I build things, break most of them, and write down what I learn.

I've never been able to sit still on one stack. I caught Rails early, when the Czech ecosystem was three blog posts and a prayer. I caught React Native early and shipped OSS bridges, one still gets downloaded. I went four years deep on a Solana protocol. Now the whole studio points at agents. Four bets, each made before it was safe.

There are two tracks here, and they run on opposite rules. The commitments run long and finish: eight years as CTO at Foundum, five years to ship one game on Steam, four years at Access from founding week to sabbatical. The experiments run cheap and die young, each with a written post-mortem. That split is deliberate. Killing small things fast is what protects the years I give to the big ones.

So the wiring is non-linear, but the method never changes: spec it, loop it, verify it, write down what it taught. Domains rotate; the discipline doesn't. It's also why the arc keeps catching waves early: you see the next one by standing in several fields at once.

I think the UI is splitting in two (task screens get generated and thrown away; the trust layer is what's worth owning) and whoever keeps building UI-heavy apps is going to regret it. So I stopped waiting and started building the plumbing for the agent era: memory, sandboxes, reasoning, routing. I don't write code by hand anymore; I direct the loop in English and own the system that catches wrong.

To be clear, the bet is about task screens in apps, not about websites: a portfolio is a trust surface and deserves design. What this site dogfoods is the split itself: a crafted face for people, an agent-readable twin at /llms.txt for the screening agents.

One human note before the receipts below: the studio isn't called Backtomeaning by accident. The sabbatical is for remembering why I build at all, not just what ships next.