Two projects, selected deliberately. Together with 13 years of platform work at Mercedes-Benz they carry the argument: enterprise scale on one side, AI-native engineering with externally checkable results on the other.
Persistent multi-agent engineering teams · experimental
A desktop control plane for supervising Claude Code, Codex and Gemini across projects and persistent agent teams: durable roles, context recovery, worker health, and what happens when human attention becomes the limiting resource.
AI-native chess engine · actively developed
A from-scratch UCI engine built primarily through coding agents, independently rated ~3570 Elo on the CCRL Blitz list. Evidence that AI-written code can earn trust the old-fashioned way: by being tested until it proves itself.
Also in daily use
Alongside them runs more than fits on one page. The part I use every day is a connected suite built for myself: taurfuel for nutrition and taurlift for training on the phone, taurscale for body composition on the desktop, and taurinsights, a small analytics backend that joins their data into one picture. They are private, and they are not portfolio pieces. I mention them for one reason: a connected multi-app suite with clean architecture used to be beyond what one person could build on the side. With coding agents, it became a byproduct of daily life.