Son Piaz

How these fit together

There are more projects on this site than one person should have. The reason is not ambition. Each one started as a problem inside the one before it, and the shortest way to describe the whole set is as a loop rather than a list.

The loop #

Mandeck the room the agents run in Kyma API one key, a hundred models What gets built watch-cli, Haynoi, Pheme, Kapt Affitor, open-affiliate tracking, terms, payouts Build to Own teaches the whole circuit calls for models returns answers needs distribution becomes the syllabus graduates run their own
Each product answers a question the one before it asked. The blue arrow is what makes it a loop rather than a list.

1. The agents run somewhere. That is Mandeck, a macOS terminal that holds many coding agents at once and lets one of them open a pane for another. Before it existed I was the wire between windows, carrying results by hand.

2. The agents need models. That is Kyma API: one key, one endpoint, a hundred models, and a measured uptime number for each of them. Mandeck made it obvious how fast a single vendor becomes the thing that stops your morning.

3. What gets built has to reach someone. This is the oldest thread here and the one I understand best. open-affiliate publishes affiliate program terms as open data, so a person and an agent read the same record. affiliate-skills hands an agent the actual playbook, one skill per stage. Affitor runs the money side: tracking, commission holds and partner payouts for SaaS brands.

4. Somebody has to know how to run the circuit. That is Build to Own: four weeks, a fixed deadline, and a public demo day, aimed at Vietnamese builders who have watched enough tutorials and shipped nothing.

5. The loop produces side projects. These are the small tools that came out of the four above, and each one is proof the loop works. watch-cli gives an agent eyes and ears for video. Haynoi types what you say in two languages at once. Pheme writes meeting notes without sending a bot into the call. Kapt and Echoly came out of the same loop earlier and are now quiet.

Where the loop is weak #

Step 3 is the weak link, and pretending otherwise would make this page useless. A two sided market is hard to start from the middle, and the open source pieces around it get plenty of attention without converting it: affiliate-skills is the most starred thing I have written and has produced no purchases at all.

Steps 1, 2 and 4 hold. Agents run, models answer, and the first cohort finished with real work handed in. Step 3 gets solved by someone using the other four, not by rebuilding the middleman.

Why this matters if you only read one page #

Every product here is the answer to a question the previous one asked. That is worth more than any of them individually, because it means the tools are tested by each other every day rather than by a demo.