WatsonBuilds.ai.
The site you're on, and the lab the rest is built in. One URL holds the whole loop: the audience surface (essays, the operator letter), the product catalog (eighteen tools you can run right now), and the public signal log where every ship and every kill gets recorded in the open. The thesis is plain — most of this was built with AI agents doing the heavy lifting, and the build itself is the marketing. Every tool that goes live is a piece of durable proof-of-work that keeps selling long after the launch post scrolls away. The job is to keep the loop turning on a calm weekly cadence: publish the work, ship the tool, log the decision, repeat. No growth hacks, no theatre — just a relentless metronome that an outsider can audit any day of the week.
Build in public and the products pull the audience, not the other way around.
Audience attention compounds when the operator ships the loop they're inside instead of describing it. A working tool earns more trust than a thread; a letter that admits what got cut earns more subscribers than one that only celebrates. So the bet is to make the body of work — tools, essays, kill log — the funnel itself. The audience arrives for the proof, converts on the products, and the revenue is a lagging read on whether the loop is honest.
- 01.Tools are durable proof-of-work. They survive the launch tweet and keep converting while you sleep.
- 02.A dense, boring weekly cadence beats one viral hit. Compounding, not lightning.
- 03.Audience, product, and the operating system all live on one URL — there's nothing to switch to.
- 04.Building real software with AI agents means the marginal cost of the next tool keeps falling. The catalog can outgrow a solo operator's hands.
Make the catalog the funnel.
If the catalog keeps growing on cadence and a handful of tools cross real, repeated usage, the audience earned by the work will support a paid operator tier without ever running a single growth campaign. The product is the distribution.
- 01.30 days of flat or negative MRR.
- 02.Weekly shipping cadence breaks and stays broken — the metronome is the moat.
- 03.Tools get used once and never again — proof-of-work that nobody returns to isn't proof of anything.
Currently wrestling with.
- 01Does the catalog page convert better with live tool previews or static thumbnails?
- 02Should the operator letter live on a hosted platform or stay self-hosted on /letters?
- 03What's the right paid wedge — per-tool, per-seat, or per-output?