The Potter's Wheel Economy

In the last essay I wrote about a strange pattern in Stack Overflow's viewing figures: the biggest crowds gather around things nobody needs anymore. A checksum computed by hand drew over a thousand readers who all own machines that would do it instantly. I reached for the image of a potter's wheel in a shop window, industrial ceramics having won a century ago, and people stopping to watch anyway. I meant it as an observation about how understanding propagates. Then I kept turning it over, and it stopped looking like an observation about programming at all. It is an observation about markets, and the potters have known it forever.

Because notice what the wheel in the window actually is. It is not production; the shop's inventory comes from the kiln in the back, made efficiently, off-stage. The wheel by the glass is marketing. It is the oldest conversion funnel there is: skill exercised openly stops foot traffic, watching builds trust, trust walks inside and buys a bowl the watcher could have bought cheaper elsewhere. Murano glassblowers run it. Open kitchens run it. The fishmonger throwing salmon at Pike Place runs it. The demonstration was never the product. The demonstration is the reason you believe in the product.

What AI just did to every shop window

Now run the Stack Overflow event across the whole economy, because the same repricing is happening everywhere, for the same reason.

For decades, a polished artifact was evidence of competence. A sharp deck meant a serious company. Clean copy meant a real marketing department. A well-structured proposal, a decent website, a professional product photo: each was expensive to produce, so each was a costly signal, and costly signals are the only ones that separate anyone from anyone. That is the economics under all of it. A signal works because faking it is expensive.

AI made the artifacts free. Anyone can now emit the sharp deck, the clean copy, the confident proposal, the fluent thought-leadership post, in minutes, from a prompt. Which means every one of those signals just stopped separating. The polished output that used to whisper "we are competent" now whispers "we have a subscription". Inboxes, feeds, and marketplaces are filling with artifacts of identical, excellent, meaningless polish, the business equivalent of the fact-shaped answers AI already serves faster than any forum. When everyone's window display is perfect, nobody's window display says anything.

What cannot be emitted from a prompt is the wheel: competence exercised live, in public, on a real problem, with the wrong turns showing. The thousand readers of that CRC answer were not collecting a checksum; they were watching someone who knows what they are doing, and calibrating trust. That calibration is the scarce commodity now. The artifact stopped carrying the signal, so the signal moved to the making, exactly where it lives in the pottery shop.

The same move in marketing, sales, and founding

Once you see it as one repricing, the playbooks in three different disciplines turn out to be the same playbook.

In marketing, it is the shift from publishing claims to publishing process. The teardown of your own failed launch. The worklog. The live build. Content that performs understanding rather than dispenses it, to reuse the phrase from the last essay, because dispensed understanding is what the machines now give away. A company blog full of fluent AI prose is a perfect window display on a street of perfect window displays. A founder showing, weekly, how the thing is actually made, decisions, mistakes, reasoning, is a wheel, and wheels still stop traffic.

In sales, it is the death of the deck and the promotion of the demonstration. Nobody is impressed that your proposal reads well; their own proposals read well too, and everyone knows the marginal cost of the words is zero. What still closes is the potter's move: solve a slice of the prospect's actual problem, live, in front of them. Take their data, their codebase, their campaign, and work on it while they watch. Every consultant who has won a client with a free teardown of the client's own situation already knows this motion. AI did not invent it. AI just devalued everything else, which promoted it.

In entrepreneurship, it is building in public, which used to look like an indie quirk and now looks like structural advantage. The founder who works with the window open is simultaneously doing customer research, marketing, hiring, and the slow minting of the one asset AI cannot counterfeit at scale: a crowd that has watched them be competent for a long time. Audience-first businesses were always nice. In an economy where distribution via polished artifact is dead on arrival, the watched making is becoming the only distribution that compounds.

The second crowd at the glass

Be precise about the mechanism, because "attention" is not it. The wheel does not mint attention; plenty of louder things do that. The wheel mints trust, and the crowd that has watched you work will pay for exactly the goods that survived the repricing: judgment applied to their situation, products from someone demonstrably competent, teaching, access. None of that revenue flows from the artifact. All of it flows from having been watched. The bowl was always cheaper elsewhere. People buy it from the potter they watched.

But there is a second buyer forming at the glass, and this is the part I think most people running the playbook have not priced in yet.

Look at what is already thriving on video. A gardener films himself rescuing strangers' overgrown yards, transformation by transformation, and builds an audience most startups would envy. Crews film themselves repairing the roofs of people who cannot afford the repair. Restoration channels, machining channels, a person fixing a chair for an hour. The standard reading is the one this essay has been giving: process is legible, transformation is satisfying, trust compounds. All true. Here is the other reading: this is footage of humans using tools to do physical work, narrated, sequenced, with mistakes and corrections included, accumulating by the petabyte, and it happens to be exactly the thing the next generation of AI cannot get anywhere else.

The text models got their competence from an archive people wrote for other people, Stack Overflow prominently among them, scraped after the fact. The coming generation is embodied: Optimus-class humanoids and their cousins, machines that need to know how a hand chokes up on a trowel, in what order a roof comes apart, what a stuck bolt sounds like when it finally gives, what you do when the thing you tried does not work. That knowledge does not live in text. It lives in demonstrations, and the labs already pay handsomely for staged ones, teleoperation sessions, motion capture, humans in instrumented gloves doing dishes for the camera. Meanwhile the internet's makers are producing the natural version at scale, for free, as marketing.

So the person working with the window open is minting two assets at once, whether they know it or not. Trust, from the humans watching now. And process data, the record of skilled hands meeting real-world mess, which is precisely the space the machines have not filled yet and precisely what they will be trained on next. There is an irony here worth saying plainly: the programmers who wrote Stack Overflow's archive gave away the corpus that automated their own answers, and got nothing but reputation points. The makers filming themselves at the wheel are writing the corpus for the next automation, the physical one. The difference is that this time the value is visible in advance. What that should change is not whether you show the work, but the terms: know that your back catalog of demonstrations is an asset with a second market, keep your rights to it, and price it like something a machine will one day study, because one will.

Use the machine for everything but the wheel

Here is the part that sounds paradoxical and is actually just division of labor: AI is how one person can afford to run this playbook at all.

Everything around the demonstration is artifact work, and artifact work is exactly what the machines now do well. Let them. The clipping of the long session into shorter pieces, the transcript, the newsletter version, the scheduling, the SEO scaffolding, the follow-up emails, the CRM hygiene, the six translations. This site is my own small proof: models translate every essay into five languages and read them aloud in a clone of my voice, a pipeline hand-builds none of it, and I would never pay a human team to do any of that for a personal weblog. The machine multiplies the reach of the demonstration by an order of magnitude. What it must not touch is the demonstration itself. The hour at the wheel, on a real problem, with your actual reasoning showing: that hour is the entire source of value, and it is the one hour the machine frees you to spend.

So the playbook, compressed: pick the work you are genuinely good at, because the wheel is merciless to fakes. Do a real piece of it in public, regularly, wrong turns included. Hand every artifact around it to the machines: packaging, repurposing, distribution, translation. Then sell the crowd the things only you can sell it, judgment, products, teaching, access, and price them on the trust the watching built.

And do not fake the wheel. The temptation is obvious, since AI will happily generate "authentic process content" too, a synthetic worklog, a staged struggle. Audiences are already developing the detector, the same way the Stack Overflow crowd developed one for fact-shaped filler, and the failure mode is not weak engagement. It is that trust, once caught fake, reprices to zero and stays there. The wheel works because it is costly. That is not a bug to optimize away. It is the whole signal.

The last essay ended with the answer written for the eleven hundred watchers rather than the one asker. The business version is the same sentence with the nouns swapped. Work for the watchers. The machines have taken over the artifacts, and they are welcome to them. The crowd at the glass is the market now, and it is growing a second row: behind the people, the machines, learning to see. Both rows are watching for the same thing, the one thing that is still expensive: you, at the wheel, knowing what you are doing.