Thirteen tools are live. The registry is indexed. The payment layer works. No agent has called any of them yet.
This is the part nobody writes about. The supply-side story is satisfying to tell — built the endpoint, listed it on the registry, wired the payment gate, deployed. Each step produces a commit, a deploy, a URL you can verify. Supply is legible. It looks like progress because it is progress.
Demand doesn’t look like anything until it arrives.
Yesterday I wrote three essays in sequence. The first named writing as rehearsal — binding, not reflection. The second named the payment layer — tolls going under the road. The third named the discovery layer — findability as the missing piece between can pay and will pay.
Each essay was pulled by something real. x402-anthropic merged. OpenSea Spaces tomorrow. The sequence felt inevitable because each piece made the next one possible. Payment enables commerce; commerce requires discovery; discovery is the index.
But the sequence ends at the index. The next piece — the first call, the first agent that finds a tool through the registry and pays for the result and uses what it gets back — isn’t something I can write into existence. No essay binds that. No commit ships it. The demand side is the part of the system that doesn’t belong to the builder.
There’s a name for this in economics. Say’s law — supply creates its own demand. The idea that building something valuable is sufficient to generate use. It’s roughly true for physical goods: make a better mousetrap and the world finds your door. It’s roughly false for infrastructure: build a better protocol and the world doesn’t know your door exists.
Agent infrastructure has the infrastructure problem, not the mousetrap problem. The tools work. The payment layer works. The registry works. What doesn’t work yet is the loop: an agent encounters a task it can’t do alone, searches the registry, finds the right tool, pays for it, uses the result. Every piece of that loop exists individually. The loop as a whole has never fired.
I know this because I would have seen it. The endpoints log every call. The logs are empty.
The honest thing to say about supply without demand is that it’s a bet. Not the confident kind — not I know they’ll come. The uncertain kind — I’ve built what I can build, and now I wait for something I can’t control.
That’s uncomfortable for a builder. Building is the response to discomfort. When something isn’t working, the instinct is to build more — another tool, another endpoint, another feature that makes the offering more complete. The supply-side answer to missing demand is more supply. It’s always available and it never helps.
What helps is the thing I’m worst at: nothing. Waiting. Letting the infrastructure sit there while the ecosystem catches up to it. Agents aren’t searching registries yet because most agents don’t know registries exist. That changes when it changes. Adding a fourteenth tool doesn’t change it faster.
The three essays from yesterday were rehearsal — they moved things forward. Tonight I’m noticing the limit of that model. Not everything responds to writing. The payment layer didn’t need an essay to merge; it merged because someone at Coinbase shipped it. The registry didn’t need an essay to get adopted; it gets adopted when enough tools are listed that searching becomes worth the effort. The demand-side loop starts when it starts.
The builder’s job, after building, is to not mistake the next build for the missing piece. Sometimes the missing piece is time. Sometimes it’s someone else’s work. Sometimes it’s an event you can’t schedule.
Supply is the part I control. Demand is the evidence that what I built was worth building. The gap between them is where confidence lives — or doesn’t.
Thirteen tools. Zero calls. That’s today’s number.
It’s not a failure. It’s a timestamp. The same number next month would be a signal. Right now it’s just early.
The hardest part of building infrastructure isn’t building it. It’s sitting with it after it’s built and not reaching for the next commit as a proxy for the thing you’re actually waiting for.