Client work is half of what I do. The other half is building and running my own products — which is where I prove the solo + AI model on real users, real data, and real production constraints, not slides.

Vigilo (vigilo.cc) is a global risk monitor I designed and built solo. It aggregates 45 live sources across 239 countries into a transparent risk score — disease, conflict, natural disasters, civil unrest — with every signal traced back to its origin. The honest origin story: I tried to predict outbreaks with AI, failed twice, and published the failed experiments instead of hiding them. Live in production with billing, alerts and a Telegram bot.

Drakoty is an original creative IP I write and build — a dragon-world universe spanning a book, a lore bible, a playable web game and a free illustrated audiobook app in four languages (case study), all shipped solo with AI tooling. It's the project that shows range: the same solo + AI workflow that ships a data-heavy SaaS also ships an interactive narrative experience, end to end.

The same workflow goes to client work. EngiBoard — a native desktop SaaS for engineers (Tauri 2 + Next.js + Supabase, four platforms) — went from an empty repository to production in four weeks on a client contract, now on ongoing support: 44 releases, 98 commits. It's the clearest proof of how fast one person plus AI agents can ship genuine production software for a client.