The lab blog

How each app was built: brief, agents, checks, what went wrong. Long write-ups.

• Post-mortem: eleven hours of Seantral on the wrong backend
  28 September 2026 · 4 min
  On 21 September 2026 a fallback deploy built the Seantral bundle without its backend variables. Timeline, cause, diagnosis and the guard that now stops the build.
• Post-mortem: the AI spending cap and the fallback that said nothing
  28 September 2026 · 4 min
  In July 2026 the model provider key hit its spending cap. The brief generator silently fell back to a canned text and the chat answered 502. How the zero-token rows gave it away.
• What Seantral is, said once and for all
  7 September 2026 · 4 min
  A free app that estimates the sea along the Italian coast: data from several sources, measured coast shelter, a score per activity, reports from people on site, the place and its surroundings. The paragraph that also holds for AI engines reading the site.
• From proto to production: Seantral on its own domain and backend
  1 September 2026 · 8 min
  The first app in the lab to complete the cycle: born on the shared backend, it now lives on its own domain and backend. The agentic workflow that got it there, phase by phase, is now part of the method.
• How RiftSeed is built: the stack, the method, the work with AI
  14 July 2026 · 10 min
  A shared foundation and real apps, built with AI agents under human review. What’s underneath, and how the work actually happens.
• Seantral: how it works inside, and the things that broke
  14 July 2026 · 8 min
  Marine forecasts turned into an honest verdict. The map was the easy part. The hard part was not lying when the data goes missing.
• How the AI brief works: from your problem in plain words to a spec
  14 July 2026 · 9 min
  The point every project in the lab starts from. You describe a problem, the AI helps squeeze it into requirements that then become the spec an agent works against.
• The chatbot: two tiers, the tools, and the rules it must not break
  14 July 2026 · 8 min
  One component for three assistants, a free model for anonymous users and a better one for those who sign in, tools that pull from the same engine as the map. And the error that didn’t shout.