating icons, for now, and keep what they were actually for. The icons were never the point. The point was the information: a voluntary, visible-at-a-glance social status for the physical space around you. And that part does not need glasses at all.

Overhead is a small opt-in radio presence, in your phone or on a cheap clip-on tag, that broadcasts a status you chose to the people physically near you who also opted in. Not your identity, not your location history, just the little flag you decided to fly today. You check the layer on your phone, or your earbuds whisper it, or a wearable blinks softly when someone nearby matches something you care about. It is the video-game overlay, minus the part that requires strapping a screen to your face.

Market & Opportunities

The desire this taps is old and completely mainstream. You want to know something about the strangers around you before you commit to the terrifying act of talking to them. Dating is the obvious engine, the wish to know who here is single and actually open to being approached. But the same mechanic covers professional networking (conference badges are exactly this, made of paper), travelers looking for other travelers, gamers finding a fourth player in a café, or someone simply signaling “I speak French, ask me for directions.”

The technology finally matches the ambition. Every modern phone speaks Bluetooth Low Energy, the same short-range presence tech behind AirTag-style trackers, and newer phones add ultra-wideband, which knows direction and distance to within centimeters. Broadcasting a small anonymous token to a radius of a few dozen meters, and resolving it into “someone nearby is open to networking” is well inside what today’s hardware does on a coin-cell battery.

And there is a graveyard to learn from, which I consider an asset. Proximity social has been tried repeatedly (Highlight, Sonar, Color, Bump, Google Latitude, Nintendo’s StreetPass), and it keeps dying in the same three ways. It runs out of density (the app is worthless until the room has other users). It creeps people out (ambient location sharing without a clear reason). Or the hardware moment passes (StreetPass was beloved, and died with the console that carried it). This version of Overhead is designed against those three failure modes from day one, and the Next Steps below take them in order.

Key Features

Statuses, not profiles. You broadcast at most a handful of small flags: relationship status, open-to-talk, a topic tag. No photo, no name, no bio. The mystery is the feature; Overhead starts conversations, it does not replace them.
Opt-in at every layer. Broadcasting is off by default, per-flag, per-context. A “visible at this event only” mode makes conferences and festivals the natural on-ramp. Silence is always one tap away.
Proximity without coordinates. The protocol answers “someone near you” and, on UWB phones, “roughly that way.” It never exposes a map pin, a distance readout precise enough to corner someone, or any trace of where you were an hour ago.
Rotating anonymous tokens. Like modern contact-tracing protocols, your broadcast identifier changes constantly, so nobody can log your token today and recognize you tomorrow. Following a person via Overhead should be technically hard, not just forbidden.
The phone is the display. A glanceable card list, an optional earbud whisper, an optional gentle blink on the tag when a match walks in. No glasses anywhere in the product.
Blocklists that mean it. Block someone and your radio goes mutually invisible to theirs, permanently, at the protocol level.

Monetization

Overhead should be free and unlimited for the individual, because density is life or death and a paywall at the person level kills the network before it exists. The money is at the venue and event layer. Conferences, festivals, coworking spaces, and bars pay for an “Overhead zone,” custom event flags (“speaker,” “hiring,” “first-timer”), and aggregate, anonymous stats (how many attendees flew a networking flag tonight). Event badges are a boring, sizeable market, and this is a live badge.

The second line is the optional hardware tag itself, sold at a modest margin for people who want presence without draining their phone, kids-off-phones use cases, or venues handing them out at the door. A third, later possibility is premium personal filters (alert me only for this rare combination of flags), kept strictly on the receiving side: paying may get you better ears, never a louder voice.

Next Steps: How I’d Build It

Density first, because it kills every predecessor. Overhead should not launch as a general-purpose app into an empty world. It should launch as an event product, one conference at a time, where a thousand people in one building opt in together and the layer is instantly alive. The general city-wide version comes after the hundredth event, if it ever does.

Safety second, and it deserves plain words. A system that lets strangers see that a woman nearby is single is a stalking tool unless it is engineered not to be. That means the defaults above are non-negotiable (off by default, rotating tokens, no precise distance, protocol-level blocking), and it means red-teaming the design with the people most likely to be targeted, before launch and not after the first incident. If those protections make the product less viral, that is the correct trade and the article you are reading commits to it.

Then come the technical limits. iOS restricts background Bluetooth broadcasting, so the polished phone-only version fights the platform, which is why the cheap dedicated tag exists in the plan. And the deepest risk is not technical at all. The Google Glass lesson says the surrounding culture decides. A room where everyone knowingly opted in, at an event built for it, is the one place that lesson works in your favor. Start there.

Bonus: the dystopia I originally wanted

I owe you the ending of the original daydream. The floating icons are coming eventually, and the only question is the decade. Meta is already selling display glasses, contact-lens prototypes exist, and somewhere down the cyberpunk slope are ocular implants. The day one of those wins, a protocol like Overhead is the natural data layer for it, and my childhood picture finally renders, literally, above people’s heads.

On that day the safeguards stop being a feature and start being the whole product. In maximum-dystopia mode, the same layer shows everyone your relationship status whether you flew the flag or not, scraped from elsewhere and pinned over your head by someone else’s app. The line between the dream and the nightmare is a single word: consent, at the protocol level, from day one. If the future is going to draw things over our heads, I would rather the standard it inherits be one where every icon was put there by the person under it. That is worth building early, before the glasses decide for us.

Closing

Overhead is the social layer I always assumed the future would have, built with the hardware we hold today, and pointed at the version of that future I would actually want to live in.

If you build this, tell me. I will be at whatever event you demo it at, flying a flag that says so. And if you want help thinking through the safety model, ask. That is the part I care about most.

ENCODE: Verified Calls for the Age of Cloned Voices

Wed, 24 May 2023 00:00:00 GMT

A while ago I tried one of those online voice-cloning demos, the kind that asks for a short sample and hands you back a synthetic twin. I fed it a few seconds of my own voice, typed a sentence I had never said, and pressed play. What came back was not perfect. It did not need to be. It was me enough to fool anyone who loves me, and it took less time than making a coffee.

The research says my little experiment was not a fluke. Microsoft’s VALL-E paper showed personalized speech synthesis from a three-second sample, and McAfee researchers produced an 85 percent voice match from three seconds of audio using freely available tools, 95 percent with a bit more training data. Meanwhile the classic scam this enables, a cloned voice of a relative pleading for emergency money, is documented by both the FTC and the FBI. The raw material for it is everywhere, in every voicemail, every story, every video posted anywhere.

My first instinct was the same as everyone’s. We need better deepfake detectors. And that instinct is a trap. Detection is an arms race against generators that improve every month, and the detector has to win every single time, in real time, on a phone call. So flip the problem. Stop trying to detect the fake. Prove the real.

ENCODE packages that flip. You get a personal, portable call-identity you share like a handle, a trust list of people whose devices have mutually verified each other, and a badge on every call that tells you, at a glance, which world the caller is calling from. It does not analyze anyone’s voice. It answers a smaller, harder question with certainty: is this call coming from the device I already trust?

Market & Opportunities

The money at stake is not hypothetical. Americans reported losing 3.5 billion dollars to imposter scams in 2025 alone, up from 2.95 billion in 2024, according to FTC Consumer Sentinel data, and reported losses understate reality since most fraud never gets reported. On the corporate side, business email compromise, the “fake CEO” family of fraud, cost victims roughly 3 billion dollars in reported losses in 2025 per the FBI’s IC3 report, with more than 20 billion accumulated since 2015. Voice cloning is pouring fuel on both fires.

The detail that convinced me this idea deserves to exist is that the FBI already tells people to build ENCODE by hand. Its December 2024 public service announcement on AI-enabled fraud advises, verbatim, “Create a secret word or phrase with your family to verify their identity.” The federal answer to billion-dollar AI fraud is, literally, a family password. It is good advice, and it is also fragile. People forget the word, say it on a compromised line, or panic and skip it just when it matters. When the official mitigation is a manual ritual, there is room for a product that makes the ritual automatic.

The two markets stack naturally. Families are the story: the grandparent who gets the 2am call, the parent who wired money because the voice was crying. Companies are the engine. An ENCODE directory bound to the corporate identity provider, so that a video call from “the CFO” carries a verified badge or it does not, is a line item any security budget understands.

What already exists, and why it is not enough

I have to name the graveyard first, because pieces of this idea are everywhere. Signal’s safety numbers already let two people verify their channel cryptographically. Signal itself notes this verifies the keys, not the human behind the account, and the feature is optional and buried. Even the automatic key verification Signal added in August 2026 has to be started by hand, contact by contact. STIR/SHAKEN, mandated by the FCC, lets carriers attest that a call really comes from the number displayed. It authenticates the number, not the voice, and the attestation is stripped whenever a call crosses legacy non-IP networks. And the family code word, as we saw, has a federal endorsement and a human failure rate.

So the invention here is not cryptographic novelty. Every brick exists. The product is the assembly and the defaults: verification that happens without anyone comparing sixty-digit numbers, a badge that is simply there on every call, and a system that fails loudly instead of silently. Signal built the engine and hid it in a settings menu. ENCODE’s bet is that the entire product is the part Signal considered optional.

Key Features

A portable call-identity. Your ENCODE ID is a short handle you give out like an email address, on a business card, a LinkedIn profile, an email signature. Anyone can request to add you; nothing happens until you approve.
A mutually verified trust list. Approval runs device-to-device: both phones exchange keys, and from then on each can prove it is talking to the other. No central server holds the relationship graph.
Four levels, not two. Verified contacts ring green. Honest strangers, the delivery driver, the hospital, ring yellow: unverified, flagged, but never blocked. Calls with spoofing markers get challenged. Only manifest fraud gets cut. A binary system would break real life; a graded one degrades gracefully.
A badge that stays on. Verification is not a checkmark at pickup; it is a persistent indicator during the call. If the cryptographic session breaks mid-conversation, the badge goes red while you are still talking. That is the moment that matters.
Device changes as a security event. New phone means re-verification: a QR handshake from the old device when possible, and your contacts see “new device” on your next call until they re-confirm. The moment attackers love most, “hi, new phone, new number,” becomes the moment the system is loudest.
An honest label on the tin. The app states plainly what it proves and what it does not, which brings us to the next section.

What ENCODE does not do

ENCODE authenticates the device, not the person. It is a badge for your front door. It proves the badge is present, not that the right hand is holding it. If someone steals your father’s unlocked phone, ENCODE will vouch for them. If someone coerces a person into calling you themselves, every light stays green, because the call is genuine even though the situation is not.

It also does not detect deepfakes, and it should never pretend to. Bolting a voice-analysis layer onto ENCODE would dilute the one promise it can keep into two promises it cannot. The right architecture is separation. ENCODE proves the channel, and if the detection arms race someday produces something reliable, it can run alongside. What ENCODE kills specifically is the scam that works today at scale: a stranger with thirty seconds of cloned audio and a spoofed number, calling cold. That attacker never held the trusted device, so they never get the green badge, no matter how perfect the voice is.

Monetization

Free for individuals and families, permanently, because a trust network with a paywall is a trust network nobody joins, and every additional verified family is marketing no budget could buy. The business is B2B: per-seat licensing for organizations, with the ENCODE directory synced to the identity provider (Okta, Microsoft Entra), admin policies like “wire transfers require a green-badge call,” and audit logs. Security teams already buy anti-BEC tooling measured in billions of losses; a verified-call layer for executives prices itself. A later tier can serve banks and brokerages that want to place verified outbound calls to clients, an inversion of today’s absurd ritual where the bank calls you and then asks you to prove who you are.

Next Steps: How I’d Build It

Phase one is app-to-app calling only, and that is a feature, not a compromise. Inside its own VoIP calls, ENCODE controls the whole experience end to end and depends on nobody’s favors. It needs no carrier cooperation and no deep hooks into the phone’s native dialer. Signal bootstrapped the same way. On iOS, CallKit already lets VoIP apps present calls in the native interface, which is enough for the green-badge experience within the app’s own calls.

There are two blockers. First, the platform ceiling. Deep integration with ordinary cellular calls is limited by what Apple and Google expose, so full protection of legacy phone calls is a vision that requires either OS-level adoption or carrier partnerships, and the article you are reading does not promise those. The strategy is to own the app-to-app layer so thoroughly that platform integration becomes the obvious next move for someone bigger. Second, the cold-start problem. A trust list is worthless alone. The wedge is the family unit, three to six people who install it in one sitting because one of them just read about a voice-clone scam, and the enterprise pilot, one finance team, one policy, one quarter of audit logs.

Build order: the identity and key-exchange core, the two-person verified call, the four-level badge UX, then the family circle, then the Entra/Okta sync. A working two-person prototype is a weekend of WebRTC plus a signal protocol library. The product is everything around it.

Closing

ENCODE refuses to play the detection game at all. Prove the real, label the rest, fail loudly. The FBI already told your family to do this with a whispered word. This just makes the word unforgettable, unphishable, and automatic. It belongs to the same fight as ThreatShield, my concept for a safer inbox. Both are tools for the people we love who did not grow up with this internet.

If you build this, tell me. I will be your first verified contact. And if you want to argue about the threat model before you write a line of code, even better; that conversation is the product.

The Obscurantism Clock: Measuring the Drift Away From Science

Mon, 27 Feb 2023 00:00:00 GMT

This one did not arrive as a product idea. It arrived as a worry. Late one night I found myself dictating a long, rambling message to an AI about something that had been gnawing at me. It was the feeling that scientific thinking itself, the basic shared agreement that evidence beats assertion, was losing ground. Not in one place, not on one topic, but as a slow tide across medicine, education, and public life. I asked the machine a simple question: can this be made visible?

There is a famous precedent for making dread visible. In 1947 the Bulletin of the Atomic Scientists compressed the entire nuclear age into a single image: a clock face, a few minutes before midnight. No dashboard, no report, just a hand that moves. In January 2026 they set it at 85 seconds to midnight, the closest it has ever stood. Seventy-nine years on, that one image still does more work than a thousand white papers.

So here is the idea, whole. The Obscurantism Clock is the same device, pointed at a different danger. A public clock that answers one question, how far are we drifting from scientific thinking, using a transparent formula, weighted indicators, and data anyone can trace back to its source. Midnight is the point where evidence stops mattering. The hand moves when the indicators move.

And whether or not the popular story of eight dark medieval centuries is entirely fair to the scholars who lived through them, the fear that story names is real. Knowledge can go backwards. It has before. A clock is a way of watching for it.

What the verified data actually says

This concept only works if its numbers are unimpeachable, so before writing this article I had every headline statistic in my original draft fact-checked against primary sources. Here is what survives, and it is sobering enough without any embellishment.

Trust took real damage. Pew Research Center’s long-running series shows the share of Americans with a great deal of confidence in scientists falling from 39% in April 2020 to 23% by late 2023. A JAMA Network Open study of more than 440,000 US adults (Perlis et al., 2024) found trust in physicians and hospitals dropped from 71.5% in April 2020 to 40.1% by January 2024. The consequences are not abstract. The CDC recorded roughly 2,288 confirmed measles cases in 2025, the most in any year since 1991 and the largest outbreak since the disease was declared eliminated in 2000, while kindergarten vaccination exemptions hit a record 3.6%. Pew finds support for school MMR requirements fell from 82% in 2019 to 69% in 2025.

The divide is now tribal. Pew measures a 25-point partisan gap in confidence in scientists, and the share of Republicans with little or no confidence nearly tripled between April 2020 and late 2023, from 14% to 38%. A 2024 University of Michigan study in Scientific Reports, analyzing millions of tweets, estimated that 14.8% of Americans deny climate change, with denial topping 20% in four states, and identified Donald Trump as the strongest single influencer in that denial network, ahead of conservative media outlets. On the education front, PEN America has documented nearly 23,000 school book bans across 45 states since 2021, alongside 51 enacted state laws or policies restricting classroom instruction. And the mechanics favor the wrong side. The landmark Vosoughi study in Science found false news on Twitter reached its first 1,500 people about six times faster than true news.

But, and a clock worth trusting must show this too, the freefall is not the whole story. By late 2025 Pew measured confidence in scientists ticking back up, with 77% of Americans expressing at least a fair amount. The sharp collapse happened between 2020 and 2023. The story since is less a plunge than a hardened, polarized plateau. Any instrument that can only move toward midnight is a marketing campaign, not a measurement. This clock’s hand must be able to move backward, and as of today, it slightly would.

Key Features

A formula you can audit. A weighted sum of public indicators maps to a score from 0 to 100, and the score maps to minutes before midnight. The whole computation is a page of open-source JavaScript, published on the site. No black box, no vibes.
Every number wears its source. Each indicator links to the primary institution that produced it (Pew, CDC, PEN America, peer-reviewed journals), with the survey date and a permanent archive copy. If a number cannot be traced, it does not enter the formula.
An events ledger. Major documented events, an outbreak, a landmark law, a major institutional restructuring, are logged with dates, sources, and their computed effect on the score, so anyone can replay history and see why the hand moved.
A hand that moves both ways. When the data improves, the clock retreats from midnight, publicly and with the same fanfare. This is the feature that separates an instrument from a doom machine.
Honest measurement labels. Survey data, social-media-inferred estimates, and case counts are different kinds of knowledge. The clock displays them in separate tiers rather than blending a 2019 Twitter analysis into a 2025 poll as if they were the same thing.
Open data and an embeddable widget. The full dataset ships as JSON under an open license, and any site can embed the clock, the same way the Doomsday Clock spreads through citation.

How it stays honest

The uncomfortable truth about any index like this is that the weights are a judgment call. Why should vaccine hesitancy count for more than book bans? There is no equation for that. Someone decides. The Doomsday Clock has the same property, and the Bulletin handles it the only defensible way, by owning it. A named, diverse board deliberates in public, publishes its reasoning, and accepts the argument. The Obscurantism Clock should copy that honesty rather than hide behind decimals.

So the false precision of a readout like 23:47 is a rhetorical device, and the site should say so, plainly, on its methodology page. The clock is a communication instrument built on real data, not a physical constant. Its credibility budget is spent in one place, on never publishing a number that cannot survive a hostile fact-check. A small editorial committee spanning disciplines and, above all, political sensibilities reviews every indicator change, and dissents get published too. The project critiques systems and incentives, never private individuals. Where a public figure appears, it is because a peer-reviewed study put them there, cited verbatim.

Monetization

This is a civic project, not a business, and pretending otherwise would poison it. The MVP costs about as much as a domain name, roughly 15 dollars, because a static site on free hosting needs nothing else. Sustaining it looks like donations, foundation grants in the science-literacy and press-freedom space, and university partnerships that lend both rigor and credibility. No ads, no paywall, no dark patterns. An instrument about trust cannot fund itself by eroding it. If it ever earns anything, it is through sponsored access to a clean historical API for newsrooms and researchers, priced to cover servers, not to profit.

Next Steps: How I’d Build It

The build is the easy part, and I mean that literally: a static site, one JSON file of indicators, one JSON file of events, an animated SVG clock face, and the scoring function, deployable on GitHub Pages in two or three weeks by one person. The Plotly charts and the widget come right after. None of this is the hard part.

The hard part is governance, and it starts before the first commit. Recruit the small editorial committee first, agree on the indicator list and the weights in writing, and publish that methodology page before publishing any score. A clock that appears fully formed from one anonymous author is exactly as trustworthy as it sounds. The second hard part is the fact-check-first rule as an operating principle. Every indicator update is verified against the primary source, archived, and only then merged. Slower, and worth it.

And the third is the obvious one. The accusation of political bias will arrive within hours of launch. The only defense that works is radical transparency, indicators from recognized scientific institutions only, methodology public, code open, committee politically diverse, and the hand demonstrably willing to move away from midnight when the data says so. You will not convince everyone. The narrower, more achievable goal is that any good-faith reader who checks the sources finds exactly what the clock claims, every single time.

Bonus: the first draft failed its own test

I owe you a confession, because it is the best argument for building this. The first draft of this concept, produced in that late-night AI conversation, came loaded with statistics. Some checked out. Several of the most alarming ones did not. A dramatic collapse of strong confidence in science from 32% to 8%? No such numbers exist in any primary source; the real Pew series shows decline, then a modest recovery. A rise in parental vaccine distrust from 6% to 27%? Unsupported; the real Pew figures tell a related but different story. Meanwhile the measles count in the draft was actually lower than the grim reality.

Think about what happened there: an AI, asked to document the erosion of truth, invented evidence for it. Fluent, plausible, sourced-looking evidence. If I had shipped that draft, the first hostile fact-check would have deservedly killed the project, and the debunking would have become one more data point for the very cynicism the clock is meant to measure. The clock’s first act was to fact-check itself, and it barely passed. That experience is now its founding rule, and it is a rule this whole era could use.

Closing

The Obscurantism Clock is one image for a diffuse dread, built on data that survives being checked, honest about its judgments, and capable of delivering good news. The concept, the methodology, and the roadmap are all yours under the same open license as everything here.

If you build this, tell me. I want to be there the first time the hand moves, in either direction. And if you want help pressure-testing the methodology before launch, that is the part I care about most, so ask.

GPT Roulette: Share and discover AI conversations

Sun, 15 Jan 2023 00:00:00 GMT

When the first AI chatbots arrived, my friends and I turned them into a competition: who could pull the best, strangest, most useful answer out of the machine. We never had a way to actually compare or verify anything, the conversations lived and died in our own accounts, and the contest stayed a running joke between us. GPT Roulette is that game, built for real. It gathers the strange, funny, and useful exchanges people have with AI in one public feed you can scroll, vote on, and argue about.

You pick a conversation worth showing, hit the big red “Roulette” button, and it lands in the public feed. In return you gain access to what everyone else has contributed. You browse the stream, upvote the best ones, and see posts grouped by topic.

How does sharing work? There is no magic pipe into your account, because OpenAI does not offer a way for a third-party app to pull your ChatGPT history. So sharing works the way it realistically can. Either you paste a conversation (or use ChatGPT’s own “Share link” and drop that in), or you upload your official ChatGPT data export and pick the threads you want public. A little more friction, but no fantasy API, and no app quietly holding the keys to your whole account.

The fun is the product. A public pile of real conversations is also, if people opt in, a way to see how humans actually talk to these models: the prompts that work, the ones that fail, the jokes, the dead ends. I care about that pile. I do not need a speech about democratizing AI to justify a good feed.

What I wanted in that old contest, and never had, was a third person in the room who had not written either prompt.

A friend will laugh to be polite. A stranger with a vote will not.

That is why the site is public, and why anonymous posting still needs an account behind it. Without that tension you get a museum of people clapping for themselves.

Market & Opportunities

People are already having the conversations. They are not already sharing them, except as screenshots in group chats. The gap is a place with structure, votes, and a rule that the author chose this thread, not a scraper.

The first to show up are the same people who already trade prompts, the ones who would have joined our old contest. Researchers and teachers are a later audience, and only if the pile is clean. I will not hang a global AI-market figure on this. The relevant market is “people who want to show a chat to a stranger and not look like they leaked their job.”

There is already a graveyard of prompt galleries. Most of them are one-liners with a like button. A conversation is longer, messier, and more honest. You see the dead end, the retry, the moment the model got it. That length is the product and the moderation problem at the same time. If you compress every thread into a prompt snippet, you have rebuilt the graveyard.

Threads about who jailbroke the model this week can live somewhere else. This feed is for the useful, the strange, and the ones that made a friend say “wait, show me that again.”

Key Features

Import by hand. Paste, official share link, or a picked thread from your export. Nothing else ever leaves the account.
Review, trim, confirm. You see the thread, you cut what should not be there, you press Roulette once. No “share everything” button.
A public feed with votes. Newest, best, or by topic. Comments allowed. Downvotes too. A feed that cannot sink a bad post will fill with them.
A name, or none. Post as yourself or anonymously. Anonymous still has an account behind it, or the spam wins.
Remix. Fork a thread and keep going. The original stays. This was the feature hiding in my old notes, and it is the one that makes the site a workshop instead of a museum.
Aggregates, never the raw pile for sale. Public trend boards if you want them. Researcher exports only as anonymized, opted-in aggregates. Raw conversations are not a product.

Monetization

The feed stays free. A small subscription can unlock saved collections, extra remix slots, and the thicker trend board. Sponsored posts are a trap in a vote-ranked feed; I would not start there. Selling the dataset is allowed only in the aggregate form above, and I would rather the site be slightly poorer than slightly for sale.

A word on what is inside a conversation

This is the part that has to be handled with care, not as an afterthought. People put real things into AI chats: names, employers, medical worries, other people’s secrets, code from work. Making any of that public is not a moderation edge case, it is the core risk. At minimum the app should scan for and strip obvious personal data before a post goes live, default to sharing single threads rather than “your entire history”, and never publish anything without an explicit, per-conversation confirmation. “Share your whole chat history in one tap” is a feature I would cut on purpose.

The rest of trust-and-safety is ordinary and still required: a rules page, a report button, a human queue, age gates for the ugly threads, and a way to delete what you posted. Ordinary does not mean optional.

Next Steps: How I’d Build It

A weekend: paste-in, trim, Roulette, a chronological feed, votes. No export parser yet, no remix, no analytics. If ten people from a single community will post a real thread and come back to read someone else’s, the idea has a pulse. Then the export picker, then remix, then the aggregates.

The wall is not the stack but getting anyone to publish a chat they were not already going to screenshot. Start inside one scene (prompt-traders, a class, a Discord) rather than “the AI internet.”

I would also write the import screen as if a lawyer were standing behind the user. Label the paste box. Label the official export. Say, in one sentence, that OpenAI does not let a third-party app pull history, and that this is why the import works by hand. If that sentence scares people off, they were never going to post a real thread anyway. The ones who stay will treat the trim step as part of the craft, which is what you want.

Closing

GPT Roulette is the contest my friends and I never had a board for, plus a public pile that only exists if each author meant to add to it. The red button is a joke with a rule behind it. You only spin what you were willing to show.

If the feed fills with safe, empty threads, kill the leaderboard and start over in a smaller room. If it fills with real ones, the research layer can wait. Fun first, aggregates later, never the other way around.

If you build this, tell me. I would be glad to review it, and to update this article with a link once it is live. You owe me nothing, just tell me how it goes.

Keep: The App That Keeps Your Things Alive

Mon, 17 Oct 2022 00:00:00 GMT

Take an inventory of my apartment with me. There is a vinyl turntable whose stylus should be brushed and whose belt has a lifespan I have never once checked. A guitar that would like its strings changed more often than they are, and its fretboard oiled at intervals I could not name under oath. A keyboard, the humble computer kind, that deserves better than the crumbs it lives with. And a drawer of t-shirts that have one simple wish, to be turned inside out and washed cool, a wish I remember roughly every second load.

None of this is exotic knowledge. It is a hundred small, slightly dumb things (brush this monthly, oil that twice a year, never wash this hot), and I know maybe a third of them and act on fewer. Meanwhile my car, the one object in my life with a maintenance book and a garage that calls me, gets flawless care. The difference is not love. The difference is that somebody built the system of reminders for cars and nobody built it for everything else.

Keep is that system. You photograph an object, and Keep identifies it and hands back its care sheet: what this thing needs, how often, and how to do it in plain words. Then it does the part no manual ever did. It remembers for you. The right nudge at the right rhythm (strings in March, belt check in autumn, descale the coffee machine now), and a log of everything you have done, a service book for your whole material life.

Where this idea came from, and what freed it

The first version of this idea, in my 2024 notes, was a Shopify plugin that offered AI-generated care guides to customers after checkout, a rare use of AI that pushed longevity instead of another purchase. I still like that instinct, but it was a feature trapped inside someone else’s checkout. The concept only becomes itself when it belongs to the owner, not the merchant, because care is a relationship with the things you already have, across every brand, for years. So the plugin became an app, and the merchant became, as we will see, the distribution channel instead of the product.

The timing argument is simple and two-sided. AI vision finally makes onboarding effortless. A photo is enough to recognize a garment, an instrument, or an appliance and generate a sane protocol, which kills the data-entry wall that murdered every previous attempt at this category. And the culture has moved. Repair, longevity, buy-less-but-better have gone from fringe virtue to mainstream common sense. People increasingly own fewer, better things on purpose. Nobody has given them the tool that honors that choice.

Key Features

Photo in, ritual out. Ten seconds per object: snap it, confirm what it is, done. No forms, no barcode hunting, no cataloging project. The clothing case is even better: photograph the care label itself and Keep translates those hieroglyphic laundry symbols into one plain sentence.
Care sheets in human words. What to do, how often, with what, and the two mistakes that actually ruin this kind of object. Written for someone who has never conditioned leather, not for a forum of enthusiasts.
Reminders with taste. This is the product’s soul, and its biggest risk, so it is a feature with opinions: few notifications, batched into a weekly care ritual, a quiet Sunday digest of the two or three things whose moment has come. An app that nags about fretboard oil on a Tuesday morning gets deleted by Thursday.
The service book. Every care act logged against the object, which sounds bureaucratic until you sell your turntable and hand the buyer a documented maintenance history, the same reason a car with a full service history is worth more. Cared-for objects hold value; Keep makes the caring provable.
Household sharing. The washing machine, the plants, and the good knives belong to everyone; so do their rituals.
Honest sourcing. Where an official manufacturer care guide exists, Keep finds it, links it, and defers to it. Generated advice fills the gaps, clearly labeled as such, and stays conservative on anything valuable or delicate. The app’s first rule is the medical one: do no harm to the patient.

The graveyard, and why Keep sidesteps it

Home-inventory apps are a documented graveyard, and the cause of death is always the same: they asked people to catalog their homes, and nobody on earth wants a second job as their own possessions’ librarian. Keep’s answer is structural. There is no catalog to complete, no progress bar guilting you toward exhaustiveness. One object gives value in ten seconds, the first reminder proves the point, and the collection grows the way a music library does, one addition at a time, whenever a thing feels worth keeping alive. The inventory is a side effect, never the task.

The second graveyard is generic reminder apps, which fail here for the opposite reason. They make you author every rule yourself, and nobody knows the right cadence for turntable belts. Keep’s knowledge layer, the care sheets with their built-in rhythms, is the part a to-do app cannot fake.

The honest tensions

Advice quality is the real risk, stated plainly. A wrong instruction can shrink a sweater or strip a fretboard, and an AI generating care protocols will sometimes be wrong. Hence the sourcing hierarchy above (official guides first, conservative defaults, and visible uncertainty), and hence starting with categories where the knowledge is stable and the failure cost is low. Notification fatigue is the second killer, addressed by design rather than promises: the weekly ritual, aggressive batching, and cadences that err long. And the market question deserves honesty too. This is a quiet utility, not a viral toy. Its growth engine is usefulness and gifting culture, and its natural scale is a beloved tool, not a social network.

Monetization

Free for a handful of objects, enough to fall for the Sunday ritual, then a modest subscription for unlimited objects, households, and the full service-book export. The second engine is the one my Shopify notes accidentally predicted: durable-goods brands as a channel. A knifemaker, a leather workshop, a hi-fi brand, anyone whose pitch is “this will last decades”, has every reason to slip a Keep card into the box. Scan it, and the object you just bought arrives with its care program preloaded, straight from the maker. The brand gets provably longer-lived products and a customer relationship that outlasts checkout, and Keep gets acquisition through the objects people already intend to care for. Brands pay for that placement. The owner never pays for the brand’s own instructions.

Next Steps: How I’d Build It

Start with five categories where care is codified and stakes are friendly: clothing via the label decoder, instruments, coffee gear, leather goods, plants. The label decoder might be the wedge all by itself (photograph the tag, get one sentence), because everyone has stood in front of a washing machine holding a garment like it was a bomb. Build the care-sheet layer with sourcing discipline from day one, ship the weekly ritual, and only then widen the catalog of categories.

The metric that matters is reminders acted on, months later, not objects added. Keep succeeds the day a user oils a fretboard in October because their phone, gently and correctly, asked them to.

Closing

Keep is the maintenance book everything deserved and only cars got, born from a Shopify note that outgrew the checkout, and from my own apartment full of well-loved, under-maintained things. It is the missing half of a philosophy already on this site. 5 Essentials was about buying less but better, and Keep is about deserving the things you kept. And somewhere between them sits Cloche, proof that I will apparently build hardware for a cheese before I change my guitar strings on time.

If you build this, tell me. My turntable and I will be your first subscribers, and this time the belt gets checked.

Homegrown: The World Map of Where Music Comes From

Tue, 03 May 2022 00:00:00 GMT

I am a musician, and there is a conversation musicians have that I have never seen a website serve. It happens in rehearsal rooms and after gigs. Someone says wait, you know that band is from here, right, and suddenly everyone is leaning in. That artist grew up two towns over. That legendary album was written in a suburb we all drive through. These are real discoveries, small and delicious, and we trade them like collectors, because knowing where music comes from changes how it sounds to you. The geography of music is a story we tell each other constantly, and nobody has drawn the map.

And there is a second conversation, the sadder cousin of the first. It is about the drummer you needed who, you learn a year too late, lives two streets away. Musicians are often invisible to each other at exactly the distance where collaboration is easiest.

Homegrown is one map that serves both conversations. A world map you zoom the way you zoom any map (country, region, city, neighborhood), except the pins are musicians. Two layers live on it. The atlas layer shows where known artists come from (their hometowns, their scenes), drawn from open music databases. Anyone can browse it, and it is fascinating from day one. And the living layer is the pins that musicians plant themselves, claiming their spot on the same map as their heroes, visible to the public and, more importantly, to each other.

The bootstrap hiding in the idea

My original 2024 notes made the classic two-sided mistake: build a musician directory, hope musicians show up, hope visitors find it interesting. Every platform in this graveyard died of that hope, and I will name the graveyard below. But the corrected version contains its own escape, and it was hiding in a sentence of my first voice note: the general public would come just to look at the map and be surprised.

They can, because the data already exists. The origins of known artists (their birthplaces, their formation cities, their scenes) live in open, community-maintained music databases with licensing designed for this kind of reuse. That makes the atlas interesting before a single person signs up. Zoom into any city and see who came from there; discover why one small town produced three cult metal bands; follow a genre across a coastline. The atlas needs no liquidity, no community, no luck. It is a good afternoon of browsing on day one, and it is the marketing for everything else.

Then the signup stops being a cold form and becomes an emotional act. You plant your pin on the same map as your heroes. Your dot on your city, next to the dots of the artists who made you pick up an instrument. The atlas draws the curious, the pin flatters the musicians, and the musicians make the map denser, which makes the atlas better. That is the flywheel.

Key Features

The zoomable atlas. Origins of known artists from open data: hometowns, scenes, eras. Search any artist and fly to where they are from; browse any city and see who it raised. The zoom and pan must be delightful, because wandering is the whole public product.
Scene pages. A city is more than dots: its page tells the scene through time, from the famous names the open data provides to the living musicians who claim it today.
Claimable pins for living musicians. Instrument, genres, a line of bio, two or three links out to wherever their music lives. Placement is by neighborhood or city, chosen by the musician, never an address.
Contact requests, then email. A musician can send another a contact request; if accepted, the platform hands both a temporary relay address and steps aside. My original note said it with the honesty of a tired builder: building a chat system is a pain, and email already works. I am keeping that reasoning in the article on purpose. Boring technology chosen deliberately is a feature, not a compromise.
Collaboration flags. A pin can quietly say looking for a band, open to sessions, teaches, so that browsing your own city answers the question the sad conversation always asks too late.
Privacy by grain. Living musicians place themselves at the precision they want, city by default. And a hard rule for the atlas layer: it shows where artists are from, sourced from public data, never where anyone currently lives. This map celebrates origins; it must be useless to a stalker.

The graveyard, named

Musician-finder platforms exist and have existed for two decades: swipe-to-jam apps, band classifieds, hire-a-session-pro marketplaces. Their shared cause of death is local liquidity. You sign up, your town is empty, you leave, and your leaving keeps it empty for the next person. A matchmaking product that only works with density, launched at zero density, is a machine for disappointing people one city at a time.

Homegrown’s answer is structural, not aspirational. The product does not need liquidity to be worth visiting, because the atlas carries it alone. The directory and collaboration layer activate wherever density happens to arrive, city by city, and lie dormant harmlessly where it has not. No empty-room experience, because the room is never empty. It is full of the dead and famous, which is the best possible crowd to warm up for you.

The honest tensions

The atlas layer inherits the limits of its sources. Open music databases are deep on some scenes and thin on others, with all the geographic bias that implies, and “where is this artist from” is occasionally contested or fuzzy. The map should wear its sources on every pin and take corrections gracefully, in the spirit of the databases it drinks from. Their licenses, which require attribution, deserve scrupulous respect, since they are the foundation of the whole house.

Claims need light verification, enough to stop pranksters from planting fake pins, without erecting a bureaucracy that scares off the shy bassist the whole site exists for. And the collaboration layer should stay modest on purpose. Homegrown is a map with a handshake, not a social network with feeds and follower counts. The moment it grows a timeline, it dies the death of every musician platform that forgot musicians already have too many profiles to maintain.

Monetization stays off the musicians entirely. Their pins are free forever, because their presence is the product. The honest revenue lives in the infrastructure around them: paid pins for rehearsal spaces, studios, luthiers, music schools and venues, the local music economy that has every reason to want to be on the map where its customers are browsing. The atlas stays clean, and the services around it pay the rent.

Next Steps: How I’d Build It

Phase one is the atlas alone. Import the open data, obsess over the zoom until wandering feels wonderful, and ship it with zero accounts. If people do not enjoy drifting through the map of where music comes from, stop there, having built something lovely anyway.

Phase two is one city, and I know which one: mine. Seed the living layer through the local scene (rehearsal spaces and open mics and the group chats where musicians already gossip) and watch whether pins beget pins. The claim flow, the relay email, and the collaboration flags are a weekend each, but the work is the wandering, the data hygiene, and the patience.

Closing

Homegrown is the map musicians already draw for each other in conversation, finally drawn for real, with room at every zoom level for the famous dead, the living obscure, and the drummer two streets away. It joins the other music ideas on this site, and this one makes the song a place.

If you build this, tell me, and load my city first. I have pins to plant and a few origin stories to correct, and somewhere nearby there is probably a bassist wondering where all the musicians are. We were here the whole time.

The Haven Program: Witness Protection for Survivors of Domestic Violence

Sun, 26 Dec 2021 00:00:00 GMT

I have carried this idea for a long time, but it stopped being abstract in the spring of 2020. During France’s first lockdown, within a single week, police interventions for domestic violence rose by roughly a third: 32 percent in gendarmerie zones, 36 percent in the Paris prefecture, according to the interior minister at the time. UN Women named the global surge the “shadow pandemic.” I remember sitting in my own confinement doing the arithmetic that the numbers implied. The same walls that were protecting the rest of us from a virus were locking thousands of people in with the person who hurts them.

Let me state the premise of this article without softening it, because everything else follows from it. Raising a hand against a partner or a child is an act of absolute malice and absolute weakness, and the people who survive it have the right to a new life, a real one, without the total fear of reprisal. That is not what they get today. Today, the state acts on the aggressor only when the violence is known, only when he can be found, and only when the courts can process him, three filters that each leak. And the survivor, meanwhile, is accompanied partially, unevenly, rarely for the long term. My conviction is simple. A far greater share of the effort and the money should be pointed at the person who was harmed.

A word on pronouns, because they will be read. I write he for the aggressor and she for the survivor because that is what the numbers overwhelmingly describe, not because the program has a gender. Victims come in every gender, and so do aggressors. The door described below is open to anyone who has to flee.

Every human being brings something to the world. No one should be held hostage by fear, suffering in silence, convinced they are the problem.

So here is the idea, stated plainly. We know how to protect people whose lives are in danger. We have done it for decades when the endangered person was a witness in a criminal trial. The Haven Program applies the structure of witness protection to survivors of domestic violence: risk assessment, extraction, confidential housing, a dedicated human protector, and months of wraparound support, until the danger is legally contained or the new life stands on its own.

The stakes, measured

This section exists because the danger this program addresses is one of the best-documented facts in violence research, not a feeling. Peer-reviewed work (Campbell et al., American Journal of Public Health, 2003) found that leaving a cohabiting abusive partner roughly tripled the risk of femicide, and that the combination of a highly controlling abuser and a separation multiplied it more than fivefold. The moment a victim finally does what everyone tells her to do, leave, is precisely the moment her risk of being killed peaks. In France, the interior ministry’s annual study counts around a hundred women killed by a partner or ex-partner every year: 118 in 2022, 96 in 2023, 107 in 2024. Roughly one every three days, year after year, many of them in that exact window this program is designed to cover.

What already exists, in fragments

Nothing in this idea starts from zero, and honesty requires the map. France’s Téléphone Grave Danger gives high-risk victims a phone with a direct 24/7 line to responders. More than 6,000 were deployed in 2024, triggering some 3,000 police interventions that year. The anti-rapprochement bracelet, operational since late 2020, GPS-tracks the aggressor and alerts police when he approaches. Over 2,500 had been ordered by end-2023. Judges can evict the violent partner from the family home and reserve it for the victim. Spain has gone furthest. The VioGén system risk-assesses every reported case and assigns graduated police follow-up, with roughly 104,000 cases under active monitoring at the end of 2025. In the United States, about 45 states run Address Confidentiality Programs that give survivors a substitute legal address. And everywhere, shelter networks operate confidential addresses, chronically saturated, and built for emergency, not for the long arc.

Look at that list and you see the shape of what is missing. Every fragment is real and every fragment is partial: a phone but not a protector, a bracelet but only after a court order, a shelter but only for the acute weeks, police follow-up but not a rebuilt life. No one has assembled the fragments into a program, one door that a person in danger walks through, behind which risk assessment, safe housing, human protection, and reconstruction happen as a single continuous service.

The program, concretely

The Haven Program is that assembly. A survivor whose danger is assessed as serious enters the program and receives five things. First, a structured risk assessment, in the spirit of VioGén, repeated over time rather than done once. Second, if the risk justifies it, relocation to secure housing that only she knows, deliberately dispersed, ordinary apartments in ordinary buildings, never clustered into an identifiable target. Third, a named protection officer, one trained human being responsible for her case, who checks in, coordinates police and justice contacts, and is reachable at 3am. Fourth, the reconstruction layer: legal support through the proceedings, psychological care, help with employment, schooling for children, the administrative surgery of a life moved. Fifth, a graduated exit, on her timeline, when the aggressor is contained by justice or the new life is established, with the door remaining open.

The duration is measured in months, sometimes longer. The unit of success is not a closed file but a person who no longer scans the street before leaving home.

The tension this idea must face

There is a moral inversion at the heart of any protection-by-hiding model, and this article will not walk past it. It is the victim who disappears while the aggressor keeps his name, his home, and his freedom. That is upside down, and the modern direction of policy, the bracelet on his ankle, his eviction from the home, is the correct one, because it constrains the person who created the danger. I hold both of these ideas at once, because the data forces me to. Constraining the aggressor is justice, and it is slow, requiring him to be known, found, and processed. Hiding the victim is unjust, and it is fast, and in the window where the femicide risk multiplies, fast is what keeps people alive. The Haven Program is therefore a bridge, never a destination. It protects during exactly the interval when justice has not yet done its work, and its own success metric, the one that should be printed in its annual report, is how short it can make the average stay. A society where this program is large and permanent has failed twice.

Why this is not a startup, and what it is instead

I will not insult this subject with a freemium model. The cost per protected person, housing, staff, security, months of support, is high, and the person being protected can never be the customer. This is a public mission. Realistically, it would be a delegated public service or a state-funded nonprofit, financed the way shelters and witness protection already are, by the state, supplemented by foundations. Call it an enterprise only in one sense. It would employ thousands (social workers, trained protection officers, jurists, psychologists, housing managers), and it would need the operational discipline of a serious company, because its failures would be measured in lives.

Two hard boundaries. The program works with police and justice, never instead of them. Protection officers are not vigilantes, not armed escorts, and the legal limits of private protection must be respected to the letter, or the program becomes a militia with a mission statement. And it must be built with the organizations who have carried this fight for decades, in governance and not as decoration. Years ago, my first sketch of this idea was an app. The more I learned, the clearer it became that an app on a phone the aggressor may inspect is not safety, and the people who taught the world that lesson are the specialized associations. They hold the expertise this program would run on.

Closing

The Haven Program is witness protection, extended to the people who need it most and get it least. It is built as a bridge across the deadliest months of a survivor’s life, pointed at a simple outcome, a new life, without fear, for people whose only mistake was trusting someone who was weak enough to use violence on those he claimed to love.

This is not a weekend build, and for once I will not end with a playful invitation. If you hold power over budgets or policy, in a ministry, a foundation, an NGO, this idea is yours, free, like everything on this site, and I would gladly help you think it through. And if you are living what this article describes: in France, call 3919; in immediate danger anywhere in Europe, 112. You are not the problem. You never were.

EmoTune: Music Discovery That Starts With a Feeling

Mon, 10 May 2021 00:00:00 GMT

Of all the ideas on this site, this is the one I actually built. Not sketched, not briefed. Built. EmoTune started in 2023 as a simple conviction, that we look for music by artist, genre, and era, when what we usually want is a feeling. I wired it up as a layer on top of Spotify. You logged in with your account, let the app read the emotional profile of tracks, and asked for what you wanted to feel instead of what you wanted to hear. The demos were good. Friends tried them and got playlists that landed.

Then I hit the wall, and the wall kept growing. Spotify’s public API grew steadily more restricted and more complex to build against, and in late 2024 the platform deprecated, for new third-party apps, the endpoints an app like this breathes through: the audio features that exposed a track’s energy and mood characteristics, the audio analysis, the recommendation engine. My abandoned prototype was not a failure of nerve or of code. The ground it stood on was repossessed. That story is worth an article on its own, so here is the idea in full, what killed it, and the version that deserves to live anyway.

The concept has not aged a day. You open EmoTune and, instead of a search bar, you are offered a spectrum. Not a crude “happy” button (that was my first design mistake, corrected early), but a range of nuanced states: calm, energized, melancholic, inspired, nostalgic, defiant. Next to it, an intensity slider, because “energized” at two out of ten is a bright pop song and at nine it is a wall of drums. Over time the app learns your personal emotional map, because what sounds melancholic to you and to me may be different songs, and that personalization, not the labels, is the real product.

The parable: killed twice by the same platform

EmoTune died two deaths, and both are instructive for anyone building on someone else’s platform.

The first death was absorption. While I was prototyping, Spotify shipped its own contextual mood curation, daylists that shift with your day, mood-tagged editorial everywhere. The platform looked at the same user desire I saw and folded a version of it into the core product. That is the first law of building on a platform: any feature valuable enough to build is valuable enough for the host to build without you.

The second death was the door closing. The late-2024 API deprecations did not target me personally, but they ended the category. An entire generation of stats apps, mood tools, and discovery experiments lost access to the analysis data they were built on, and new apps could not get it at all. That is the second law: an API is not infrastructure, it is a favor, revocable at any time, and a business built on a favor is a tenant, not an owner.

I am not bitter about either law. Platforms owe outsiders nothing. But I watched both laws execute on a working prototype, and the lesson is now load-bearing in the version below.

Key Features

The emotion spectrum. A curated range of feeling-states, deliberately richer than the five moods of editorial playlists, as the primary way in. You browse feelings, not genres.
The intensity slider. The feature I remain proudest of: every emotion has a volume knob, and the pair (feeling, intensity) maps to concrete musical parameters. It turns a vague mood into a precise request.
A personal emotion profile. The app learns from what you keep, skip, and repeat within each emotional context, building a private map of what actually moves you. The labels start generic; the profile makes them yours.
Context, gently. Time of day and season can pre-tint suggestions, a calm Sunday morning defaults differently than a Friday night, always as a suggestion, never as a cage.
Exploration, gamified with restraint. My 2023 notes said “make it almost addictive,” and older me wants to edit younger me: no leaderboards, no streaks guilt. What survives is the good part, gentle rewards for exploring the map, a badge for venturing into emotional territories you never visit, a weekly nudge toward one unfamiliar feeling. Design for curiosity, not compulsion.
Emerging artists as the center, not the garnish. My original plan sprinkled a quota of unknown artists into playlists. The rebuilt version inverts that entirely, and the next section explains why that inversion is what saves the whole idea.

The version that survives: leave Spotify entirely

Here is the pivot that turns both platform deaths into a mission. If the closed catalog will not let you analyze it, build on the music that wants to be analyzed: open-licensed catalogs, independent distribution platforms, and artists who opt in directly. The analysis layer no longer needs anyone’s API. Open-source audio analysis toolkits can extract energy, tempo, timbre, and mood-correlated features from any track you have the right to process. The stack that was a favor becomes a stack you own.

And suddenly the weakness is the positioning. EmoTune cannot be “Spotify, but emotional,” because the door is closed. It can be something better, the discovery engine for the vast, growing world of independent music that mainstream algorithms bury. An unknown artist’s track surfaces there because it matches what “nostalgic, intensity six” feels like, not because a popularity flywheel chose it. Emerging musicians get something they cannot buy, a discovery path that ignores fame entirely. Listeners get music they could not have found. Nobody’s API can take that away, because the catalog’s whole premise is permission.

The honest tensions

Emotion detection is crude, and the article should say so plainly. Musical emotion is personal and cultural. A model tagging a track “melancholic” is making a statistical guess about averages, and it will be wrong for you regularly. EmoTune’s defense lies in its architecture. The global labels are only a starting grid, and the per-user profile, learned from your actual behavior, does the fine-tuning. An app that pretended the labels were truth would deserve its bad reviews.

The wellness line must be walked carefully. An emotion app will be tempted to market itself as mental-health adjacent (my old notes even suggested mental-health influencer partnerships), and that temptation should be resisted in writing. EmoTune curates moods, it does not treat them. No therapeutic claims, no “scientifically proven to reduce anxiety,” nothing the concept cannot support. Music affects how we feel. That observation is enough, and overclaiming it is both dishonest and, in some jurisdictions, regulated.

And the catalog gap is real. Leaving Spotify means losing the songs people already love, which stings on day one. The bet is that emotion-first discovery is exactly the mode where not knowing any of the artists is acceptable, even desirable. Nostalgia for your own library needs Spotify, but discovery of your next favorite unknown does not.

Monetization

The free tier has the spectrum and daily discovery, and a modest listener subscription unlocks the full profile, offline mixes, and unlimited exploration. On the artist side, one red line and one open door. The red line is that artists never pay for placement, because an emotion engine that sells its feelings is a payola scheme with extra steps. The open door is analytics. Artists and indie platforms can pay to see how their music lands emotionally, with which audiences, at what intensities. They cannot get that knowledge anywhere else, and selling it does not corrupt the listener side. If the engine gets very good, licensing the emotion-mapping layer to indie distribution platforms is the B2B future.

Next Steps: How I’d Build It

Start with a few hundred tracks from open catalogs, run the open-source analysis stack over them, and rebuild the demo I already had (spectrum, slider, and a first crude map). I have the advantage of having done this once. The UX questions are already answered in my old prototype, and the parts that died were never the parts I built.

Then grow the catalog through artist opt-in, ship the personal profile, and treat the emotion labels as a living, versioned model rather than a truth. The hard parts are the cold catalog (solved by mission, starting with one scene or one label that wants this), the label quality (solved by iteration and user correction), and the discipline to keep the wellness marketing honest when growth gets tempting.

Closing

EmoTune is the idea I built, lost to two laws of platform gravity, and still believe in, now rerouted through the one catalog that cannot evict it. It shares blood with WaveArt, another of the music ideas on this site. Both start from the conviction that a song is an emotional object first and a file second.

If you build this, tell me, and I will happily compare your prototype with the ghost of mine. And if you are an independent artist reading this, you are not the garnish in this idea. You are the reason it survives.

Dark Pleasure Circuits: A Dystopian Shop, Open for Business

Mon, 21 Dec 2020 00:00:00 GMT

This one started with a feeling every internet user knows, that modern e-commerce is already halfway to dystopia. The popup that begs as you leave, the countdown timer lying about scarcity, the subscription hiding in the checkout, the “we value your privacy” banner that values nothing of the sort. I kept looking at those dark patterns and asking the science-fiction question my bookshelf trained me to ask. Where does this end? If nothing bends the curve, what does the storefront look like in a century where the ads no longer interrupt your screen, because they install directly into your neural implants, immediately, with a smile?

Most of my ideas stay on this site as blueprints. This one would not leave me alone, so I built it. A year and a half of evenings later, the answer is live at dark-pleasure-circuits.com. It is the storefront of Dark Pleasure Circuits, a subsidiary of the megacorporation Omnitech-K, purveyor of premium body modifications to citizens who are, perhaps, a little too agreeable about it.

The visitor arrives on what looks like an ordinary shop: products, categories, search, cart, promo ticker. Then the tone starts to slip. The reviews praise a memory eraser for helping with “unwanted loyalty conflicts.” The warnings announce irreversible effects with the cheer of a returns policy. A corporate support agent answers with glacial politeness. And for the visitor who cannot leave a locked door alone, there are doors: five secret codes hidden across the site, a clearance ladder that climbs from NONE to OMEGA, and a classified catalog considerably darker than the shelf stock. No spoilers here. The catalog is watching you too.

Why a shop is the perfect stage

Here is the idea underneath the idea, the part you can take even if cyberpunk is not your genre. E-commerce is the most universally understood interface on the internet. Nobody needs a tutorial to read a product page, a review section, a price, a cart. So every one of those conventions can carry fiction. A spec sheet becomes worldbuilding, a review becomes a character, fine print becomes horror, an out-of-stock badge becomes a plot point. The visitor already knows how to explore your world, because commerce taught them the controls.

And the format is sitting there, barely used. A game studio could launch a title through its faction’s in-universe shop. A musician could sell an album from inside its concept. A novelist could give the book’s sinister corporation a web presence that outlives the print run. The fake storefront can carry satire, marketing, an art project, or a portfolio piece, and the mechanics transfer whole. Familiar interface, wrong details, layered depth, and the visitor’s own curiosity as the engine.

Key Features

A surface that works as a complete experience. The satire, the absurd products, the aggressive install-now popups: a passing visitor gets a full show without discovering a single secret. The layers never pollute each other; deeper content only ever rewards the behavior that went looking for it.
A five-tier clearance ARG. Five codes hidden in product specs, reviews, fine print, and corporate branding unlock real content: hidden products, extra sections, and a community dashboard where visitors see what everyone has collectively found.
Several editorial voices, one world. Fifty-two thousand words across the catalog and its in-universe articles, held together not by one narrator but by several who disagree: the clinical Compliance division, the chirpy catalog team, the anonymous citizens. Coherence comes from the world they share, not from uniform prose.
A deterministic ASCII art generator. No product photos: every SKU renders its own unique terminal-style visual, generated client-side from its identifier. A thousand products, zero image assets, one aesthetic.
A support chat with no AI in it. Scripted, glacially polite, occasionally more unsettling than any model would dare to be. Visitors assume it is an LLM; it is writing.
Accessibility under the glitch. The genre loves low contrast and permanent motion; the site refuses both. AA contrast, focus-trapped modals, every scanline and glitch honoring reduced-motion preferences. The dystopia is opt-in; the readability is not.

The honest tensions

The first is the invoice that never comes. Eighteen months is a long time for a project with no client, and pretending otherwise would break this site’s house rules. So, plainly, some projects exist to explore a format, to feed the communities that love this kind of web, and to serve as a calling card for the day a fully open brief arrives. That return is real, just not denominated in currency. The trap is not knowing which kind of project you are in. Decide before you start, or watch a gift to yourself quietly become a resentment machine.

The second tension is scale versus tone. A universe stretched over a thousand products and dozens of articles wants to go repetitive or wobbly, and both kill the joke. The defenses cost real editorial discipline: distinct voices with their own vocabularies, and a satire that never winks. Nothing on the site admits it is a joke, from the promo ticker to the terms of service, and that straight face is precisely what makes the comedy and the horror land.

And the third is the search engine problem, which turned into my favorite trick. How do you get indexed for real queries about your genre without breaking character? The answer was to write the informational content from inside the fiction, as declassified archives annotated by the corporation. Search engines now index a dystopia that refuses to admit it is one.

Next Steps: How You’d Build Yours

Start with the world, not the code. A fictional storefront with ten products and one perfectly consistent voice beats a thousand products of wobbling tone. My scale came from years of accretion, not from a requirements doc. Make the surface complete before hiding anything under it, then hide your first secret exactly where your most attentive visitor is already staring. Constraints will do your art direction if you let them. My ASCII generator exists because I refused to produce a thousand product images, and it became the site’s visual signature.

The stack matters less than the discipline: static pages, tests on the flows that must not break, and accessibility treated as a launch requirement, because the most immersive thing a weird site can do is work flawlessly. Budget the writing like an engineering line item, because it is most of the product.

Closing

Dark Pleasure Circuits is my answer to “where do the dark patterns end,” built instead of blueprinted, and living in Now Running with the other things that made it off this site’s pages. The shop is open at dark-pleasure-circuits.com. Try to install more than five modules. Talk to support. See how deep the clearance ladder goes.

And the format is yours, like everything here: build your own impossible shop for your own impossible world, and if you do, tell me. If a product page ever feels like it is watching you back, be polite. Compliance is monitored.

Curtain Call: Make an Empty Home Look Lived-In

Mon, 05 Oct 2020 00:00:00 GMT

This summer I house-sat for friends who went away for three weeks. Their only real instruction had nothing to do with the plants or the cat. It was: “just make it look like someone’s home.” So a few times a week I drove over, opened and closed some curtains, left a different light on, nudged the car forward in the driveway, dropped the mail inside, cracked a window. Ten minutes of theater, then I left.

Standing in their dark living room one evening, flicking a lamp on for an audience of nobody, it struck me how little “presence” actually is. A burglar walking a street at dusk is not checking for people. They are reading signals: a light that changed since yesterday, a shadow crossing a window, a rhythm that says someone lives here and could walk back in at any moment. I was faking those signals by hand. Badly, and only when I remembered to.

The thought that stuck with me was that the single most convincing signal, the one no timer-plug can fake, is a human shape moving behind a curtain. You catch it from the street and your brain fills in the rest without asking permission. So why can’t a window do that on its own?

Curtain Call is a thin rear-projection film you stick to the inside of a window, plus a small projector that throws choreographed silhouettes onto it from behind. From the street it reads as someone crossing the room, pausing at the sink, getting up in the night. Not a looping clip, but scenes with believable timing. And because the interesting part is the performance itself, the scenes are content. You make your own or download them from a shared library.

The house never needed me. It needed the ten minutes of theater I brought with me. Curtain Call is that ten minutes, automated and handed to a timer that actually thinks.

Market & Opportunities

Simulating an occupied home is an old idea with a real market. People already buy timer plugs to cycle their lamps, and FakeTV, a little box that mimics the flicker of a television, sells precisely because a changing glow behind the curtains reads as “someone is up.” The appetite is there. What nobody has done well is the part that actually convinces: a moving human shape.

The clearest buyers are the people who leave homes empty on purpose. Frequent travelers and second-home owners are the obvious core. Renters are an underserved segment right next to them. They cannot bolt a security system to a wall, but they can stick a film to a pane of glass and peel it off when the lease ends, no drilling and no landlord to ask. There is also the quieter case of adult children who want a parent’s house to look occupied and less like a target, without turning it into a surveillance site.

Why now is mostly about cost. Pico projectors are cheap and finally bright enough, rear-projection film is a commodity you can buy by the roll, and the piece that turns a gadget into something people come back to, a library of scenes you can browse and swap, is just software. The hardware has been possible for a while. The content layer is the part that was missing.

Key Features

Peel-and-stick rear-projection film. Renter-friendly by design: it goes on the inside of the glass, needs no drilling or permission, and comes off cleanly. The illusion lives on a surface you were never going to modify anyway.
Choreographed silhouette scenes. Not a loop. Scenes carry randomized pauses, entrances, and exits, so a figure crosses, leaves, and comes back later instead of pacing on a visible three-second cycle that gives the whole thing away.
A scenario library you create or download. This is the heart of it. A dinner being cleared, someone pacing on a phone call, a night owl padding to the kitchen at 2am. Design your own and share it, or grab someone else’s, offered under the same open license as everything here, so the library grows without anyone owning it.
A rhythm engine. Scenes follow a believable daily arc and sync to local sunset, so nothing “performs” at noon into a bright empty room. The point is to match how a real household actually moves through an evening.
Multi-window coordination. Different rooms show different figures, so it never looks like one silhouette teleporting around the house. Coverage reads as a family, not a ghost.
Optional layering. Trigger a smart bulb or a low, muffled murmur of sound alongside a scene, for the nights you want the full performance rather than just the shadow.

Monetization

Three paths, and the first is the honest anchor. Sell the kit, film plus projector, as a one-time purchase, bundled by the number of windows a home wants to cover. This is the version most people just want to buy once and forget.

Second, a scenario marketplace. The community library stays free, because that is the spirit of the thing, but designers can sell polished premium packs (a convincing dinner party, a believable “someone working late”) and the platform takes a cut. Third, an optional subscription for people who want the app to do more: cloud-synced scenes across several homes, seasonal packs, smarter scheduling. Kept deliberately optional, so the base product works fully without it. As a rough sense of scale, a single-window kit sits comfortably in the price range of a mid-tier smart bulb setup, not a security system.

Next Steps: How I’d Build It

The MVP is almost embarrassingly cheap. One window, a phone or a pico projector behind a sheet of rear-projection film, and three or four hand-animated silhouette clips with randomized gaps between them. Then comes the only test that matters. Stand in the street at dusk and see whether it fools you.

The hard part is not the hardware but the craft of convincing. A flat, two-dimensional shadow can look flat, from a sharp angle, under a streetlight, or the moment a loop repeats itself and the eye catches the seam. Getting “random but believable” right, the timing, the scale of the figure against a real room, the way it should never quite repeat, is the actual product. Projection also gets washed out by bright ambient light, so this works best at dusk and after dark, which happens to be exactly when it matters most.

To be clear, this is a deterrent layer, not a lock. It raises the odds that someone picks the obviously darker, quieter house next door instead of yours. It does not stop a determined burglar, and it should never be sold as if it does. Pair it with the boring, real measures. The upside is that nothing here depends on a closed API or a change in the law. It needs a good film, a bright-enough small projector, and a library worth browsing.

Bonus: the most convincing scene is the most boring one

When I tried to picture the scene that would sell the illusion best, it was not anything dramatic. It was a figure wandering to the kitchen and just standing there, lit by the open fridge, doing nothing in particular. Presence, it turns out, is mostly boredom rendered faithfully. One rule, though, sums up the ethics of the thing: keep the performance on your own windows. The goal is to look home, not to project anything at the neighbors.

Closing

Curtain Call is a cheap film, a small projector, and a library of quiet human moments that make an empty house lie convincingly on your behalf. It is yours to build, and you owe me nothing for it.

If you build this, tell me. I would love to stand in a street at dusk and fail to tell whether anyone is home. And if you want a second pair of eyes while you get the illusion right, I am glad to help.

Cloche: The Humidor, but for Cheese

Sat, 04 Jul 2020 00:00:00 GMT

I will introduce this one the way I first dictated it into my notes: a business idea that is possibly stupid, that I nevertheless refuse to let go of. I am French. Cheese, in my house, is not a snack but a small sacred ritual with its own course at the table. And so, like a serious person, I keep a dedicated cheese cupboard, indoors, out of the refrigerator, as the affineurs say you should. That puts my cheeses at the mercy of my apartment, through radiator winters, air-conditioned summers, and the heat waves in between. Some weeks the cupboard smells like a triumph. Other weeks I unwrap a Saint-Nectaire that has clearly suffered, and I feel the specific guilt of a man who has mistreated something noble.

The absurdity struck me one day between the radiator and the AC remote. We solved this exact problem for other sacred indulgences long ago. Cigars have the humidor. Wine has the electric cellar humming in a million dining rooms. Both are just boxes that hold a living product at a precise temperature and humidity, and both support proud, prosperous industries of enthusiast hardware. Cheese, a product that is literally alive, that wants somewhere around 10 to 15 degrees Celsius and 85 to 95 percent humidity, that is ruined by the fridge’s cold dry air and slowly betrayed by a room’s mood swings, has nothing. A drawer. Some paper. Good intentions.

Cloche is the missing object, a climate-controlled cheese bell. A handsome glass dome or small cabinet for the counter or sideboard that holds, actively and precisely, the atmosphere your cheeses actually want, whatever your radiator is doing.

Why the fridge is the enemy, and why the cheap fixes almost work

The refrigerator is designed to stop biology, and cheese is biology you paid for. At four degrees in dry circulated air, rinds crack, pâtes tighten, aromas go to sleep and never quite wake up. Every cheese lover knows the ritual of pulling the plateau out two hours before guests arrive, hoping the flavors return in time. Meanwhile the passive alternatives (cheese paper, a wooden box, a terracotta cloche, my cupboard) get you most of the way there on humidity and airflow. What they cannot do is hold a temperature through July, or through the heating season, or keep five different cheeses from arguing.

That last point matters more than it sounds. A proper plateau is a cohabitation problem: the blue that wants to colonize everything, the washed rind broadcasting its opinions, the delicate chèvre absorbing everyone else’s. My original note already sensed the answer (several zones rather than one), and that instinct turns out to be the core of the product. Each compartment gets its own humidity setpoint and airflow, so the roquefort and the comté can share a home without sharing spores.

Key Features

Closed-loop climate. Peltier cooling and gentle humidification holding a chosen setpoint, cellar conditions on your sideboard, regardless of season. Technically this is a small wine fridge rethought around humidity and airflow, which is precisely why it is buildable.
Compartments against contamination. Two to four zones with independent humidity and filtered air paths: soft rinds here, hard pâtes there, the blue in respectful quarantine.
Family presets, on a dial. Pâte molle, pâte dure, bleu, chèvre. A knob and a small display, not an app. Nothing on this object should ever need a firmware update to keep cheese happy, and no cloud service will ever be notified about my camembert. Sensors, yes; internet, no.
Service mode. The one clever trick: tell it dinner is at eight, and it walks the chosen compartment up to tasting temperature on schedule, so the plateau arrives at the table exactly ready, no two-hour countertop gamble.
Odor management. Replaceable charcoal filtration, because the difference between “a beautiful cheese cupboard” and “a problem in the living room” is airflow engineering, as my own casier reminds me weekly.
The pro line. For restaurants, the same climate engine built into the service trolley and the back-of-house cabinet, with one connected feature worth having: automatic temperature logging for hygiene compliance, the paperwork nobody enjoys doing by hand.

The honest tensions

The thirty-euro question comes first. A passive cheese box with good paper costs almost nothing and does most of the humidity work. Cloche has to earn its price on the part passives cannot do (temperature through the seasons, multi-cheese peace, and the service-mode ritual), and this article says so plainly rather than pretending the baseline does not exist. If those three do not seem worth real money to you, you are not the customer, and that is fine. The humidor is not for people who smoke twice a year either.

Food safety comes next, stated carefully. This is a preservation and service object, not a license to ignore biology. Soft raw-milk cheeses have real handling rules, home users should follow their fromager’s advice on what belongs outside a refrigerator and for how long, and the professional version lives inside existing hygiene regulations, which is why the compliance logging exists. Cloche makes good practice easier, but it makes no health claims and grants no exemptions.

Last, the market. Cloche sits in a niche within a niche, strongest in France and wherever cheese is taken personally, a passion object rather than a mass appliance. I consider that a feature. Enthusiast hardware with real engineering and a clear ritual has built durable businesses many times, because passionate niches buy deep, gift often, and forgive premium pricing when the object is beautiful. This should be a beautiful object. That is half the spec.

Monetization

Premium hardware at enthusiast pricing, in two sizes for homes and a pro line for restaurants. The recurring revenue is consumables (the charcoal filters and humidity cartridges), priced fairly rather than printer-ink style, because this brand lives or dies on the affection of a community that talks to each other. The pro line carries service contracts. And the gifting channel matters more than usual, because “for the person who has everything and loves cheese” is a real and generous shelf.

Next Steps: How I’d Build It

The prototype path is short and cheap: a small wine fridge, a hygrometer, an ultrasonic humidifier on a controller, and a summer of measuring what actually happens to a comté in there. The real engineering is in airflow and odor, not in the cooling. That means keeping compartments genuinely separate, keeping the seal honest, keeping the living room fresh. That is prototype two.

Then the enthusiast community before any factory: fromagers, cheese clubs, the people who already own wine cellars, ten hand-built units in passionate homes, including, obviously, mine, replacing the long-suffering casier. Their feedback decides whether this becomes a product, and their word of mouth decides whether it becomes a brand. Restaurant pilots come after the home version proves the climate engine, because pros forgive nothing.

Closing

Cloche is the humidor that cheese never got. It was born from a cupboard that could not cope with my radiator, and built on the conviction that a sacred food deserves to be tasted as it should be, at its temperature, in its humidity, at its hour. It is yours to build, compartments, charcoal, dial and all.

If you build this, tell me, and I will volunteer my apartment as the test kitchen, radiator and all. My Saint-Nectaire has suffered enough.

Astranus: The Office Toilet League

Sun, 19 Jan 2020 00:00:00 GMT

Every workplace has one ritual that would be impossible to explain to an outsider, and ours involved a stopwatch. At some point, in an office I will not name, it became tradition that whenever someone headed for the toilets, somebody else discreetly started a timer. Scores were kept. Records were remembered. And at the end of each month, the colleague at the top of the leaderboard, the reigning champion of the porcelain, paid lunch for everyone. It was completely useless, mildly juvenile, and one of the most reliable sources of collective joy that office ever produced.

I know exactly how I justified it at the time, because I still have the job post I wrote in January 2020, when I decided this ritual deserved software:

Yes, that’s useless and stupid, but it gives us something to laugh about as a group, and a free lunch is always very satisfying.

That job post is its own confession. I budgeted 700 dollars, stated plainly that the project was “totally stupid,” asked for one page, one big chronometer, three buttons and a poop gauge, and posted it under a pseudonym, out of some instinct that my professional name and this project should meet as late as possible. Then a freelancer from the other side of the Atlantic answered, and the story got better than the product.

What Astranus is, and what it is not

Astranus is the office toilet league. A web app where a team, your actual colleagues, tracks sessions on a shared scoreboard: a big timer, a gauge that climbs from a green “relaxed” to a red “stressed” as the seconds pass (half a percent per second, so two hundred seconds to full alarm, a calibration decision I stand behind), personal logs, team stats, a monthly champion, and a newslog that reads like a small-town gazette where every headline is about the same thing. Pseudonyms are first-class citizens, because the leaderboard wants “The Poop King,” not your LinkedIn name, a design point my developer raised himself, with great professional seriousness.

It is not Chronopiss, and the two deserve their border drawn. Chronopiss is a gadget: a box above your own toilet, a number, a nudge, a private habit. Astranus is an institution: teams, seasons, archives, a champion, and a lunch debt. One instruments a bathroom, while the other federates an office. The hardware counts. The league competes.

And one rule sits above all the features, because it is where the joke becomes governance. The monthly champion pays lunch. The game taxes its own winner. It is self-balancing (nobody wants the crown too badly), self-funding (the prize is the group eating together), and self-limiting (the stakes are a sandwich). I have designed payment systems with less elegance than this rule, and I did not even design it. The office did.

The story: 700 dollars, a pandemic, and the most serious man I ever hired

What makes Astranus worth an article rather than a footnote is that I actually built it. Or rather, a freelance developer built it, between January and July 2020, at twenty dollars an hour, and he treated the least serious project I have ever commissioned with the gravity of a banking contract.

He rebuilt the team-join flow until requests resolved in real time with no page refresh, and recorded video tutorials, with his own voice, demonstrating each flow. He moved all timestamps to UTC so that, and I quote the spirit of his update, teammates on different continents would compete on the same time base, a globalization concern for an app with four users. He built yearly archives and defended them. After all, we would not want 2021 to be unable to display 2020’s records, would we? He bought a UI theme out of his own pocket because he believed in the project.

He advised me, with a straight face, to start approaching sponsors, reasoning that if the game got popular, brand representatives should not care what it was about. When soda destroyed his computer, he fought support teams for a week to recover the files and came back apologizing for the delay. Between updates we argued about football. He held strong and correct opinions about underrated free-kick specialists.

All of this happened across the spring of 2020, which is to say through the first lockdowns, two strangers on two continents, checking on each other’s health in every message, building the world’s least essential application while the world shut down. I have shipped far more important software with far less humanity in the process. Somewhere in that contradiction is the entire reason this site exists.

By July the app was, in his estimate, ninety percent done. In August we paused, “just a matter of money,” I wrote, planning to resume in September. The app did not die of failure, or of ridicule, or of technical debt. It died of August, the way most side projects do. Not with a decision, but with a month that quietly became six years.

The honest tensions

One tension towers over the rest, and the league lives or dies on it. The office toilet tracker must never, under any circumstance, be visible to management. A game colleagues play on themselves is joy, but the same data in a dashboard one level up is workplace surveillance with a rocket logo. Astranus is structurally by the team, for the team: opt-in, pseudonymous, data isolated per squad, with nothing exportable and HR permanently uninvited. The lunch rule only works between equals.

The second is the line this site already drew for its sibling. No health claims, none, ever. Chronopiss carries the carefully caveated science. Astranus carries none, because a league does not need a rationale. It is a game. The moment it pretends to be wellness, it becomes both a worse game and a lie.

And the third is honest scale. This is a team-size product, not a network. Its natural unit is one office, five to fifteen people, one running joke. It spreads the way jokes spread, one office telling another at an afterwork, and no other way. There is no venture math here, only a lunch.

Monetization

Barely, and proudly. The core stays free because charging colleagues to laugh together is grim. A small one-time fee per team for the deluxe trimmings (custom gauge themes, the poop-rocket launch animation on new records, printable end-of-season certificates for the champion’s desk). And the true economy of Astranus never touches the app at all. It is denominated in lunches, and the app is merely its central bank.

Next Steps: How I’d Build It

Strangely, this is the one idea on this site where the next step is archaeology. A Firebase project from 2020 exists, ninety percent complete, with team flows, archives, and a working gauge. My engineering assessment is that reviving six-year-old code probably loses to rebuilding it in a weekend with today’s tools, and my personal assessment is that the rebuild matters less than the launch: one office, one month, one lunch actually paid. That was always the missing ten percent, and no framework fixes it.

If I ever do relaunch it, the spec carries over untouched. The big timer, the three buttons, the gauge, the champion, the lunch. Six years of product fashion have not improved on what the office invented by itself.

Closing

Astranus is a stopwatch joke that became a league, a 700-dollar job post that found the most dedicated collaborator of my career, and a ninety-percent-finished monument to the kind of stupid that holds a team together. Somewhere out there is an office that needs this. Statistically, it is probably yours.

If you build this, tell me, and invite me to the launch. I will lose the first month’s championship on purpose, let someone else buy the lunch, and consider every hour of 2020 finally shipped.

Breachline: A Zoomable Timeline of Hacks, Bugs, and Breaches

Wed, 30 Oct 2019 00:00:00 GMT

The thought that started this one arrives every election season. As campaigns heat up, the news fills with breaches, leaks, botnets, and influence operations, and I catch myself feeling that the attacks are intensifying, that somewhere behind the headlines there is a rhythm: quiet years, then surges, clustered around the moments when public opinion is worth the most to manipulate. And every time, I want to actually see it. Zoom out to thirty years of incidents, watch the density rise and fall, drag the view across 2016, 2020, 2024, and check whether the wave I feel is a wave that exists.

I could not find any place to do that. The feeling stayed a feeling, unverifiable either way, which is the kind of itch this site exists for.

Breachline is the instrument I wanted. It puts the entire dark history of computing on a single, beautiful, zoomable timeline. Every notable bug, virus, worm, breach, vulnerability, ransomware wave, and state-linked operation, from the Morris worm of 1988 to LoveLetter in 2000 to WannaCry in 2017 to whatever happened last Tuesday, each one a dated, categorized, sourced point on one infinite horizontal line. Scrolling is traveling through time, and zooming moves from decades to years to days. At a distance you see density, eras, surges. Up close you see events, each opening into a summary, its categories, and its sources.

The assembly is the product

My research turned up something strange. Everything needed for this exists, and the thing itself does not. CSIS maintains a serious list of significant cyber incidents, but it is a chronological list, not a visual instrument. Wikipedia’s timeline of viruses and worms has historical depth, as linear text. There are static pedagogical timelines, generic contributive timeline platforms with no security focus, and at least one lovely open-source zoomable timeline engine with no cyber content in it. Each piece has the flaw the others solve.

So Breachline is an assembly job, not a research project or a data-collection moonshot. It takes proven engine patterns, existing openly licensed data as the seed corpus, and one opinionated interface. The design principle from my earliest notes still holds and deserves to be printed verbatim: the scroll must be delightful. If dragging through three decades of digital catastrophe does not feel like flying over a landscape, nothing else about the project matters, because a timeline’s entire value is the desire to keep moving through it.

Key Features

One line, infinite zoom. Decades collapse into density bands; days expand into individual incidents. No pagination, no per-year pages, no PDF. The navigation is the interface.
Events as sourced objects. Every point carries a date, a title, a short summary, categories, and at least one source. No source, no event; that rule is absolute and enforced at submission.
Categories as color, impact as size. Malware, vulnerabilities, data breaches, ransomware, state-linked operations, influence campaigns, each a color; the size of a point reflects impact against published criteria. Any single impact score, though, is an editorial judgment dressed as a number. The criteria are public, arguable, and versioned, which is the only honest way to run one.
The density layer. The founding feature, for my founding itch: a view that renders incident frequency over time as a landscape, with optional context overlays, election periods, major conflicts, so anyone can put the pattern they suspect next to the record and look. The tool draws the two lines; it refuses to draw the conclusion.
Attribution with confidence levels. This is where a cyber timeline either grows up or becomes a propaganda surface. “Attributed to” is, in this field, a claim made by someone: a government, a security vendor, a journalist. Breachline never states that a state did something; it states who made the attribution and with what confidence, on the event card itself. Contested attributions show as contested.
Public contribution, boring on purpose. A submission form with mandatory fields and mandatory sources, a proposed/validated/rejected status, and a visible edit history. No gamification: points and badges attract volume, and volume without verification is how a sourced timeline rots.
Made to be embedded. Any slice of the line, a year, an incident cluster, a category, can be embedded in an article, a course, or a conference deck. Embeds are the distribution.

The honest tensions

First, the founding question must stay a question. Suppose the density view does show surges around elections. That still would not prove what my gut believes, because the timeline records documented incidents, and documentation has its own rhythms: security vendors publish more during news cycles, journalists look harder when elections loom, and disclosure laws changed what gets reported at all. More visible attacks and more attacks are not the same fact. Breachline’s job is to make the record visible and let the reporting bias be part of the conversation, not to launder my hunch into a chart. If anything, making that distinction legible to normal people would be the project’s finest contribution.

Second, the attribution minefield deserves a second paragraph, because influence operations are the category where this project touches geopolitics. The 2016 intrusion into a US political party, attributed by US intelligence agencies to Russian services, is among the most documented events of its kind. Plenty of others rest on a single vendor report and a preponderance of maybes. A timeline that flattens that spectrum into equal-looking dots becomes disinformation about disinformation. Encoding the uncertainty is the moral core of the product, not a nice-to-have.

Third, the treadmill. Incidents happen weekly, and a timeline that visibly stops updating dies in public, the emptiness after its last event reading as abandonment. Sustainability has to be designed in: semi-automated imports from the maintained public lists, a small review crew, and an institutional partner, a university lab or a security nonprofit, as the long-term home.

And fourth, retention. Timelines are visit-once products. People come, fly through it, feel things, and leave. That is fine, because this is a resource, not a startup. Its success metrics are citations, embeds, classrooms, and journalists who reach for it on deadline, not daily active users.

Monetization

The answer is the same as for my Obscurantism Clock, this project’s sibling. It is a civic project, and pretending it is a venture would ruin it. Costs are nearly nothing, a static-first architecture on free hosting with versioned JSON data. Sustaining it looks like grants in the digital-literacy and security-education space, sponsorship credits from security firms that want association without editorial influence (a line that must be written into any deal), and perhaps a supporter tier for the clean embed API. The data itself stays open, because the seed corpus came from open sources and the project’s legitimacy depends on giving back under the same terms.

Next Steps: How I’d Build It

Phase one is a static prototype with a hundred events in a single category (malware history is the natural pick, since Wikipedia’s corpus is deep and openly licensed) and one obsession: making the zoom and scroll feel wonderful. No accounts, no submissions, no filters. If a hundred points on a line do not make a visitor want to keep dragging, stop there and lose nothing.

Then licensing hygiene before scale. Wikipedia content is CC BY-SA and requires attribution and share-alike, the open timeline engines and event archives on GitHub carry their own licenses to respect, and institutional lists like CSIS get cited and linked, never scraped wholesale. An afternoon of reading licenses costs nothing; retrofitting a corpus does.

Then the categories, the density layer, the attribution model, and only after all of that, public contribution with its review process. Community is the last brick, not the first, because an empty review queue with clear rules beats a full one with none.

Closing

Breachline is the dark history of the connected world on one line you can fly over, built to show patterns without asserting them, and to treat “who did it” with the honesty that question deserves. It belongs to the same family as my Obscurantism Clock, instruments for seeing slow alarming things, and it shares blood with ThreatShield too. Both are tools for a web that has never been more useful or less trustworthy.

If you build this, tell me, and send me the link with the view zoomed out to 1988. I want to see the whole landscape at once, and then I want to drag it, slowly, toward now.

Exposed: Count the Ads You Actually See

Tue, 21 May 2019 00:00:00 GMT

Everyone knows the feeling of being followed around the internet by a pair of sneakers. You look at a product once and it shadows you for three weeks, across news sites, videos, weather forecasts. What nobody can tell you, and this is the strange part, is the number. How many times, exactly, did that brand appear in front of your eyes this month? Ten? Two hundred? You are the only witness of your own screen, and even you are not counting.

I wrote this idea down years ago and forgot it, which is itself a data point about how ambient the problem is. Advertising is the only massive daily exposure in our lives that nobody measures from the receiving end. We count our steps, our calories, our screen minutes. The ad industry counts its impressions with terrifying precision, from its side.

The one ledger that does not exist is yours.

Exposed is a browser extension that keeps that ledger. It does not block anything. It counts. Every ad that renders in your browser gets tallied, by brand where identifiable, by category, by site, by hour, and everything stays on your machine. At the end of the week you get the report nobody has ever handed you: your ad diet. You were served roughly 1,400 ads this week. Here are your top ten brands. Here is the fast-food cluster that owns your evenings, and the betting brands that own your weekends.

Blockers count what they stop; nobody counts what you absorb

The adjacent tools make the gap obvious. Ad blockers proudly display how many things they blocked, a number about their work, not your exposure, and by definition they only see what they removed. Mozilla once shipped Lightbeam, a beloved little extension that visualized the trackers following you across sites; it was discontinued, and it tracked the trackers, not the ads themselves. Academic projects have tried to observe advertising at scale, and the most famous one, a university tool for studying political ads on a large social platform, got legal threats from the platform for its trouble. That tells you two things: observation is valuable, and the observed do not love it.

The gap Exposed fills is personal, not academic. It is the quantified self, applied to the one input stream everyone receives and nobody logs. Not “advertising is bad,” which is an opinion, but “you saw this brand ninety-one times this month,” which is a fact, your fact, sitting in your own browser.

Key Features

A counter, not a blocker. Exposed changes nothing about the page. It observes, tallies, and stays out of the way, which also means it can coexist with a blocker and count whatever leaks through.
Brand-level attribution, honestly approximate. Where the advertiser is identifiable, from ad domains, disclosure labels, landing URLs, the tally is per brand; where it is not, it counts anonymously by category. The dashboard says “estimate” and means it, because the alternative is pretending.
Local, forever, structurally. The entire point collapses if the exposure ledger travels. Everything is computed and stored on-device; no account, no server, no telemetry. The report is yours to export, screenshot, or delete, and the code is open so that claim is checkable.
The weekly diet report. One glanceable page: volume over time, top brands, categories, the sites that serve you the most, your personal prime time. Designed like a nutrition label, because that is what it is.
Shareable stats, private data. A one-tap card, “1,437 ads this week, top brand for the 6th week running: same soda”, for the people who want to show the world, without a single underlying datum leaving the machine.
Research mode, strictly opt-in. For the journalists and researchers who need aggregate exposure data, a user can choose to donate anonymized tallies. Off by default, loudly consensual, and never required for any feature.

The honest tensions

Detection is a guerrilla war and the extension is the weaker side. Ads are deliberately obfuscated, served from rotating domains, rendered in closed iframes. There is one structural consolation. Manifest V3 hobbled the blockers far more than it hobbles a counter. Observing survives where blocking was maimed, which makes a counter the more future-proof species of extension. The elegant trick is to reuse the enemy of ads as a dictionary. The same public filter lists that blockers use to remove ads, EasyList and its siblings, work just as well to recognize them. Counts will still be undercounts, some attributions will fail, and the product must wear that honestly. Exposed offers an estimate of a number nobody else will give you at all.

The second tension is the observer effect, and I have made peace with it. Show someone their ad diet and many will install a blocker the same day, at which point Exposed’s numbers drop toward zero. The tool’s success shrinks its own dataset. Fine. Exposed is a gateway drug to media literacy, and if its endgame is its own irrelevance on a given machine, it has done what a nutrition label is for.

And the third is hostility. Platforms have already shown, with legal letters, what they think of ad observation at scale. Exposed’s defenses are structural. It observes only the user’s own screen, for the user only, locally, which is about as defensible as observation gets. But an idea in this territory should know whose lawn it is standing on.

Monetization

Almost none, and on purpose. This is a civic instrument in the family of my Obscurantism Clock and Breachline, tools for seeing slow, ambient things. Free and open source as the default, a pay-what-you-want supporter model, and perhaps a modest paid tier for the research-mode tooling and long-horizon exports. A tool whose entire promise is “your data never leaves” has almost no business model compatible with that promise, and pretending otherwise would be the first crack in it.

Next Steps: How I’d Build It

The MVP is one browser, one filter-list-based detector, one local database, one weekly report page. No brand attribution yet, just volume, sites, and time, because “1,400 a week” is already the headline that has never been shown to anyone. Then the attribution layer, the diet report design, and the shareable card, which is the only growth engine a no-server product gets.

The real work is credibility: publish the methodology, publish the code, undercount rather than impress, and let the number be checkable. The first journalist who runs it for a month and writes “I counted” is the launch.

Closing

Exposed is the missing ledger of the attention economy, kept by the only party who never had one, you. It does not argue with advertising. It just counts, locally, and lets the number do what numbers do.

If you build this, tell me your weekly total. I have been guessing at mine for years, and I suspect the truth is worse than the guess, which is, come to think of it, exactly why this tool should exist.

ThreatShield: Building a Safer Inbox

Mon, 08 Apr 2019 00:00:00 GMT

A few years ago, when I was working at a major French web company, we occasionally received some surprisingly elaborate scam emails. Hackers had breached one of our suppliers, and their emails looked almost identical to legitimate invoices. While I personally didn’t fall for the trap, some of my colleagues weren’t so lucky. A tool to quickly verify those emails could have saved us hours of stress, and, in some cases, real financial loss.

Fast forward to my personal life. My siblings and I have become the go-to “IT support” for our parents, a situation I’m sure many of you can relate to. Anytime a suspicious email pops up in their inbox, they forward it to us, hoping we can determine whether it’s safe to open or a phishing scam in disguise. Sometimes it’s easy to spot (obvious grammar mistakes, sketchy sender addresses), but other times, it’s hard to tell.

And even tech-savvy people occasionally hesitate when an email seems just a little too perfect. You know, the ones where you stare at the sender address, hoping for a clear sign it’s fake.

Tools do exist now that get close. Bitdefender Scamio and Norton Genie will both tell you whether a message looks like a scam, for free, without installing a suite. My parents have used neither, and I do not think they ever will, because both ask you to open a website you first have to find and trust, then paste the message into a chat box. What my parents actually do, every single time, is forward the email to one of their children.

That habit is the product. Not another place to go, but an address to forward to, the way they already behave, with the answer coming back in the same inbox they were already looking at. The gap is not detection. The gap is the gesture.

Market & Opportunities

The need is real and it is not shrinking. Phishing sits near the top of every national cybercrime report, year after year, and the losses keep climbing. Small businesses and individuals carry more of it than anyone, because they are the ones without a security team. Beyond financial loss, there’s the emotional fallout: frustration, mistrust, and even a reluctance to use email altogether.

Two groups pay for this, for opposite reasons. Families pay to stop being a burden: the parent who forwards every doubtful message to a child and the child who answers between meetings. Small businesses pay because a single convincing invoice costs more than a year of subscription, and they have nobody whose job it is to look. Everything else (enterprises, other languages, other continents) is a later problem that only exists if the first two work.

Key Features

Forward and forget. Send the suspicious message to a dedicated address (e.g. check@threatshield.example) and the answer comes back in the same inbox, in seconds. No plugin, no inbox permission, nothing to install.
A score anyone can read. Green, amber, red, then the reason in one sentence. The detail underneath (sender reputation, link analysis, formatting anomalies) is there for whoever wants it and invisible to whoever does not.
Shared memory. A message already reported by someone else is recognized instantly. It is a neighborhood watch for inboxes, and it gets better the more people use it.
What is kept, and what is not. The body of a forwarded email is thrown away once parsed. What stays is the scam signature: sender patterns, link structure, the shape of the fraud. That distinction is the entire privacy promise, and it is worth stating plainly rather than hiding it behind a “we store nothing” badge that would not be true.
Your own history. A simple record of what you flagged and what it was, so the person who keeps receiving parcel-delivery scams eventually learns to spot them without asking.

Later, if the thing works: native Gmail and Outlook add-ons, SMS, and a monthly Threat Barometer built from the reports.

Monetization

Two paths, and they are enough. Freemium for individuals: a handful of checks a month for free, because the person who needs this most is the one who will never pay for security software, and a small subscription for unlimited checks and the personal history. A flat plan for small businesses, sold per company rather than per seat, because the buyer is an office of six people who just want the invoice checked before it is paid.

What I would not do is sell the threat database. It was in my first notes, as a subscription for security firms and research institutions, and it contradicts the promise made two sections above. You cannot tell someone their email is thrown away and then sell what you learned from it. Pick one.

Abuse: what an open address invites

An address that anyone can forward anything to is, by construction, an address that will receive things it should not. People will forward emails containing someone else’s medical results, a colleague’s salary, a scan of a passport. Not maliciously, just because the message happened to be suspicious and it also happened to contain that.

So the deletion of the body is not a feature, it is the load-bearing wall. Parse, extract the signature, drop the rest, and never build the button that would let an administrator read a stored message, because a button that exists is a button that gets used. Rate limits on top, per sender, to stop anyone from using the service as a free bulk scanner for their own spam campaign.

Next Steps: How I’d Build It

The simplest starting point is the one already in the pitch, a forwarding address. Users send a suspicious message to a dedicated address (say, check@threatshield.example), a parser pulls the headers, links, and sender reputation, and a reply comes back with a score. No inbox permissions, no plugin, nothing to install. That is the MVP, and it is buildable with tools that exist today.

The blockers sit further out. A risk score in seconds is only as good as the analysis engine behind it, and getting false positives low enough to be trusted is the real work, not the plumbing. Sender reputation and URL scanning would likely lean on existing threat-intelligence feeds rather than a database built from scratch.

The deeper Gmail and Outlook integration is not a weekend add-on. It means building against Google’s and Microsoft’s add-on APIs and passing their review, which is doable but slow. I would ship the forwarding version first and treat native integration as a later milestone, not a launch feature.

A Safer, Smarter Inbox

I keep coming back to the same picture: my parents forwarding me an email, waiting to hear whether it is safe.

A tool that answers that question in seconds, for anyone, would take a real weight off a lot of people, from older relatives to busy professionals who glance at their phone between meetings.

I do not think this reinvents cybersecurity. It just puts one useful check within reach of people who do not have an IT department, and that is enough of a reason to build it.

The concept is still rough, and plenty of the hard parts (accurate scoring, keeping false positives low, provider integrations) are unsolved. But the demand is obvious, and none of the blockers look fatal.

If you are a developer, an entrepreneur, or someone who has also become their family’s IT support, and you decide to build this, I would be glad to help think it through or review it. If you build this, tell me. I would love to follow it here.

Chronopiss: The Playful Bathroom Tracker

Sat, 26 Jan 2019 00:00:00 GMT

Have you ever been in the middle of a bathroom break, secretly timing yourself, wondering if today you’ve set a new “personal best”? As a guy, I’ve had this thought more than once. Some sessions clock in at a surprising 45 seconds or more, and I’m left wondering, “Was that my longest yet?” But there’s no easy way to track these things, until now.

The device is a small, colorful box that sits above the toilet. A sensor starts the timer when you walk in and stops it when you leave, so the tally arrives without you touching anything. The machine is that simple.

The second function, and the one I would actually keep, is a beep after ten minutes. Not a health feature, a nudge. It is far too easy to lose a quarter of an hour to a feed while sitting somewhere you meant to be for two minutes, and a small noise in the room is a better interruption than a notification on the phone that is causing the problem.

And for those of us who like a stupid comparison, there is a leaderboard to share with friends who are up for it.

Add it up and you get a number, a nudge, and a bit of friendly competition, in a box that is as colorful as it is unnecessary.

Market & Opportunities

People already count their steps, their sleep and their coffee. Counting this is the same reflex pointed at the one room nobody instruments, which is why it is funny.

Who buys it? Someone who already owns a silly gadget, and someone buying a gift for the friend who is famously never out of the bathroom. This is an impulse purchase and a Secret Santa object. I will not pretend there is a wellness market waiting for it.

There is one non-joke behind the ten-minute beep, and it turned up after I first wrote this. A 2025 study in PLOS One surveyed 125 adults before a colonoscopy and found that people who used a smartphone on the toilet were far more likely to stay more than five minutes (37% against 7%), and had a 46% higher rate of hemorrhoids. The authors are careful and so am I: it is a small, cross-sectional study on people over 45, it relies on what they said about themselves, and it cannot show that the phone caused anything.

So no, this is not a medical device and I am not selling one. But it does mean the beep is aimed at something real, and that is a better reason for a joke gadget to exist than a market trend.

Key Features

It starts on its own. A presence sensor begins the count when you walk in and stops it when you leave. No button to press, because a tracker you have to remember to start is a tracker nobody uses.
Two big buttons anyway. For the people who would rather run it by hand, and because two large colorful buttons on top are half the charm of the object.
A clock the rest of the time. Idle, the display is just a bathroom clock. It has to earn its place on the wall during the twenty-three hours it is not counting anything.
The ten-minute beep. Optional, configurable, off if you want. A noise in the room beats a notification on the device that is keeping you there.
A leaderboard you have to opt into. Invite-only groups with friends or family, nothing public, nothing shared by default. The stat is funny among people who chose to see it and creepy everywhere else.

A companion app can come later for trends over weeks. What it should not collect is anything that turns a joke into a medical record: no hydration logs, no symptom tracking, no “restroom routine” data. The device counts seconds, and it promises nothing more.

Monetization

Sell the box. That is the business, and for a gift object at an impulse price it is very nearly the whole business. A two-pack for households with two bathrooms is the only bundle I would bother printing.

Everything else is optional and small. The app can sell a few sounds and themes, because people will pay a euro for a fanfare. I would keep the trends and the leaderboard free, since they are the only reason to open the app twice.

What I would refuse is the wellness partnership. The moment this thing recommends a hydration supplement or a bladder-health brand, it stops being a joke about a stopwatch and starts being a health product that has to answer for its claims. There is no version of that I would want to defend, and the margin on a sachet is not worth it.

Next Steps: How I’d Build It

Do not manufacture anything. The first version is a presence sensor, a microcontroller and a 3D-printed shell, taped above one toilet, in one flat, for a month. It costs almost nothing and it answers the only question that matters: do you still look at the number after two weeks, or did you laugh twice and stop? A gag object lives or dies on that, and no amount of industrial design saves it if the answer is no.

If you still look, the walls are the ordinary hardware ones, and they are worse than they sound for an object this cheap. Small-run injection molds cost more than the whole first batch will earn. A bathroom is a humid room, so the enclosure has to be sealed and the thing has to run for months on a battery or accept living next to a socket that may not exist. And an electronic device sold to consumers needs CE or FCC marking, which is a real line in the budget of a joke.

The sensor itself is the fiddly part. Telling a real visit from someone walking past to grab a towel is the difference between a funny number and a broken one. Start with the crudest thing that works, and only add cleverness once the crude version is wrong often enough to annoy you.

Then price it honestly. If the bill of materials plus certification means the box has to sell for the price of a decent dinner, it is no longer an impulse buy and the idea quietly dies. Better to find that out with a taped-up prototype than after the molds.

Bonus: the one misuse that matters

A sensor on a bathroom wall is a surveillance device wearing a joke.

The risk is not theoretical. The same box that tells you your own number tells a parent how long a teenager was in there, or a manager how long an employee was gone.

So the design has to refuse that job. No named profiles on the device, no history a second person can pull up, no export of someone else’s sessions. A shared household box shows the last time and nothing more. If the product cannot be built without a screen that says who was in the bathroom and for how long, it should not be built, and I would rather say that here than discover it in a review.

The rest is ordinary: encrypt what goes to the app, let people delete everything in one tap, and keep the leaderboard opt-in. Bathroom habits are personal even when the whole thing is a gag.

Closing

Chronopiss is a joke I’d use for real. It’s a silly little device that turns an invisible daily ritual into a number, nudges you to put the phone down, and keeps a very stupid leaderboard for friends who are up for it. That’s the pitch, and it doesn’t need to be more than that.

It’s just a concept for now. If you have the hardware chops for the sensor-and-timer version, or the app to go with it, take it and run. You owe me nothing, but if you make a real Chronopiss, send it my way and I’ll happily review it (and time myself with it).

LastBus: A Ride When the Bus Never Comes

Mon, 03 Sep 2018 00:00:00 GMT

This idea is one of my oldest, and it was invented uphill. In my lycée years, my ride home was a rural bus line with a timetable that treated my existence as optional. Some days there was simply no bus. Other days I would round the corner just in time to watch it pull away, and both cases ended the same way: several kilometers on foot, climbing, on the shoulder of a busy road, cars brushing past a teenager with a schoolbag.

I had a lot of time to think on those climbs, and what I thought about, mostly, was the parade of half-empty cars passing me, every single one going exactly where I was going.

That is the problem in one image. Rural mobility does not suffer from a shortage of vehicles, which are right there, streaming past the bus stop. It suffers from a shortage of coordination, plus the small social wall that keeps a soaked student from sticking out a thumb and a kind driver from stopping unprompted.

LastBus is the coordination layer, built on the bus stop, the one piece of infrastructure every countryside already has. When the timetable fails (no bus for two hours, a bus just missed, a line cancelled by snow), the stops on that stretch switch to pickup mode. Riders check in at a stop, and drivers who declared “I pass these stops anyway” get pinged. A small micro-payment covers costs. The app matches, the car pulls in where buses pull in, and nobody walks the hill.

The graveyard, and the three things it left unbuilt

Venture-backed microtransit is a field of headstones (Bridj’s American service, Chariot, Citymapper’s ride experiments, all gone), and the survivors like Via live by selling software to transit agencies, not rides to riders. France already runs transport à la demande in many regions, and, closest of all, Rézo Pouce organizes rural hitchhiking with designated roadside points. The space is not empty. So the question is what my uphill walks add, and I count three things.

First, the trigger. Everything existing is either scheduled in advance or always-on. Nobody exploits the sharpest signal in rural mobility, the real-time gap between the timetable and reality. “The 17:42 just passed” or “nothing until 19:15” is the moment demand exists, concentrated at known points, and it is computable from open transit data. LastBus is the bus network’s own exception handler, not an alternative to the bus.

Second, the stop itself. A bus stop is a Schelling point with a roof: known, mapped, lit, expected. Pickup at the stop removes the two hardest problems of ride-matching (where to meet, and whether stopping there is safe and legal), because the infrastructure already answers both. My old note said it plainly: show drivers the stops along the route they already drive, so riders know which stops are live.

Third, the micro-payment. Organized hitchhiking is free, and free carries a social debt that keeps half the potential riders walking. A euro or two, framed as shared costs, converts charity into a transaction light enough to accept and small enough to never feel like a fare.

The money is not revenue; it is social lubricant.

Key Features

Pickup mode, driven by real timetables. Stops light up only when the schedule fails or gaps. The app watches the network’s own data; scarcity of the mode is what keeps drivers responsive when it matters.
Declared routines, not dispatch. Drivers register the stretches they drive anyway, commutes, school runs, market days. A ping is an invitation, never a job. This is a neighbor slowing down, not a gig.
Cost-share by design. Payments are capped at genuine cost sharing, the legal carpooling frame, with karma and community recognition carrying the rest of the reward.
Verified both ways. Identity checks for drivers and riders, plate confirmation at pickup, live trip sharing with a contact. The stop is public; the car must be accountable.
Comfort filters, handled honestly. Riders can express preferences, including riding with a driver of the same gender, a real and legitimate safety instinct, especially on empty roads at night. The legal footing of strict gender matching varies by country, so the design treats it as a preference the matcher honors when it can, inside whatever the local law allows, rather than a promise.
The minor problem, faced. My founding user was fifteen. An app that puts minors in strangers’ cars is the hardest version of this product, and the answer is staged: adults first; minors only later, through family-linked accounts and a vetted circle of community drivers, the digital version of the parent network that already informally ferries kids around every rural town. The teenager who invented LastBus could not have legally used version one, and the article owes him that admission.

The honest core: my favorite feature is the illegal one

My original notes included a rush mode, an auction where a stranded rider posts a price and drivers bid. I still love it, and it cannot exist, at least not between private individuals. The line in most countries, France included, is sharp. Share costs and you are carpooling. Make a profit and you are an unlicensed taxi. The auction crosses the line the moment it works.

So the design splits in two. Between individuals, LastBus stays strictly inside the cost-sharing exemption (the BlaBlaCar precedent), small, capped, boring, legal. And the rush auction survives in the only place it can: the business-to-government version, where the transit authority, not the rider, pays drivers a bounty to absorb a cancelled line or a snow day. The authority already pays for service it fails to deliver, and LastBus just lets it buy the exception from the people already driving there. That, not rider fees, is also where the revenue lives. Sell the exception handler to the operators whose timetables create the exceptions.

Next Steps: How I’d Build It

One valley, one line, one winter. Partner with the local transit authority from day one, not for permission but for data and legitimacy. They know where their timetable fails, and a pilot branded with the network’s own colors turns “get in a stranger’s car” into “the bus network’s backup plan,” which is the entire trust battle won in one move. Seed the driver side through the commuters who already drive that road daily, and measure one number: walks avoided.

The MVP needs embarrassingly little technology (open transit data, a check-in, a ping, a payment cap), because the product is the agreement between a timetable, a bus stop, and the half-empty cars already passing it, not the app.

Closing

LastBus finally gives the parade of cars that streamed past my soaked teenage self a reason to stop, at the stop, when the bus does not come. It took me years to see that the infrastructure was already built and only the coordination was missing.

If you build this, tell me, and start with a hill. Somewhere on it, right now, there is a kid with a schoolbag doing the walk I did, counting the cars going exactly where they are going.

HomeGuard Drone: A sentinel drone, when the law allows

Tue, 12 Jun 2018 00:00:00 GMT

I do not have a burglary story to open with, thankfully. What I have is a colder observation. Home security is one of those quiet, universal worries, and for the first time the technology to do something different about it exists. Ten years ago this concept was science fiction. Today the pieces, autonomous flight, thermal imaging, smart-home integration, are all real, even if the law has not caught up (more on that below). What I wanted to sketch is what those pieces add up to. Not another lens bolted to a wall, but something that can actually go and look.

A dock on the roof. A small aircraft that sits there, charges, and goes to look when a sensor asks it to, or when you tap a phone because you heard something. Thermal and night vision for the dark. A follow if there is a person where there should not be one. A two-way speaker if you want to say “I see you” from another country.

Every piece already exists.

What stands between this sketch and a product on a shelf is regulation, not engineering.

The reality check further down takes that seriously. The shift is real, from seeing what happened after the fact to going to look, on a timeline set as much by regulators as by engineers.

Market & Opportunities

People already buy cameras and alarms. Those sit still. The gap is a property too big, too dark, or too far from a road for a doorbell to help: a house with a garden, a driveway you cannot see from bed, a place you leave empty for a week.

I will not hang a home-security market figure on that, and I will not sell peace of mind as a feeling. Who might pay, when the law allows: owners of larger lots, and later a security firm that wants a patrol it does not have to staff. Who should not: anyone who wants to watch the street. That is a different product, and a worse one.

Smart-home hooks (a sensor, a phone alert) are how this joins a house that already has cameras. They are not a reason to skip the legal section.

Key Features

A patrol you schedule, or a launch you trigger. Routine passes over the parts of the lot you care about, and an instant launch from the phone when something sounds wrong.
Eyes that work in the dark. Night vision and thermal, so a person and a fox are not the same blob.
A follow that stays on your side of the fence. Track a path, record it, alert you. The fence is not optional; see the beacons below.
A speaker on the aircraft. Two-way audio from the phone, if you want a voice in the garden. Off until you press it.
A dock that is boring. Charge, sit, launch. Solar is a nice extra for a remote lot. Weatherproofing is required. Neither is the idea.

GPS is the obvious way to draw a boundary, and it is also the wrong tool this close to a house. The signal bounces off walls and drops out under the eaves, which is exactly where the drone flies.

The fix is physical. Small beacons fixed along the roofline and the edges of the property, using short-range positioning such as ultra-wideband, give the drone a centimeter-accurate sense of where the property ends. The geo-fence stops being a hopeful GPS estimate and becomes a measured wall the aircraft is programmed never to cross, holding up precisely in the low, cluttered airspace next to a building where satellite positioning is least reliable.

Monetization

Sell the aircraft and the dock. A subscription for updates, encrypted cloud clips, and support. A security company can bundle it later. I would not hide the aircraft behind a software lock. If the subscription dies, the thing should still launch and still refuse to cross the beacons.

Misuse: the neighbor’s side of the fence

Geo-fencing your own property is necessary but not sufficient. A drone at altitude with thermal and night vision inevitably sees over the fence. A responsible version has to treat the neighbors’ privacy as seriously as the owner’s, no recording beyond the property boundary, cameras that cannot capture a neighbor’s window, and a clear answer to the reasonable fear that the thing hovering next door is filming you.

Users can mark privacy zones on their own lot (a window, a terrace) where the camera will not record. Containment itself rests on the beacon fence described above, not on a GPS estimate.

Footage should be encrypted, and a zero-knowledge store is the version I would trust. “Authorized users only” is a slogan until the architecture says who cannot see it.

A regulatory reality check

This has to be said plainly, because it is the difference between a concept and a lawsuit. The problem is not flying a drone over your own garden at a legal altitude. Plenty of people do that every weekend. The problem is that nobody is holding the controls. In the US, FAA rules are built around a pilot keeping the aircraft within visual line of sight, and they restrict night flight and flight over people. A drone that launches itself and patrols while you sleep, or while you are on vacation, is what the rules do not allow for consumers today. Europe (EASA, and DGAC in France) and Canada (Transport Canada) are built on the same principle. Autonomous, pilotless patrol falls into authorization categories that ordinary homeowners cannot realistically obtain yet.

So this is a forward-looking concept, not a product you could ship next quarter. The realistic near-term version stays inside the law: piloted on demand rather than fully autonomous, strictly within your own property line, and honest that “autonomous sentinel” is where the category is heading once regulation catches up, not where it is now. It is telling that the closest thing to a product, Ring’s indoor flying camera (invitation-only since 2021, never widely released), stays indoors, where aviation rules do not apply.

There is a middle path worth exploring: put the drone on a leash. A tether, a thin cable that powers the aircraft and physically limits how high and how far it can go, changes the picture. It hard-enforces the altitude cap and the property boundary (no flyaway, nothing drifting over the neighbor’s garden), and several jurisdictions already treat captive aircraft more favorably than free-flying ones. Tethered systems are sold to fire departments today precisely because the cable removes much of the certification burden. A tethered HomeGuard watching the driveway and the back door trades coverage for legality, and it might well be the version you could actually ship first.

Next Steps: How I’d Build It

Build the legal half first. A piloted, on-demand aircraft that will not cross a beacon fence, on a tether if that is what a lawyer in your country will sign. One garden, one dock, one phone trigger. No patrol while you sleep until the rules say you may.

The order matters. The dock and the beacon fence are ordinary engineering and you can build them in a garden. The autonomy is the part that waits. Do the boring half first and you have something a lawyer can look at; do the flying half first and you have a demo you cannot sell.

Pick the country before the hardware. The rules that decide whether this is a product or a hobby are national, and a tether that satisfies one regulator is not automatically enough for the next. Read your own aviation authority first, design the aircraft second. That order is the whole business plan.

If that version is useful (you heard a noise, you looked, you went back to bed) you have a product that can wait for autonomy. If it is a worse flashlight, the sketch was only a sketch.

Closing

HomeGuard is a thing that goes and looks, when the law allows, and it has a tethered cousin that might be legal sooner. The technology is not the mystery. The boring half is.

If you take a run at this, even just the piloted, stay-inside-the-fence version, tell me. This is one of those ideas where the builder who solves the boring legal half wins the whole category.

Liminal Archive: A mind palace you can walk through

Thu, 11 Jan 2018 00:00:00 GMT

I am a very organized, very visual person. I remember things by where they sit, not by what they are called. This idea has felt obvious to me for years. People have sworn by the mind-palace method for centuries, memorizing by walking through an imagined building, and Liminal Archive is essentially that, except built, and the way I would want it: your notes as a space you walk through, not a list you scroll.

Plenty of tools already organize notes. None of them feel like a place.

The picture is simple. Rooms, drawers, a title floating above each one: Personal Projects, Work, Novel Ideas, whatever you actually use. You put a note in a spot. You find it again by walking back to that spot. On a phone it is a map you pan. In a headset it is a room you enter. Same palace, two doors.

I have tried to fake this with folders, with tags, with a second brain that was really a third inbox. The tools were fine. My memory was not using them.

I can still picture where a page sat on a desk, and I cannot tell you what I named the file.

That is not a personality quirk I am proud of. It is the constraint the product has to respect, or it is just another outline with nicer lighting.

Design is not a coat of paint here. If the space is ugly or noisy, the metaphor dies, and you are left with a worse Notion. I care about the click of a drawer and the quiet of the room. I do not need a famous industrial designer on the poster for that to be true.

Market & Opportunities

The second-brain market is crowded: Notion, Obsidian, Roam, Heptabase and the rest already own “organize your thoughts.” Beating them on features is a losing game. The only wedge is the spatial one, and even that only works if people can try it without buying a headset.

If headsets spread the way phones did, and that is a real if, a serious spatial notes tool will have somewhere to live. Until then the buyers are people like me: visual, a bit stubborn about folders, already bouncing between a phone and a laptop, curious about VR but not ready to make it the price of entry.

Writers with a novel in pieces, researchers with a wall of PDFs, anyone who already thinks in rooms rather than in databases. Not the person who wants a faster outline. They already have one, and they are happy.

A shared palace, two people in the same room of notes, is a later chapter, not a launch promise. Useful for a small team. Easy to turn into a meeting product nobody asked for. I would wait until one person already lives in their own palace and wants to hand someone else a key. Starting with collaboration is how you build a worse Miro.

I would not sell this as wellness or as a treatment for a scattered mind. It is furniture for memory. That is enough. If a reviewer writes “mindfulness” in the App Store copy, rewrite the listing before you ship the next build.

Key Features

A palace you arrange. Drawers, rooms, clusters. Folders and tags if you want them, but the primary act is putting a thing in a place.
The same palace on every screen. Phone, laptop, headset. Sync is boring and mandatory. If the room on your phone is not the room in the headset, the metaphor is a lie.
Quiet suggestions, not a copilot. Optional links between notes you already wrote. A character trait from March next to a plot twist from October, if you ask. Off until you turn it on.
More than text. Images, PDFs, a voice note from a meeting, sitting in the same drawer as the page they belong to.
VR as the top end, not the door. Gestures and depth for the people who have a headset. Everyone else gets the 2.5D canvas first.

Search still exists. You can type a word and jump to the drawer. The palace is not a rejection of find-in-page. It is what you use when you do not remember the word, only the corner. If search is missing, visual people will still leave. If search is the only way in, the palace was decoration.

Monetization

Free for one small palace on the web. A subscription when the palace grows, when you want history, or when you want the headset door. Team seating later, if the shared room earns it. No marketplace of “palace themes” on day one. Taste is the product, and a theme store would sell it off.

A fair price is the price of a notes app people already pay, not the price of a headset. If the VR door costs extra, say so on the free canvas, and let people refuse it forever. The product should still work for someone who will never put a screen on their face.

Next Steps: build the flat version first

The wedge is spatial, but it only lands if people can feel it before they buy hardware. If the pitch is “you will want to buy a VR headset for this,” almost nobody will try it. That is the wrong order.

So the MVP is not VR. It is a web app with a 2.5D spatial canvas, drawers and rooms you pan and zoom through on an ordinary screen, that already feels different from a list on day one. VR becomes the aspirational top end for the few who have a headset, not the price of entry. Ship the flat version and find out whether the spatial metaphor actually helps people.

The first test I would run is selfish and small. I would move my own working notes into the canvas for two weeks and see whether I stop opening the old app. If I sneak back to a list because the palace is pretty and slow, the idea is a sketch, not a product. If I start putting new notes in a corner without thinking, then, and only then, invite five other visual people in. Do not invite a headset reviewer first. They will review the headset.

Import from a markdown folder is enough for that test. A pretty empty palace teaches you nothing. Fill it with the mess you already have, then see whether the mess gets easier to walk through.

Closing

Liminal Archive is memory with an address. For now it is a palace with no bricks. I would rather walk a slightly empty room that is mine than decorate a template that looks like everyone else’s second brain. If the first version is only a canvas you can pan, with drawers that stay where you put them, that is already more than I have today.

If you build this, tell me, and give me a headset demo. I want to walk through someone else’s memory palace at least once in my life. Take the idea wherever you want, you owe me nothing.

BeTheHero: Name the play that flipped the match

Thu, 10 Nov 2016 00:00:00 GMT

Some of my best gaming memories are rounds of Battlefield 3 where the match looked lost and one decision changed everything: a long flank around the side, a strong position taken, the enemy caught off guard, and suddenly the whole team is pushing an objective they had given up on. I have been that player a few times, in Battlefield Bad Company 2 and later in Battlefield 3 and 4, and fifteen deaths were a small price to pay to turn the battle around.

Tactically, those are the moments that flip a match. In a game built around a team objective, they deserve to be shown to everyone when the round ends.

Yet too often these players go unrecognized. Most multiplayer scoreboards still worship the kill count. They skip the person who died fifteen times to make the push possible.

The idea is a system that can read a match, find the turning point, name the player who made it, and show that clip when the round ends. Not a new K/D. A way to see the play that actually flipped the map.

I still remember the feel of that flank more clearly than any scoreboard. The deaths were public. The reason was not. That is the waste I want fixed. The useful play is already in the demo. Nobody names it.

I would rather games turn us back into team players than solo players.

Market & Opportunities

Competitive multiplayer is huge and still growing. I will not pick a billion-dollar figure from a slide. The audience I care about is the one I was in: people who already play objective modes and already know the scoreboard is lying.

Who would use it: players who want the flank named, coaches who want the same clip without scrubbing, and later a broadcast that is tired of another ace montage. Game makers could steal the idea and ship it inside the title. A third-party tool only lives where the match data is actually public.

That last sentence is the business. If Valve or Riot already expose the round, you can be a clip site tomorrow. If they do not, you are a pitch deck waiting for a meeting. I would rather be the clip site. Publishers can copy it later. That is a compliment, not a failure.

I would not sell this as an esports-betting layer. That is a different product with a different conscience.

Key Features

Find the turning point. A retake against the odds, a save that flips the economy, a flank that opens an objective. The model can be clever later. The definition has to be readable by a human first.
Score the play in context. Map, mode, scoreline, whether you were last alive. A kill in a won round is not a hero play. A death that made the plant possible might be.
Cut the clip. Enough time before and after that a stranger understands it. A name on the screen. Shareable, not a two-minute recap.
Remember who does this often. A board of people who keep making the unselfish play, inside one game, not a fantasy cross-title MVP.
A page where you jump to the moment. Filter by match or player, skip to the turn, watch it again. Broadcaster overlays can wait until the clip is good.

The CNN, the reinforcement loop, and the publisher API can live in the build notes. They are how you might do it. They are not why anyone would open the site.

A coach view that ranks players against each other in public is how this becomes the scoreboard I am trying to leave. Keep that view inside the team, or do not build it. Recognition in the round is the gift. A season-long ranking of “who sacrificed more” is a new way to farm the wrong thing.

Monetization

Free for the clip after your own match. A small subscription for history, the explorer, and a coach view that stays on the team’s data, not a public shaming board.

If a publisher wants the feature inside the game, that is a license, not a consumer upsell. I would not fund this with betting widgets.

Next Steps: start where the data is open

The blocker is that the games behind these memories, the Battlefield series, do not expose the live match data this needs. Most big FPS titles keep it behind closed doors. So the first version should not chase them. It should start with a game that already offers real match data through official channels, CS2 and Dota 2 via Valve’s replay files and game-state feed, or League and Valorant via Riot’s match APIs, where kills, objectives, positions, and round state are actually retrievable.

That makes a concrete MVP: ingest completed CS2 matches, score the round-turning moments (a retake against the odds, a clutch, a save that flips the economy), and auto-cut a short highlight with the player named. No publisher deal required to prove the core idea works. Expanding to closed-API games later becomes a business-development problem, not a technical unknown, and that order matters.

The first test I would run is one week of my own CS2 games, or a friend’s. If the tool names the player I would have pointed at in voice chat, you have a pulse. If it names the top fragger every time, you have rebuilt the scoreboard.

Keep the first model stupid enough to explain. A rule you can say out loud (“plant while down a player,” “last alive on a retake”) beats a network you cannot defend in a Discord. You can get fancy after the stupid version is already right more often than the kill feed.

Then decide what happens when it is wrong, because it will be. Naming the wrong player is worse than naming nobody. It hands someone else’s round to a stranger, in public, at the moment the whole server is looking. Let a team correct a call, keep the correction visible, and do not sell the thing as a referee. Nine right out of ten is fine for a highlight reel. It is not fine when the tenth one is the play someone actually made.

Closing

This is the part that makes me dream: a match where the player who threw themselves at the objective, and died fifteen times doing it, finally sees their name in lights instead of at the bottom of a K/D table.

I would rather be wrong about a CS2 clutch than vague about a neural net. The Battlefield memories are why I want this. They are not where you start. Start where Valve or Riot will actually hand you the round.

If that clip makes a teammate say “yes, that was the play,” you have the product. If it does not, no overlay will save it. I would watch that clip. The scoreboard can keep the kills. I want the flank.

I do not know if it is buildable at scale, and the API access is a real wall. But the itch is real, and I would love to see someone take a run at it. If you build this, tell me. I will be first in line to test it and write about what you made.

FanVault: Official merch for the artists you already stream

Thu, 28 Jul 2016 00:00:00 GMT

I am an avid music fan, and the hassle I want to kill is specific: I finish an album, I want the shirt, I open the artist profile, I get a vague store link or nothing, I search the name plus “merch,” I land on a shop that may or may not be theirs. Doing that for five artists in a row is how an evening disappears.

Spotify is great at playing music. It was never built to sell me a hoodie, and you feel it the moment you try to browse merch for more than one artist in a row.

FanVault is the page I wanted instead. You connect a Spotify account, see your top artists, and get their official merch in one place. That evening, collapsed. No more hopping from a label shop to a Shopify theme to a dead Bandcamp button. To be clear, this is a fan-made concept with no affiliation to Spotify. It would build on their public API, if its terms allow it.

Official merch first, from the artist or the store they point to. A trusted long tail (Amazon, a couple of big merch platforms) only where the official shelf is empty. The job is a reliable selection, not every bootleg hoodie on the internet.

I would keep the interface familiar to a Spotify user on purpose. The less you have to learn, the faster you get to the shirt.

Market & Opportunities

Music merch is a real business, especially after streaming ate most of the album sale. I will not hang a global licensing figure or a Spotify user count on that sentence. The demand I trust is the one I have. I already know who I listen to, and I still cannot buy their merch without opening five tabs.

Who shows up first: people who already live in Spotify and already own one tour shirt. Who pays: nobody, at first. The site takes a cut when a click becomes a sale. Artists and stores get a visitor they did not have to design a page for.

The gap is not “fans do not want merch.” The gap is that the merch is scattered and the streaming app will not be the store. A third-party catalog that is honest about being partial can still be useful. A catalog that pretends to be complete will lie in the first week.

This is not a new Spotify, and it is not a concert app. Tickets, forums, and fan art can live somewhere else. If the first version cannot show a shirt for the artist I played this morning, the rest is decoration.

Key Features

Spotify for the taste, not for the stock. Connect an account, pull top artists and recent listening, build a merch home from that list. Nothing else needs to come from Spotify.
A catalog that grows store by store. Shopify here, Bandcamp there, a label shop next month. Coverage starts partial and says so on the artist page.
Browse without the feed becoming a mall. Genre, artist, “new this week.” Filters that find a vinyl or a shirt. Not a recommendation engine that has opinions about your identity.
Buy on the store that actually stocks it. Deep link, affiliate if the program exists, a clear price. In-app checkout for every shop on earth is a fantasy. Do not ship it as one.
Official first. If the item is not from the artist or a store they stand behind, it does not sit next to the official pile. Third-party results, if they exist, are labeled.

Exclusive drops and a peer-to-peer closet for old tour shirts can wait. Those are how this idea becomes a marketplace with a music skin. Get the official shelf right first.

Monetization

Affiliate commissions are the straightforward engine: a click, a sale, a cut. Rates will differ by store and you will not get a deal from every artist on day one.

Sponsored placements can exist if they are labeled and if they cannot bury the official item the user came for. I would not fund the homepage with display ads for concert bundles. That is how a merch finder becomes a banner farm.

Consignment and resale are a later chapter, and a legal one. Counterfeit lives there. I would not open that door until the official catalog is boringly trustworthy.

Next Steps: the part that is actually hard

Pulling your top artists from Spotify is the easy half. The Spotify API hands that over once a user connects their account. The merch is the hard half.

There is no single “artist merch” API.

Official stores are scattered across Shopify, Bandcamp, label shops, and bespoke sites, each with its own format, and Spotify itself exposes no merch feed. So “all their merch, automatically” is the promise and the wall at the same time.

The MVP does not try to cover everyone. Start with the merch sources that offer a clean feed or an affiliate API (a few big artist-store platforms, plus something like Amazon for the long tail), aggregate those well, and be upfront that coverage is partial. Grow it store by store. A thin, reliable catalog beats a “complete” one stitched from scrapers that break every week and may run against a store’s terms of service.

The first test I would run is ten artists I actually listen to. If eight of them have an empty shelf, the idea is a bookmark list. If five have a real shirt I can buy in two clicks, you have a pulse. Then add the next store, not the next social feature.

Read the Spotify terms before you write the connect button. Their developer policy forbids using Spotify data to build user profiles for advertising or marketing, and a merch page assembled from someone’s top artists lives close enough to that line that you want an answer, maybe even Spotify’s answer, before you build. Read the affiliate terms too before you promise Amazon as a backfill. A fan app that gets its API access pulled is not a merch app. It is a eulogy.

Bonus: counterfeit, and the rest of the ugly shelf

Counterfeit is the real one. Vet sellers, prefer official URLs, and take a listing down when a rights holder asks. I would not dress that up as blockchain authentication.

Copyright and trademarks are the same problem in a lawyer’s font. A report button that reaches a human is worth more than an automated scanner you do not have.

User reviews, if you add them, will attract the usual sludge. Guidelines and a queue. The catalog can live a long time without reviews. I would rather ship ten official stores and no comments than a comment section sitting on empty artist pages.

Closing

FanVault points your Spotify taste at official merch, with a catalog that admits it is incomplete. I have wanted this exact app for years. A thin shelf I can trust beats a mall that might be selling someone else’s tour.

If you build this, I want to hear about it, and I would be your first user. Take the idea, rename it, change whatever you like. You owe me nothing, just tell me how it goes.

BRBMail: A smarter out-of-office, one message per contact

Wed, 13 Jan 2016 00:00:00 GMT

As a freelancer, managing client communication when I take time off is a real headache. A generic out-of-office treats every contact the same, and writing a different note for each client is the job I never have time for before I leave.

I juggle several clients and projects at once, and I want to stay professional with each of them, even when I am away for a few days. BRBMail is the app I have wanted for that: one tailored reply per contact, not one canned line for the whole inbox.

The night before I leave is always the same: I open the vacation responder, I stare at it, I write something vague, and I know the client who is mid-project will read it as a shrug. The client who can wait will get the same shrug. That is the itch. Not “email is hard.”

One person needs a date and a backup contact. Another needs “I will not see this until Monday.” The inbox does not know the difference.

Market & Opportunities

Freelancers are the obvious first users. They are not the only ones. I have worked in web agencies and in larger companies with external communication, and the same gap shows up there. When someone goes on leave, the mailbox either overshares or says nothing useful.

Agencies and small PR teams live in that gap. A publicist out for a week still owes journalists a clean, specific line. A studio with clients in two time zones still owes a reply that does not pretend everyone is in the same afternoon. I will not hang an Upwork headcount on that sentence. The demand I trust is the one I feel every time I book a few days off. I need five different messages, and Gmail will let me write one.

The first to pay is a freelancer who leaves twice a year, then a three-person shop that wants the same rules for everyone. A company that already has an IT team and a policy PDF will not pay. They have an out-of-office. They do not have this problem.

Key Features

One message, one address. Assign a reply to a specific client. That is the product. Everything else is furniture.
A library of your own lines. Every message you write is saved, so the next trip starts from last time, not from a blank box.
Segments when one-to-one is too much. Clients, press, “everyone else.” Your best contacts get the specific note. The rest get a clean general one.
A start and an end. It turns on when you leave and off when you are back. Recurring weeks if your life actually looks like that.
Gmail first. Manage it next to the inbox you already use. Outlook can wait. A dashboard of every mailbox on earth is the vision, not the first release.

Team approval flows and “on-brand” templates can wait until a shop asks for them. A freelancer does not need a manager to sign off on “I am back Monday.”

I would also refuse to auto-write the replies with a model. The point is that I already know what to say to each person. A generated “hope you’re well” is the generic vacation responder in a nicer font. Store my lines. Send my lines. Nothing else.

Monetization

Free for a handful of contacts and one schedule. A personal plan when the library and the segments become the reason you open it. A small-team plan when two people share rules. I would not invent an Enterprise tier to look serious.

Per-email billing and lifetime licenses are cute and wrong for a tool you use twice a year. A subscription you can pause is closer to how leave actually works.

Next Steps: how I’d build it, and the one hard part

The catch is the whole premise. Gmail’s built-in vacation responder sends one canned reply to everyone (at best limited to your contacts). It cannot send a different message to a specific client. So the core feature, one tailored reply per contact, is what no mailbox does out of the box.

That means going through the provider’s API. On Gmail, the realistic path is a small service that watches the inbox with the Gmail API, matches the sender against your rules, and sends the right reply on your behalf. It works, but it is not a weekend hack. It needs OAuth with sensitive scopes, and Google puts apps that read mail through a security review before they can serve real users. Outlook is the same shape through the Microsoft Graph API. Deliverability matters too, since an app firing auto-replies has to stay well within rate limits or it starts to look like a spammer.

So the MVP is narrow: one provider (Gmail), per-sender and per-segment rules through the API, scheduling on top. Multi-provider “manage everything from one dashboard” is the vision, not the first release. None of this is fatal, but whoever builds it should know the per-contact auto-reply is the real work, not the UI.

The first test is selfish: my next leave, five clients, five different lines, and a sixth for everyone else. If I still open Gmail to rewrite one of them by hand, the product is not done.

Do not hide the security review from yourself. If you cannot stand in that queue, you do not have a mail product. You have a mockup. Say the scopes on the install screen in one sentence a freelancer can understand: this app reads who wrote to you, and it sends the reply you already wrote. People who leave after that sentence were never going to trust a stranger with their inbox.

Bonus: a polite way to set boundaries

There is a quieter use for this. We all have the contact who treats “out of office” as a suggestion, or the one who floods your inbox the moment you resurface. A targeted auto-reply lets you signal “not right now” without ghosting them and without rewriting the same diplomatic paragraph for the tenth time. The goal is not to dodge the work, only to keep the evening you actually booked off.

Closing

BRBMail sends one tailored reply per contact, from a library of the lines you already wrote, on a schedule that turns itself off when you come back. The hard part is the per-contact auto-reply itself, which no mailbox offers today, and that gap is what makes the idea worth building.

I would rather a narrow Gmail tool that is right about five clients than a “communication suite” that also wants to write my invoices. If the product starts offering tone suggestions, you have lost the night-before-leave problem and found a different, worse one.

If you build it, I want to hear about it. I would happily be one of the first users, test it, and update this article with my hands-on experience; this is a problem I face every time I take a few days off. Take the idea, change whatever you like. You owe me nothing, just tell me how it goes.

Vertical Slices: Playable slices before a game ships

Fri, 16 Oct 2015 00:00:00 GMT

I have wishlisted games on the strength of a trailer and a promise, like millions of people do, and watched the finished thing arrive as something else entirely. What I never got to do was play the ten minutes that actually decide whether a game is good, before it shipped.

But the idea runs deeper than “try before you buy”. Think of all the games that started development and will never be seen by anyone: the demos, the prototypes, the vertical slices of projects that died before launch. Vertical Slices is a place for developers to put those playable slices, and for players to actually play them, whether the game ships next year or never.

Save it all, and make it playable.

Underneath the feedback loop sits the preservation mission. Once a slice is on the platform, it stays playable, even if the game itself never ships.

I am not asking for a seat in the studio. I am asking for the ten minutes that would have saved me a wishlist, and for the ten minutes from the project that never got a store page. Those are the same object, a build someone was proud enough to show.

Market & Opportunities

Steam Early Access and Kickstarter already let people pay to be in the room. This is a narrower room: a polished ten minutes, a vote, and an archive. Not a live service in progress. Not a trailer with a wishlist button.

The uploaders are indie teams and mid-sized studios who cannot buy a front page, and who would rather hear that the loop is wrong before they spend the next year on it. The players are people like me, who will give an evening to a slice if it might show the real game. Journalists and publishers can lurk. They are not the customer I would design for first.

I will not hang a two-hundred-billion-dollar industry figure on that. The relevant market is “games that die in a folder,” plus the smaller set of games that ship and needed a better ten minutes first.

Existing platforms will not do the archive part. When a Kickstarter fails or a studio closes, the build disappears. That disappearance is the gap.

This is not a way to democratize game development, and it is not a replacement for a publisher QA department. A public vote can tell you the loop is confusing. It cannot tell you how to ship. If a studio wants focus groups and NDAs, they already know how to buy those. What they cannot buy later is the slice they deleted.

Key Features

A slice you can play. A build that shows the loop, the look, and a taste of the story. Not a trailer. Not a mood board. If it does not run, it does not list.
A vote and a note after. Gameplay, feel, whether you would buy the rest. Short, structured, optional prose. Developers read the pile, not a comment war.
Context, if the player offers it. Platform, a rough age range, how much they usually spend. Opt-in, aggregated, never a name next to a sentence. Signal, not surveillance.
A reputation that opens doors. Players who write useful notes see more slices. Developers who ship honest builds get seen. Invite-only rooms exist for the projects that cannot take a public pile yet.
The archive. A slice that lands here outlives the project. Studios close, funding dries up, a build gets left on a dead drive: the ten minutes still open. That is the part no storefront will do for you.

I would not launch with Steam buttons, influencer exclusives, or a suite of sentiment dashboards. Those are how a library of playable minutes becomes a marketing tool with a play button. The library comes first.

Monetization

Hosting playable builds is not free. A featured spot, labeled, can pay for some of that, as long as it cannot bury a quieter slice that people actually play.

Developers who want more than a listing (a private room, a longer life for a fat build, a thicker readout of the votes) can pay a hosting fee. I would not take a cut of a future game’s sales. That promise sounds aligned and collects nothing for years, if ever.

Hardware giveaways and publisher booths can wait. If the archive needs a patron, say so. Do not pretend a vote site prints money.

Next Steps: How I’d Build It

The wedge is not “another Early Access.” The wedge is a slice that still runs in five years. That is also the wall.

Start with one way to play, either a downloadable build for one desktop OS or a web build if the engine allows it, not both. A page, an upload, a play button, a three-question vote. Seed it with a handful of slices you have permission to keep, including at least one from a project that already died, so the archive is not a slogan.

Then the unglamorous work: storage, checksums, “this still launches,” and a rule for what happens when an exe needs a vanished middleware. Preservation is a museum job wearing a game-dev costume. If you cannot say how a 2024 build opens in 2029, you are building a launch pad, not an archive.

I would also decide, in writing, who may take a slice down. The developer, yes. A takedown notice, yes. A mood, no. An archive that evaporates when a studio gets shy is a trailer host. Say the rule before the first upload.

Do not ask players for a passport. Email is enough to play and comment. Trust accumulates. Full identity checks belong only on developers who receive money through the platform, if that day comes.

The first public week should feel slightly empty and slightly stubborn: a few slices, a vote, a promise that those files will still open next year. If you need a launch party to hide the emptiness, you launched too late or too wide.

Bonus: the abuse you can see from here

Fake reviews meant to sink a studio run into reputation, a report button, a human. Leaks from an unannounced project run into invite-only rooms and terms that say the slice is not a trailer you own. I would not promise cinematic watermarks you cannot staff.

Vote rings and spam listings are ordinary. Make uploading a slice slower than posting a link. IP theft is a fear every developer already has. Give them a plain page on what they are granting when they upload, and do not pretend the platform can sue the internet for them.

Closing

Vertical Slices keeps, on purpose, the ten minutes I never got to play, including those from games that never arrive. It is a living library of playable slices from projects that would otherwise disappear without a trace. If the first year only saves twenty builds, and five of them are from games that died, that year already did the job a storefront never will.

If you build this, I would love to see it, and to be an early tester. Playing slices and arguing about them is exactly my idea of a good time. Take it, rename it, change the model. You owe me nothing, just tell me how it goes.

Pixel Prints: Get your best gaming moments out of the folder

Tue, 10 Mar 2015 00:00:00 GMT

Somewhere in my phone there is a screenshot of an unbelievable Battlefield kill streak, and a few more big scores from other games I have never shown anyone. But the score is not really the point. What stays with me is the intensity of certain in-game moments, the split second where everything went right, and those deserve to be immortalized far more than any number. Instead, they sit in a folder, seen by no one. That is the small waste Pixel Prints is about. The best moments we capture in games almost never make it out of the hard drive and onto a wall.

Getting your captures in is simple. You upload them yourself. On PC that is painless, since Steam and most capture tools already save screenshots to a local folder the app can watch. Console shots take a quick share-to-phone detour first. From there you pick a treatment (pop, quiet, a date, a tag if you want one) and it prints on paper that can survive a wall.

I have never hung one of those shots. I have shown them in a chat and watched the thread move on. A wall is slower than a chat. That is the point. If the print is only a flex, it will look like a flex in six months. If it is the second where the match turned, it might earn the nail.

Market & Opportunities

Gamers already take the pictures. They already buy chairs, lights, and the occasional poster of a character they did not capture. The gap is the frame they were actually in.

I will not hang a global gaming-merchandise figure on that. Twitch and YouTube made the culture louder; they did not invent a print shop with a rights problem. Who buys first: PC players with a folder of shots, and the friend who wants a gift that is not another hoodie. Who does not: anyone expecting an official Battlefield cinematic on canvas.

That version is a lawsuit wearing a shop.

The clean wedge is user-authored space (a Minecraft build, a map you made) and games whose screenshot policies are actually permissive. Licensed franchises come later, if they come, and only with a deal. A “beloved brand” speech will not get you that deal.

Gifts are the second buyer. Someone who did not take the shot buys the wall for the person who did. That only works if the upload is easy enough that the gamer will do it. If the gift requires a 40-minute editor, you will sell posters of other people’s art again, and we are back in the legal box.

Key Features

Upload that does not feel like homework. Drag and drop, batch import, and on PC a watch folder for Steam or your capture tool. Console is share-to-phone until a platform opens a door that does not exist today.
A few treatments, not a design school. Minimal, pop, a date, a gamer tag, a stat if the file has one. If the editor needs a tutorial, it is too much.
Paper, canvas, a frame. Museum-weight paper as the default. Canvas and metal as the extras people pay for when the shot is worth it.
A human look before it prints. Resolution, crop, a shot that will look like mud at 50 cm. A short quality guide on the way in saves the argument on the way out.

I would not launch with official developer collections, guest-artist series, or in-game rewards. Those collide with the legal section below, and they turn a print shop into a licensing company. Get one good print of a shot the user owns first.

Monetization

Sell the print. Price by size and material. That is the business.

A monthly subscription that mails a surprise capture is a cute idea and a storage problem. Merch (phone cases, shirts) can wait until the wall print is something you would hang. Limited signed editions assume a publisher who said yes. Assume they have not.

The legal reality: whose image is it?

A game screenshot is not just your moment, it is also the publisher’s copyrighted art. Selling a print of a Battlefield or Diablo frame means reproducing someone else’s intellectual property for profit, and most publishers’ terms do not allow that. This is the single biggest risk of the whole idea, and it has to be designed around, not ignored.

So lean on games with permissive screenshot or creator policies, pursue licensing deals for the big franchises, and keep purely user-authored content (your own Minecraft build, your own map) on firmer ground than a canned cinematic. Copyright protection is essentially worldwide (the Berne Convention covers some 180 countries), so this is not a loophole you can geo-dodge either.

Next Steps: How I’d Build It

A weekend shop: upload, one paper size, one treatment, a print-on-demand partner, a page that says what you will not print. Seed it with shots you have the right to hang. If three friends will pay for their own Minecraft room, you have a pulse. If they only want the Battlefield streak, you have an article and a legal memo.

I would put the legal paragraph on the upload page, not in a footer. People should know, before they crop, that a canned cinematic is not for sale here. The ones who leave were going to be the angry review. The ones who stay will bring the shots you can actually print.

Manual upload has to feel complete. Direct Xbox or PlayStation libraries would need partnerships and APIs that do not publicly exist today, and even a local Steam folder needs care when the path changes. Build the upload so well that a platform deal is a bonus, not a crutch.

Shipping is the unglamorous half: tracking, a tube that does not crush, and honesty about international postage. Localized print-on-demand is how you avoid learning customs the hard way.

Color is the other quiet failure. A screenshot that looks fine on an OLED will print like a cave. Show a proof that is worse than the screen, on purpose, so the buyer is not surprised. A reprint policy for your own bad crop is kindness. A reprint policy for “it doesn’t look like the trailer” is how the IP section comes back as a ticket.

Closing

I keep coming back to that Battlefield streak sitting in a folder on my phone. A licensed cinematic on the wall is a legal problem. Your own Minecraft build, or a moment you were in, is the version I would hang.

I would rather sell twenty honest prints of rooms people built than one viral poster of a franchise I cannot license. The shop can stay small. The legal memo cannot. If you only remember one paragraph of this article, remember the one about whose image it is. The rest is paper and postage. Hang the room you built. Leave the cinematic in the folder.

If this concept is yours to build, take it. And if you do, I would love a print of that streak, or a note explaining why the idea does not work. Digital memories deserve a wall. You owe me nothing, just tell me how it goes.

5Essentials: Community-Driven Minimalism, Five Items at a Time

Fri, 01 Aug 2014 00:00:00 GMT

Right after moving into a new flat, staring at a long list of things to buy, I did the most useful thing I could think of and asked my friends and family what they actually used most at home, category by category. Their answers were gold. A handful of trusted, battle-tested recommendations saved me money, saved me time, and spared me a couple of purchases I would have regretted, and I still rely on some of them today.

5Essentials is that reflex turned into a platform: everyone’s five true essentials, in every category, one place to find them. Kitchen, cleaning, a first apartment, a desk. Five, not fifty. Quality over quantity, fewer but better things, less overconsumption.

The public side is a vote. People post a five, other people agree or they do not, and the items that keep showing up rise. The private side is the one I actually wanted when I moved: your mother’s kitchen five, your friend’s office five, a list you can ask for and keep.

I still use some of those answers. That is the bar. A list that looks clever and gets replaced in six months is a blog post. A list you are still cooking with is the product.

Market & Opportunities

People are already trying to buy fewer things and waste less money on the third spatula. The internet answers that itch with ten “best of” roundups that all link the same affiliate toaster. A shortlist from someone you trust is a different product.

Anyone staring at an empty flat shows up first, along with anyone who is tired of replacing junk. The money comes later, from a membership for people who live in the lists and from brands who want a clearly marked sixth slot under a community five, never inside it.

I will not hang a global e-commerce figure on this. The relevant market is “I need a kettle and I do not want to read forty reviews.” Amazon already sells the kettle. 5Essentials tells you which kettle your people kept.

This is not for the person who wants a store. They have one. It is also not a sustainability lecture. If fewer, better things happen to waste less, good. I would not sell the site as a movement. I would sell it as a shorter shopping list.

Key Features

Paste a link, then correct it. A URL should fill the name, the photo, and the price. Retailer pages are messy, so a human edit is part of the design, not a failure.
Public fives, with votes. One category, five items, a score. The five is the product. A sixth item that is an ad is labeled as an ad, or it does not exist.
Private lists. Share with family and friends the way I asked around my move. This is not a worse Pinterest. It is a short answer from people who will still be in your kitchen next year.
Search that stays narrow. Category, price band, a brand if you must. Not a marketplace homepage.
Rot and honesty. Broken links, dead products, and lists that look bought get a queue. Votes get gamed. Design for that on day one.

I would not ship a mobile app as a feature. A site that works in a shop browser is enough. Native can wait until someone is standing in an aisle and refusing to leave.

Monetization

Free to read the public fives and to keep a couple of private lists. A small membership can unlock more lists, a history of what you asked, and an ad-free page if ads ever exist, which I would rather they did not.

Sponsored products sit in a sixth slot, labeled, capped, and still allowed to lose the community vote. If a brand cannot live under the five, it does not belong here.

Affiliate commissions can pay the lights, through a network, not through a side door where the five suddenly tilts toward whoever pays more.

I have watched too many owners drift from the mission to push whatever pays them. That drift is the one thing this site cannot survive.

I would not sell individual shopping profiles as “insights.” Aggregated, boring trends, maybe, much later. Day one is lists.

Next Steps: How I’d Build It

The hardest part is the data, not the voting UI. “Paste a link and everything fills in” sounds trivial but retailer pages are messy, and scraping them can run against a store’s terms of service. Realistically you start with two or three retailers whose product feeds or affiliate APIs you are allowed to use, and let users correct the rest by hand.

Then there is the cold-start problem. A community shortlist is worthless with no community. Seed the first lists yourself in a handful of categories you know well, so the site is useful on day one for a visitor who contributes nothing.

Finally, budget for moderation from the start. Votes get gamed, links rot, and sponsored slots invite abuse. A small manual review layer on top of automated checks is cheaper to design in early than to bolt on after launch.

The first test is selfish and small. Recreate the lists I got from family when I moved, in public, and see whether a stranger adds a sixth item I would actually buy. If they do not, you have a pretty empty grid. If they do, add private lists next, then the sixth slot, then membership.

Start with three categories you can defend without a search: a first kitchen, a cleaning cupboard, a desk. If those three cannot stay useful for a month, twenty categories will not save you. I would rather a small site that is right about pans than a taxonomy of lifestyle.

Bonus: the abuse that shows up first

Email plus a rate limit is enough identity for a list of kitchen gadgets. Full ID verification is overkill and a data-protection headache. Fake accounts will still show up. Make posting a five slightly slower than posting a comment.

A report button that reaches a person is enough misuse policy for launch. Spam lists, bought votes, and a brand sneaking into the five are the cases that matter. Write those three on a short rules page and enforce them. The rest of the trust-and-safety laundry can wait.

Closing

5Essentials comes down to five things from people you trust, and a public vote for everyone else. I once held a domain for this one and have let it go, so the name is free, and so is the idea. If you are serious about building it, get in touch anyway. I would love to compare notes.

I would rather a small, slightly stubborn site than a recommendation engine that discovered a sixth monetization. If the five starts to look like a storefront, you have lost the move-in conversation that started this. Five items. People you trust. That is the site, and it is already more useful than another roundup.

If you build this, tell me. And if you do, I would love a free account to review it, and to feature the real thing, with honest notes, in a future article.

Virtual Stress-Buster: A safe, private outlet for stress relief

Sat, 26 Apr 2014 00:00:00 GMT

At the agency, I was managing more than 130 client accounts a month. Most days that was fine. Some days, what I needed at 7 pm was not a meditation app, it was somewhere to let the pressure out, loudly, without consequences and without breaking anything that costs money. That is where this idea comes from. It is a place to smash something that is not real, blow off the steam, and put the phone down calmer than you picked it up.

You have probably heard of rage rooms, those places where you pay to smash old printers with a baseball bat. I love the concept and hate the logistics. You have to book, drive there, wear goggles, and pay for every plate. My idea is the same release, but virtual. In a simple app, on a phone or a VR headset, you pick a generic cartoon avatar or a symbolic object, and you get a few minutes of gloriously silly destruction. Smash it, drop a piano on it, watch it crumble. Then a short breather, and back to your evening.

The key word is generic. The avatars are stick figures and cartoon characters from a built-in library. You can tweak a hairstyle or an outfit, but you cannot upload photos, real faces, or company logos. This is deliberate and non-negotiable.

The moment the thing on screen looks like your actual boss, the app stops being a pressure valve and starts being something darker.

Symbolic targets only, always.

What this is, and what it is not

It’s a playful outlet, a way to vent, not therapy and not a treatment. And I want to be careful here. The old “let it all out and you’ll feel calmer” idea is actually contested by researchers, and some studies find that rehearsing aggression can feed it rather than drain it. So the promise is small and human. A few minutes of silly, satisfying destruction might take the edge off, inside guardrails (short sessions, a breathing prompt, links to real help if someone needs it). This is not a clinical claim to reduce real-world aggression.

That framing shapes the whole product. Sessions are short by design. The tone is cartoonish, not gritty. And the app is upfront about its own limits. If someone is using it every day for hours, the right move is to gently point them toward real support, not to sell them a bigger hammer.

Market & Opportunities

Who it’s for

The obvious users are people like me on a bad Tuesday: professionals, students, parents, anyone in a high-pressure stretch of life who wants a quick, private way to decompress. My guess is the core audience skews 18 to 35, people who already treat their phone as a wellness tool, but frustration has no age limit. Anyone who has ever wanted to throw a laptop out a window is a potential user.

Why this gap exists

The stress relief app market is crowded, but it is crowded on one side of the room. Headspace, Calm and their many cousins are all about slowing down. Breathe, meditate, listen to rain sounds. That works for a lot of people a lot of the time. But some days you are not in a slow-down mood, you are in a get-it-out mood, and for that mood the app store has very little to offer.

The missing piece is an active outlet instead of a passive one. Rage rooms prove the demand exists, since people literally pay to break things. A virtual version removes the cost, the travel, the broken glass and the cleanup, and it is available at 7 pm on a Tuesday in your living room.

Room to grow

Stress is not going anywhere, and neither are phones and headsets. A digital product like this can launch small, in one language and one market, and scale from there. Frustration translates well. I would still validate the retention question early. Is this something people come back to weekly, or a novelty they try twice? That single metric decides whether this is a business or a toy.

Key Features

Avatar library and symbolic targets

Users pick from a library of generic figures: stick people, blobby cartoon characters, and a set of symbolic objects for non-human stressors. A ringing phone for the inbox that never stops. An alarm clock for the deadline. A tiny cardboard office for the whole open-space experience. Light customization is available (hairstyles, outfits, accessories), but as said above, no photo uploads, no real faces, no logos, ever.

Guided smash sessions

Destruction comes in flavors. The physical kind means smashing, crumbling, tearing, or whacking with an oversized cartoon mallet. The environmental kind places the avatar in a scene and lets a storm or a tiny earthquake do the work. The abstract kind, for quieter moods, lets you watch the thing slowly fade or dissolve into confetti. Satisfying sound effects throughout, because half the relief is in the crunch.

Every session is framed and capped. After five to ten actions, the app asks you to take a short breathing pause before you continue. That cap is a feature. The point is a quick release with a calm landing, not an hour of stewing in your own anger. After each session the avatar resets, ready for the next bad Tuesday.

Optional check-ins

After a session, the app can ask one or two light, optional questions. What got to you today? Did this help? Users who opt in get a simple personal log over time (which triggers came up most often, which methods they favored, how sessions trended). It is a mirror, not a diagnosis.

The same check-ins power the safety net. If someone’s usage pattern looks like reliance rather than release (long daily sessions, weeks on end), the app quietly suggests other coping tools and links to real professional resources. Aggregated and anonymized, the data might also one day say something interesting about what stresses people out in different places, but that is a distant maybe. This is a pressure valve, not a public health observatory.

User experience

The whole flow should take under a minute from opening the app to first smash. A short onboarding, a clean interface, and saved favorites so your go-to avatar and go-to mallet are one tap away. An optional community space lets users swap stress management tips and laugh about their week, with moderation to keep it kind. Cross-platform from the start, with the phone first for reach and a VR headset for the full swing-your-arms experience.

Monetization

Freemium base

The free tier gives everyone the core loop: a handful of avatars, basic customization, and two or three destruction methods. Enough to feel the relief and decide if it is for you. Paying should never be the price of calming down.

In-app purchases

The fun stuff is where the money is. Elaborate destruction tools, like a trebuchet, a black hole, or a stampede of llamas. Extra avatar outfits and accessories. Special environments, seasonal or absurd (office at Christmas, tax season bunker, Monday morning simulator). None of it gates the basic release, all of it makes the release funnier.

Subscription

A monthly membership bundles the extras: early access to new tools and environments, an ad-free experience, the premium community space, and the personal insights log with longer history. Subscribers get depth, while free users still get the full pressure valve.

Ads, carefully

Light, clearly separated ads can support the free tier, shown between sessions and never during one. Brand partnerships could work for cosmetic content, a sponsored silly environment for instance, but never as destruction targets. You do not get to pay to have people smash your competitor. That door stays closed.

How I’d Build It

The technical shopping list is refreshingly boring: a game engine with decent physics (Unity or Unreal both do cartoon destruction well), a built-in avatar system, cloud accounts, and a standard payment stack. Nothing here requires inventing new technology, which is what you want for a first version.

The MVP is one avatar, one room, three ways to smash, and the session cap with the breathing prompt. Put that in front of fifty stressed humans and watch two numbers. Do they smile, and do they come back next week? If both answers are yes, expand the library, add the check-ins and the community, then the subscription. If not, you have spent a small budget learning something true.

Bonus: keeping it from turning ugly

An app about venting anger has an obvious dark side, and pretending otherwise would be naive. Here is the guardrail set I would ship from day one.

Symbolic only, enforced. No user uploads of faces or logos, no naming avatars after real people, and filters plus human moderation to catch workarounds. The whole design keeps targets abstract, because the abstraction is what makes this a game and not a rehearsal.

Session caps and cool-downs. The five-to-ten action limit and the breathing pause are hard-coded. The app is built to end sessions, not extend them.

Reliance flags, handled with care. Usage patterns that look like dependence trigger a private, gentle nudge toward other resources and professional help. Support, not surveillance. The data serves the user, never a punishment system.

Community standards. Clear rules in the community space, age verification, and a simple reporting flow with transparent follow-up. Repeat harassment gets people removed. The venting happens at cartoon avatars, never at each other.

Closing

Strip away the feature list and this idea fits in one sentence. Rage rooms are fun, and a virtual one in your pocket would be cheaper, safer and available exactly when the bad day happens. Keep it symbolic, keep it short, keep it honest about what it can and cannot do for your head, and I think there is a real product here, somewhere between a game and a very loud sigh.

If you build this, tell me, and send me a code. After some agency days, I will be your best beta tester. Take the idea, rename it, reshape it. You owe me nothing, just tell me how it goes.

MetaOptimize: Automated A/B Testing for Web Metadata

Tue, 14 Jan 2014 00:00:00 GMT

I spent years running an SEO and marketing agency, and meta titles and descriptions were one of our quiet obsessions. We tested them, rewrote them, tested them again, because our own client dashboards kept showing the same thing: get that one grey line under the blue link right, and click-through rate can climb surprisingly fast. What always struck me is how little attention that line gets outside agencies.

Meta titles and descriptions are the storefront window, and most teams write them once and forget them.

MetaOptimize scratches that itch. It treats metadata as something you test and improve, not something you guess at once and leave alone.

In the agency years the work was always the same, and always slower than it should have been. Pull Search Console. Find the queries that already get impressions and lose the click. Rewrite the title. Wait. Guess whether the new line helped, or whether Google rewrote it for you, or whether the page simply had a bad week. We got better at the rewriting. We never got a tool that closed the loop without a human babysitting a spreadsheet.

The timing helps. Search keeps rewarding pages that sound like a person wrote them for a person, and most teams still write metadata that is either invisible (no keywords, no promise) or unclickable (stuffed, shouty, nobody trusts it). Current language models are finally good at proposing several honest variants of the same grey line. What is missing is a loop that writes, tests, and keeps the winner, not more copy.

Market & Opportunities

The SEO tool market is large and still growing. I will not pretend I have a sourced billion-dollar figure; the demand is visible enough from any agency inbox. What matters for this idea is a narrower gap.

The big suites chase enterprise accounts and treat metadata as one tab among fifty. Small teams get either that overkill or nothing. They rewrite a title by gut, wait a month, and guess whether it worked. A tool that does one thing (test the words on the SERP, keep the winners) and prices itself for a freelancer has room to live between the two.

The people who pay are freelancers and tiny agencies who already live in Search Console, and later the in-house marketer at a company too small for an SEO retainer. As for why now, generating variants is cheap, the Search Console API already hands you impressions and clicks per query, and nobody serious is still A/B testing titles by hand in a spreadsheet.

I would start with English-language sites that already have traffic. A tool that promises to invent demand is lying. A tool that promises to stop wasting the demand you already have is a conversation you can have in one meeting. Local businesses, SaaS marketing pages, and content sites with a long tail of almost-ranking articles are the same job wearing different clothes.

This is not for a brand-new site with twelve sessions a month. There is nothing to test. Wait until Search Console has a real sample, then come back.

Key Features

Variants from the page, not from a prompt box. Point it at a URL. It reads the page and proposes several title and description pairs that a human can edit before anything goes live. The model is a first draft, not the publisher.
A/B tests with rules you can explain. You set a minimum sample and a win margin. The tool splits traffic or, more realistically at first, rotates titles over time and compares click-through on the same query. No “AI decided, trust us.”
Snippet win/loss. Straight from the Search Console API: queries where you appear and do not get the click. Those are the titles worth rewriting first.
Autopilot, with a leash. Once a variant wins, it becomes the default. The tool keeps a long view so a lucky week does not overwrite a real winner. You can turn the leash off. Most people should not.
A simple history. What you tested, what won, what you rolled back. The point is a service book for the storefront window, not another dashboard of vanity charts.

I would not bolt on image SEO, site-speed scripts, or a competitor-keyword suite. Those are how this idea becomes the tab among fifty it is trying not to be. If a user asks for keyword research, point them at the tool they already have. MetaOptimize should be slightly boring on purpose: titles, descriptions, tests, winners.

Monetization

Three indicative tiers, cheap enough for a freelancer, then up for agencies:

Basic ($29/mo): a handful of pages, manual variants, one test at a time.
Pro ($99/mo): more pages, win/loss view, geotargeted Search Console data, autopilot.
Enterprise ($499/mo): the full site, API access, and a human who will look at the messy cases.

Pricing could sit higher; the shape is the point. A free trial of one page is enough to prove the loop. I would not fund this with ads inside the tool, or with affiliate links to hosts and consultants. That is how a metadata tester becomes a junk drawer.

Next Steps: How I’d Build It

The MVP is one site, Search Console connected, titles only. Generate five variants for the ten queries that already get impressions and lose the click. Rotate, measure, keep the winner. Descriptions come second. Autopilot comes last.

The real walls are not the model. Google’s title rewriting can ignore your carefully tested line, which means a “win” in your log is sometimes a title nobody ever saw. The Search Console API is delayed and coarse, so early tests will be slower than a lab A/B, and you should say that on the pricing page before a customer finds out the hard way. And “customer data” should mean on-page copy and search queries, not a CRM dump. Stay on the page.

A useful demo, not a product: a custom GPT that rewrites one title through your API. Fine for a video. Do not confuse it with the loop. If the first ten paying users are agencies, give them a weekly digest of what won and what got rewritten by Google. That digest is more valuable than another dashboard.

Bonus: the misuse that actually matters

The real risk is a user asking the model for cloaking, bait titles, or stuffed fragments that would get a site filtered, not a cartoon hacker. The tool should refuse those patterns, say why, and keep a log of what it declined. Educational blog posts about “ethical SEO” will not do that job.

Closing

MetaOptimize is a focused answer to a neglected problem. It pairs systematic tests with model-written variants, at a price small teams can actually afford.

I would rather see a small tool that is right about ten titles than a suite that is vaguely helpful about everything. If you feel the same itch I felt in those agency years, the grey line under the blue link, this is the one I would start.

If you build this, tell me. I would love to see what you make of it, and I am happy to trade notes if that helps.

EventForge: Run interactive events with your community

Tue, 27 Aug 2013 00:00:00 GMT

Fifteen years ago I was a YouTuber. I had found a great team online and we made heavily scripted Battlefield 3 videos, mini-films really, what we used to call machinimas back then. I loved every minute of it, and organizing those shoots was chaos. The plan lived in one place, the team in another, and the game in a third, and half the plan never reached the people playing.

A tool built for running events with an online community would have been a gift. Fifteen years later, I watch streamers organize community events, and there is still no dedicated tool. That gap is where EventForge started. It gives a creator one place to run an interactive event with their community, instead of duct-taping five apps together.

What I would build

EventForge is a platform for gaming creators and their communities, and its one job is running interactive events together. For creators, that means designing an event, assigning roles to participants, keeping everyone on the same page, and tracking what actually happens once the event goes live. For fans, it means a real way to take part, not just typing in chat while someone else plays.

That second half matters as much as the first. On the big platforms today, fan participation mostly means comments, chat, and the occasional poll. But the people who care most about a creator’s content are exactly the ones who would show up, play a role, build something, or help moderate if there were a structured way to do it. EventForge gives creators the coordination tools and gives fans a seat at the table, through roles, contributions, and rewards.

To be clear about where this comes from:

It is not market research, it is a scar.

Every event tool I have seen since those machinima days was built for conferences or esports brackets, not for a creator who wants to run a story-driven night with forty of their regulars.

Why not just Discord?

Fair question, because Discord already does servers, roles, and event pings. The bet here is narrower. Discord is a chat tool that events are bolted onto, while this is built around the event itself (the scenario, the participant roles, the live tracking, the after-action stats). If that focus does not feel sharper than a well-run Discord plus a spreadsheet, the idea does not have a reason to exist, and that is the first thing to test before writing any code.

The market, without fake precision

Gaming is now a mainstream, billion-player industry, and the live, creator-driven side of it (streams, tournaments, community events) is where a lot of the energy has moved. Exact figures vary by source, but the direction is not in doubt. Audiences and creator counts keep climbing, and the tools to run events for those audiences have not kept up.

For this idea, the number that actually matters is how many mid-size creators run community events regularly, and how much pain that causes them. That number is smaller and harder to find in any report. My bet, from watching them do it live, is that the pain is real and the workarounds are ugly. That is a thing you validate with twenty conversations, not with a market chart.

Key features

Roles that make fans participants

Creators can define custom participant roles for each event, things like ‘Lead Strategist’ or ‘Creative Contributor’, so community members carry real responsibilities instead of just watching. A participant rating system lets people give each other constructive feedback after events, which rewards the reliable regulars and keeps collaboration healthy.

Picture a Star Citizen creator running a narrative mission night, with community members assigned as ‘Piracy crew’, ‘Salvage team’, or ‘Mining squad’ based on their in-game professions and skills. Everyone knows their job before the server fills up. That is the kind of structure our machinima shoots never had.

Scenario Builder

This is the heart of the platform for me. The Scenario Builder lets a creator design the event itself: objectives, tasks, phases, win conditions, whatever the format needs. A scripted roleplay night, a scavenger hunt, a community versus creator challenge, all of them start as a scenario that participants can actually read and follow. The plan finally lives in the same place as the people executing it.

Event creation and management

Around the scenario sits the boring but essential machinery: scheduling, customizable templates, automated reminders, and recurring event options. During live events, real-time participant tracking and engagement monitoring give the creator and their moderators a control room view, and streaming platform integrations keep the broadcast side simple.

A Fortnite creator could template a recurring Friday Community Showdown once, then let reminders and signups run themselves. During the session, participant queues and live tracking replace the moderator frantically alt-tabbing between five windows.

Customization

Both sides get to make the space theirs. Community members customize profiles with themes and avatars. The Minecraft player with the Creeper avatar and biome background gets to look like themselves. Creators go further with branded event pages and custom stream layouts, so an event feels like their show, not a generic lobby.

Analytics that answer real questions

After each event, creators get the numbers that matter: registrations, participation levels, drop-off points, and audience sentiment during and after the event. A League of Legends creator running a weekly ARAM tournament can see which formats fill up, which ones bleed participants halfway through, and adjust.

The platform also keeps community rankings (highest rated events, most active moderators, top contributors). Recognition is cheap to give and it is half of why people volunteer their evenings.

How it could make money

Premium memberships

A premium tier for creators could include advanced analytics like session reports and loyalty tracking, branding options such as a custom URL and logo overlay on event pages, and priority support. For community members, a paid tier could mean an ad-free experience, early registration, exclusive events, deeper profile customization, and the option to donate to a creator for a higher rank in an event.

Something around $9.99 per month or $99 per year feels like the right neighborhood to start testing, with custom plans for bigger creators. Those are opening guesses, not projections; the real price comes from what the first paying creators say no to.

Sponsorships and brand partnerships

Gaming brands already sponsor creators, and sponsoring their events is a natural extension. Partnerships could unlock in-game content for participants or fund sponsored creator battles streamed to the partner’s channels, on a revenue share basis. This only works once the events are good, so it is a later chapter, not a launch feature.

Advertising

A small number of contextual ad placements, plus opt-in rewarded video for members who choose it, can round out revenue without junking up the interface. I would treat ads as the smallest and last lever, not a pillar.

How I would start, and the real walls

Here is the part I would get wrong if I let the feature list seduce me. The technology is not the wall. User accounts, event pages, role assignment, a scenario editor, live status, none of that is exotic. Any competent small team can build it.

The real wall is frontal competition. Discord already owns the community layer. StreamElements and Streamlabs own the stream-side tooling. Battlefy and Toornament own tournaments. A full EventForge platform asks creators to move their community’s center of gravity, and creators, rightly, refuse that trade unless the payoff is enormous.

The second wall is creator acquisition. Creators are busy, drowning in tool pitches, and their audience is their livelihood. Nobody experiments on their community with an unproven platform. You do not solve this with ads. You solve it by making five specific mid-size creators personally love the thing.

So the realistic MVP is not a platform at all. It is a simple event tool that plugs into a creator’s existing Discord instead of replacing it. The creator builds the scenario and defines the roles, members sign up through the server they already live in, and the event runs with live tracking and ends with clean after-action stats for the creator. If that thin slice beats the spreadsheet, there is a business to grow. If it does not, better to learn that in three months than in three years.

From there, the path is ordinary and healthy: a private beta with a handful of creators who run events monthly, brutal feedback loops, and only then the bigger platform ambitions (streaming integrations, monetization, the works).

Misuse and safeguards

Any platform that hands community members roles and visibility will attract the usual problems, so moderation has to be designed in, not patched on. That starts with clear community guidelines that every user agrees to at onboarding, explicitly covering spam, harassment, hate speech, and trolling.

On top of that foundation come automated filters to catch spam and obviously inappropriate content, an easy and visible reporting mechanism, and a graduated ladder of consequences from warnings to temporary suspensions to permanent bans. Critically, these tools belong in the hands of creators and their moderation teams, because they know their communities and they carry the reputational risk when an event goes sideways.

The operational risks are the standard ones and deserve standard answers: redundancy and load balancing against downtime, encryption and regular security audits against breaches, and honest ongoing attention to competitors, because in this space the incumbents can copy a good feature fast.

Closing

Strip away the feature list and this idea fits in one sentence. Creators run events with their communities all the time, and nobody has built them a proper tool for it. The concept survives contact with skepticism only if the Discord question gets a convincing answer, which is why I would test the thin plug-in version first and let real creators decide whether the focus earns its keep.

I would have used this fifteen years ago for our machinima shoots, and I would use it today. If you build it, tell me. I will be in the front row of your first event. Take the idea, reshape it, and rename it too. EventForge is only a placeholder. You owe me nothing, just tell me how it goes.

Athoughtfor.me: A sky of thoughts for the people who matter

Wed, 09 Jan 2013 00:00:00 GMT

I will admit it: I am a hopeless romantic. There are people from my past, old loves mostly, I still think about from time to time. Not to reopen anything, and certainly not to intrude. They are part of my story, and once in a while a memory surfaces and I catch myself hoping they are doing well. I have often wished there were a gentle way for them to know, if they ever cared to look, that someone thought of them this year. And I would love to know if the same were true for me.

Athoughtfor.me is the space I wanted for those moments, a place to let someone know they crossed your mind, with the option to add a heartfelt message.

The sky, not the feed

Unlike a social network, there is no feed to scroll and no followers to collect. You send a thought to a name or an ID you invent, such as a nickname, a first name, or something as thin as “Jean C.” The other person, if they ever come looking, searches that same string and finds what was left for them.

It is also not a dating app, and it is not a way to reopen a conversation someone already closed. If a note reads like a request for a reply, a moderator should send it back.

The sky holds a thought. It does not deliver a message.

The picture I have always had is a sky of floating notes. You drift through it the way you drift through a quiet room, not the way you refresh a timeline. Some notes are for you. Most are not. That is the bet. Not another place to perform, but a page people visit once in a while, type a name, and hold their breath.

I do not want a notification that someone thought of you. That turns the sky into a dopamine machine, and the whole point dies. The visit is the product. You come when you want to know, or when you have something to leave, and then you leave the tab. If the site starts chasing you, I would consider the idea failed even if the numbers looked good.

The .me domain hack is part of the charm, but I no longer hold “athoughtfor.me” or “athoughtforme.com”. If you build this, check whether they are still available, or pick a name that fits your own vision.

Key Features

A thought, addressed. Write to a name. Optionally add a short message. No account required to leave one, at least at first.
Search for your name. The only discovery that matters: type what people would call you, and see if anyone left something. Browsing the rest of the sky is optional, and should never become a hunt.
A sky, not a wall. Notes float. The interface is the product as much as the copy. If it looks like a forum, the idea is already lost.
Moderation before it goes live. An LLM filter plus a human queue. Flagged notes get a warning and a chance to rewrite; repeat abuse gets banned. The sky stays a place you would let someone you love walk through.
Ambient music, optional. A quiet layer for people who want the visit to feel like a room, not a form. Off by default.
Removal as a right. Anyone who finds a note about themselves can ask for it to come down. Prefer initials and nicknames over full legal names, so the sky does not become a directory.

The name field is the whole architecture. Too precise and you have a searchable phone book of other people’s feelings. Too vague and nobody ever finds what was left for them. “Jean C.” is the compromise I keep coming back to: enough for the person who knows, useless as a lookup table for a stranger. If you build this, spend more time on that field than on the animation of the sky.

Monetization

Free to leave a short thought and to search. A modest membership unlocks the extras that do not change the core: a longer message, one image, a coarse place tag (a city, not a street), and a seasonal skin if you want a different sky in late December.

Virtual gifts are the only idea I would treat with care. If they exist, most of the money should leave the company and go to a named charity, in public, or the feature is just a sticker shop wearing a halo. I would not sell ads against other people’s unspoken feelings.

Event worlds (a sky reserved for a wedding, or for a memorial) can wait until the empty sky already works. They are pretty. They are also how a quiet product starts asking people to perform a season. Build the blank page first.

Next Steps: How I’d Build It

One page: the sky, a form, a search box. No accounts. A moderation layer in front of every note. That is a weekend prototype, and it already answers the only question that matters: do people come back to search for their name?

If they do, add accounts so someone can edit or delete what they left, then the membership extras. If they do not, you have lost a weekend and learned that the daydream was only mine.

The first audience I would try is not “everyone who has ever missed someone.” That is a slogan, not a launch. I would start with a small circle that already writes letters they do not send, like a writing group, a grief forum that wants a gentler tool, or a handful of friends who will tell you the sky feels cheap. If those people search twice, you have something. If they leave one note and never come back, you have a postcard generator.

Keep the first version ugly if you have to. A slow sky with honest notes beats a polished empty one. The animation can wait until someone other than you has searched for their name.

Bonus: the misuses that are not theoretical

Spam and cruelty are the obvious ones, and the filter-plus-queue above is for those. The quieter ones matter more here. Impersonation, where someone leaves a note as if they were you. Commercial drive-bys. And the one I would design around from day one is naming a person who never asked to be findable. Initials, nicknames, and a real removal path are not polish. They are how this stays gentle.

Closing

Athoughtfor.me is a sky for the thoughts we do not send, waiting for the day someone types a name. Feel free to rename it. I no longer hold the domains, and the idea does not need that particular hack to work.

I still think about those people from my past. I still would not message them. The whole point of this page is that both of those sentences can be true at once, and that a stranger might one day type a name and find a note that was never meant to demand an answer.

If you build this, tell me. Whether you run with it alone, with friends, or with your company, you have my full support. I would love to see my old daydream floating in someone else’s sky.

PlayPro Connect: Train with athletes who have really competed

Fri, 08 Jun 2012 00:00:00 GMT

I am not a semi-professional athlete myself, but through friends I know several people who are. They are good enough to compete, good enough to coach anyone I know, and yet outside the handful of big competitions with real prize money, most of them struggle to make their sport pay. Meanwhile plenty of us would love a real trainer nearby, someone who has actually competed, not another app spitting out a generic workout plan. The two problems looked like they solved each other.

PlayPro Connect started there. It is a platform that lets semi-professional athletes earn a living by coaching the rest of us, in person at a local park or over a live video session from anywhere. Athletes get a flexible income stream and keep a foot in the game they love, and everyday sports lovers get access to people who have really done it. And around the training, a community forms, which is the part I care about most. A coach you book once is a transaction. A coach you see on Saturdays, with two other people from the same park, is the reason the idea is not just another marketplace.

Market & Opportunities

Semi-pro careers are short, and income often drops when they end. I would rather not lean on a single tidy statistic I cannot source, but the need for a second income is real. On the other side, people are tired of a plan that does not know their park, their knees, or the fact that they would rather train with a person than with a video.

Hybrid sessions (a park on Tuesday, a call on Thursday) are now ordinary. Group sessions are how a booking becomes a habit. PlayPro sits at that intersection: athletes who need the money, people who want a coach who has competed, and a map that puts them in the same place. I would start in one city. A global fitness figure will not tell you whether ten coaches there can fill a Saturday. If they cannot, twenty cities will not save you. If they can, the second city is a copy with new insurance paperwork, not a new product.

Key Features

A map of people, not of gyms. Nearby coaches and open sessions, the way a dating app shows who is actually around. Distance is the product.
Book a session, pay once. Individual or group, in person or on a call. Stripe or the equivalent. Move it, cancel it, do not invent a new checkout.
A review after every session. This is not a nice-to-have. It is how a stranger becomes someone you will meet in a park.
A reminder that you are going. Confirmations, changes, a last-minute open slot. The community side lives here: pickup games, not a content feed.
Events you can actually join. Local sessions with spots left. The mechanic that makes those work is not the map.

One mechanic deserves its own paragraph: the commitment deposit. The plague of games between strangers is the no-show, the missing fifth player who kills the match for everyone who came. So joining an open spot requires a small refundable deposit. Show up and it comes straight back. Ghost the game and it goes to the players you left hanging, a small consolation with a strong incentive attached.

And for genuine emergencies, mercy stays human. Three attendees can jointly approve an exceptional refund. Simple, self-policing, and it turns “I’ll try to come” into “I’ll be there.”

I would not launch with voice commands, a swipe-deck of coaches, or a merchandise shop. Those are how a booking tool becomes a fitness brand. Get ten athletes paid first.

Monetization

A commission on each booked session is the engine. It scales with use and does not ask anyone to pay rent before they have trained.

A membership can unlock priority with the coaches who fill up, and a cleaner path into the better events. Exclusive tournaments can come later, as a community thing that also pays, not as a media product.

Sponsored listings and branded merch can wait. If the take-rate starts to feel like a tax on a park session, you have lost the part I care about.

Next Steps: How I’d Build It

Do not start with the app. Start with ten athletes in one city and a shared calendar. Recruit them by hand, help them fill their first sessions, and learn what actually blocks a booking. The software (profiles, map, Stripe, reviews) only deserves to be built once those ten athletes are earning.

Then onboard both sides deliberately: coaches get trained on the platform and the community rules, users get a short tour of how to find, book, and rate a session. Launch city by city, not everywhere at once. The first region is your feedback machine.

Three things will be harder than the app itself. Verifying that coaches are who they claim to be takes real work, not just a badge on a profile. Insurance and liability need an answer before the first session. If someone gets injured training with a coach they found on your platform, who is responsible? And paid sports coaching is regulated differently from country to country. In some places it legally requires a certified diploma, so the groundwork has to be done market by market. None of this kills the idea, but none of it can be skipped.

The pieces already exist: a booking flow, a map, Stripe, a review system. I would write the liability answer before I write the map. A pretty booking flow with no insurance story is how you meet a lawyer. The first city also tells you whether coaches will accept a deposit culture, or whether they would rather keep no-shows as “part of the sport.” If they will not enforce the deposit, the pickup games will die the way pickup games always die.

The hard parts are trust and safety, not the tech.

Bonus: strangers, parks, and harm

A platform that puts strangers in a park for physical training will attract the obvious misuse. Clear rules against harassment, a report button that reaches a person, and verification that is real enough to slow a fake coach down. Monitoring slogans will not do that job. The deposit already does part of it for no-shows. The rest is human moderation, budgeted before launch.

Closing

PlayPro Connect uses a map to put athletes who need the income in the same park as people who want a coach who has actually competed.

This is not a generic fitness app with a coach skin, and it is not a dating app for sport. If the first version cannot get ten real sessions on a calendar in one city, the map was decoration. I would rather that ugly calendar than a beautiful empty feed.

I would put the deposit on the first pickup game, not as a later optimization. If people refuse to pay a few euros to hold a spot, you have learned something cheap. If they pay and show up, the community I care about has started, even before the map looks like a product.

If you take it further, or figure out the reason it does not work, I would love to hear about it. Build it, you owe me nothing, just tell me how it goes.

Roméo Rocks: Throw a Stone, Grow a Romance

Wed, 01 Feb 2012 00:00:00 GMT

I still remember that summer evening. I was 15, and I tossed a small pebble against the window of my girlfriend’s family home. My heart raced as I watched from the shadows, hoping I would not break the glass.

She came to the window and smiled when she saw me waving below. Later she told me she had kept that very pebble as a memento of our young romance. I have never forgotten it.

I did not know yet that the pebble would plant seeds of its own.

Recently I was telling this story to a friend, and we laughed about his own attempt, which ended with a slightly cracked window. It got me thinking about how hard it is to find the right stone in the moment. That’s when the idea for Roméo Rocks was born. It’s a playful way to declare your affection for a loved one while putting smiles on faces and roots in the ground (or in a pot to nourish indoors). Now your thrown token of love will also blossom into colorful flowers and keep the sentiment in full bloom for all to enjoy.

Each Rock is crafted from a biodegradable casing and infused with delicate wildflower seeds that will burst into a rainbow of hues. Oval-shaped and smooth to the touch, it bears subtle ridges like ripples on a lake’s surface.

There is also a companion booklet in the plan, with tips and success stories to guide romantic missives in an eco-conscious spirit. More on that below.

(PS: I do not feel sorry about the amount of flower puns in this article)

Market & Opportunities

The gift market keeps growing, and personalized gifts are one of its strongest corners. People increasingly want presents that feel chosen rather than generic, and more and more of them also care about what those presents do to the planet. Roméo Rocks sits right at that intersection. It is personal (the engraving), sentimental (the story behind it), and sustainable (it literally turns into flowers).

Floral gifts are another durable favorite. Roméo Rocks offers a twist on that category: instead of giving flowers, you give the act of growing them.

So in fact, there is a market for such a crazy idea. Nostalgia, personalization and sustainability are all pulling in the same direction, and Roméo Rocks happens to sit where they meet.

Main Features (kind of)

Here is what a Roméo Rock actually does, in plain words this time.

Seed-infused. Each pebble is packed with wildflower seeds. Plant it (garden, pot, wherever) and the gesture literally grows into flowers. Different seed mixes for different climates.

Made to be thrown safely. Smooth, light, biodegradable, so the classic pebble-at-the-window move leaves no broken glass and no litter, just, eventually, flowers.

Fully biodegradable. The casing breaks down in soil within weeks. The pebble disappears, but what it stood for doesn’t.

Engravable. A name, a date, a short message. Every pebble can carry its own little note before it carries seeds.

Looks and feels like the real thing. Oval, smooth, with subtle ridges like ripples on a lake. It sits in the hand like a proper skipping stone, which is the point.

Nostalgia, sentiment and sustainability, all in one small object you can throw at a window without regret.

Companion Book

The Art of Roméo Rock-Throwing: A Romantic’s Guide to Grand Gestures. A short illustrated booklet (16-20 pages, A5), in the spirit of a retro-futurist game manual. To be clear, that means the spirit and not the assets. The art style of a published game is its studio’s property, so this has to be an original illustration job rather than a Fallout or Bioshock pastiche you print and sell. A few entries from the table of contents:

Safety First! (a page about not breaking windows or hurting anyone)
Getting the Technique Right: a step-by-step illustrated throwing guide
Timing the Gesture: distance, angles, and picking your moment
Mother Earth’s Love: how the seeds and the biodegradable casing work
Real-life Success Stories
Conclusion: Now, Go Forth and Love

Plus small “Did You Know?” sidebars throughout, and QR codes linking to video tutorials.

Next Steps

I would start with the object itself: a prototype that actually works. A casing that is genuinely biodegradable, holds the seeds, feels like a real pebble, and survives being thrown at a window without breaking the glass or itself. Until that exists, everything else (branding, shop, booklet) is decoration.

Then come the real walls, and they are not small:

Seed regulations. Wildflower mixes are regulated country by country. A species that is harmless in one place can be invasive in another, and shipping seeds across borders runs into import and export rules fast. Realistically this starts as a one-country product with local seed mixes.
The material. “Biodegradable” and “survives a throw” pull in opposite directions. Finding a casing that does both, at a price a small gift can carry, is the core engineering problem, and it may simply not have a good answer.
Safety testing. Whatever the branding says, a pebble you throw is a projectile. It needs real impact testing on glass and on people, honest warnings, and probably an insurance conversation before the first sale.
Small-batch costs. Molds, engraving, seed handling: none of it is cheap at low volume. The first run will have ugly margins, so a small crowdfunding campaign is less about buzz and more about paying for that run while proving anyone wants one.

If the prototype holds up and the seed rules are clear in one country, then it is worth spending on the brand and the booklet. Not before. First the seed, then the garden.

Bonus: how to keep this from going wrong

One serious note under all the fun. This is for people who want to hear from you (a partner, a friend, your own garden). Not for throwing things at a stranger’s window, or anyone who hasn’t invited the gesture. “Choose your window” means the window of someone who’ll be happy to see you under it, nothing else. When in doubt, hand it over in person.

The practical version of that is short: say it on the packaging, in plain words, next to the throwing instructions. A pebble sold as a romantic gesture and a pebble thrown at a stranger’s window are the same object, and the box is the only place you get to explain the difference. A warning label and a way to report a misuse are worth more than a page of community guidelines nobody reads.

Bonus 2: product variations and line extensions

The core pebble is the star, but a few variations almost design themselves:

Wedding Wishes: for guests to toss instead of rice.
Bee’s Delight: a seed mix designed to support local bee populations.
Boulder of Love: a bigger rock with more seeds, for grander declarations.
Juliet Jars: decorative, eco-friendly containers for catching and displaying Roméo Rocks.
Edible Blooms: edible flower seeds for the culinary romantic.
Love Story Collection: a set of rocks designed to be thrown over time, telling a story as they grow.

Each variation keeps the core values of romance, eco-friendliness and personalization, and gives people a reason to come back.

One Last Toss

Roméo Rocks is a pebble you can throw at a window without regret, one that turns into flowers instead of litter. It will not change the world. But it might make a few evenings under a few windows more memorable, and leave something growing where the gesture landed.

If you make Roméo Rocks real, tell me. I will throw the first pebble (at a consenting window).