// personal computing environment · v2026.07
Ladi. 18 years shipping. Builder, tinkerer, ex-CTO. I ship some, I shelve most, I learn from all. Now betting on agents.
/ the bets
Public bets on the agent era. Where there's an essay, the full argument is one click away.
/ the human
These days I don't write code by hand. I write everything in English, directing agents. What I produce is the orchestration layer: the spec, the constraints, the verification that catches wrong before it ships. The machines type; the judgment stays mine.
What I'm after now is one level up: machines that build machines, systems that find the problems, draft the products, and leave me the two jobs that matter, taste and verification. Everything on this site is that thesis, dogfooded.
I build in the open.Most of it never leaves the workbench.That's where the taste comes from.
/ four waves, caught early
Early Rails adopter in Czechia when the ecosystem was three blog posts and a prayer. Learned that shipping beats waiting for the docs.
ruby · convention-over-configBet on cross-platform before it was obvious. Shipped OSS bridges, one hit 121 stars. Distribution taught more than any framework did.
react-native · objective-c · javaFour years on Access Protocol, founding week to sabbatical: the platform, design system and SDKs around a Solana protocol, then the AI-agents pivot from inside. Not a tourist.
typescript · go · solanaThe whole studio turned toward agents: LoomBrain, sandcaster, plugins, reasoning. Building the plumbing while the interface layer sorts itself out.
typescript · claude · mcp/ how I build
Every post-mortem on this site taught me the same lesson from a different angle: big leaps carry your old mistakes with them; small verified loops surface them while they're still cheap. This is the machine every project runs on now. Scroll to assemble it.
/ the lab · field notes
A machine that finds problems I should work on. It found two shipped products; it also taught me the real bottleneck is elsewhere.
A psychometric interviewer you talk to. Paused on a hard voice-vs-transcript problem.
Idea to paid Chrome product in four days. Nobody paid: the ICP could not operate the delivery mechanism. Best tuition I bought all year.
A writing assistant that learns your voice through corrections. Killed with a clear head: the loop demands too much from the human.
A focus coach that talks you back to work. Cool product; it never helped me, so I shelved it.
Rebuilding my brother's ISP back-office from interviews and database archaeology. Active this month.
Teaching agents to buy things and mail physical letters: Medusa adapters, QR payments, Česká pošta API.
Four paper cuts, four tiny apps: audio presets, icon hiding, meeting audio, ambient music. "Too small to build" is a dying category.
Export X articles to Markdown for my AI. 500+ strangers turned out to want the same thing.
A tiny falling-blocks machine, because a workshop should have at least one toy. Click to play: arrows move, space drops.
Most of what I build never launches, and that's the point. Shipped, shelved, or still learning, it all gets logged. See the whole lab →
/ selected work
Every project has its reasons written down: what it is, why, what happened, what it taught me. All projects, with the post-mortems →
/ writing
Written the way I build: I direct in English, my agents type. The opinions, the learnings, and the scars are mine.
The Book of Aaru took five years, one full restart, and a small studio to ship. I already wrote the goodbye; this is the build: where the years actually went, why restarting fixed the assets but not the decisions, how the pace of AI made me rethink the whole path, and the 1%-loop lesson I'd run any project on now.
read →logLocal-first is the sacred cow of second brains: files on your disk, plain text, outlives every app. I think it's the wrong architecture for the agent era, and I'm betting LoomBrain on the opposite: a brain that lives on the internet so agents can reach it while my computer sleeps. Build log: three ground-up rebuilds in four months, the WHY-annotation idea, and an empty user count.
read →essayThe comforting line is that taste is the one thing AI can't take from you. But nobody can define it, teach it, or prove who has it, and stripped of the mystique, taste is just an evaluation function: what tells a feedback loop better from worse. We already built the machines that learn one. The edge was never having taste. It's running the loop.
read →/ ask sam · live
Sam is live now: a real-time video avatar with my face and a voice cloned from mine, an AI agent grounded in my notes. He answers in the first person, as me — and if you ask whether he's really me, he'll tell you the truth.
Talk to Sam live →
