From App Factories to a Reasoning Compiler
I didn’t plan to build a compiler — I just wanted to maximize out of the AI agents I had.
Sergei Belousov
Personal website and blog
Latest posts
Launching the Shuttle with 1970 Code: How We Revived a Forgotten NASA Program
T+0.0. ALT 10. VEL 0. GAM 90.0. Vertical launch. Cape Canaveral, Pad 39. Computer: not a UNIVAC 1100.
Environment Over Intelligence: How a 27B Model on a Laptop Builds Mobile Apps
Late at night. The room is dark. A phone simulator loads on my laptop screen — running a habit tracker. With bar charts for daily stats, pie charts for categories, a calendar with colored dots for each habit, and streak counting. An hour earlier none of this existed. Just a terminal with a blinking cursor and a single line: /build habtrack "Habit tracker with daily check-ins, streak counting,…
Teaching an Agent to Improve Its Own Scaffolding Between Tasks
Most of my time I spend vibe-coding on small projects — it works great there. Drop in some context, get code back, tweak it, ship it. But at some point I got curious: how does this hold up on a genuinely large codebase with serious legacy baggage?
How I Built a Junior QA Engineer on Top of Claude Code
I’m building Etnamute — a system where AI agents write mobile apps from idea to App Store. User interview, market research, spec, design, code, marketing — all automated. My role came down to one thing: open the finished app and check if it works.
I Built an AI Mobile Developer. Here’s What It Actually Does.
I’ve been experimenting with automating mobile app development end-to-end. Not the “type a prompt and pray” kind — a structured pipeline where AI interviews you, researches the market, writes a spec, builds the app, checks its own work, and deploys to the App Store.