How one engineer ships all of this — safely.

Every project on this site was built the same way, with a method I authored and maintain: an operating system for building software with AI agents.

The rules are simple and unforgiving. Work lands in small, safe, independently mergeable changes. A plan maps every change to the requirement it serves, and a lint gate reads the plan like a compiler — stale claims fail the build. Every decision gets a written record. AI agents build in parallel lanes without stepping on each other, and fresh-context agents review work they didn't write. A green gate — not a vibe — decides "done."

And the most important rule of all: nothing deploys itself. Every deploy in every project is a human action with a typed confirmation. AI writes a great deal of the code here. It never decides what ships.

This is AI-assisted development for professionals, not vibe coding. It's how a practice of one maintains sixteen active repositories — enterprise systems, live commerce, event platforms, generative art — without dropping any of them. It's also the thing clients most often actually need: not a model, a method. If your team is trying to adopt AI without giving up engineering discipline, this page is the service.