BLOG
Notes on building AI agents, physical systems, and the tools that connect ideas to something real.
Selected notes
Building an Agent That Learns From Its Work
OpenKyrozen is an experiment in making an AI agent improve through the work it already performs.
Teaching a Card Game Agent to Think in States
Kards AI turns a complex card game into a deterministic environment for simulation, self-play, and reinforcement-learning research.
Giving AI Agents a Safe Path to Real Hardware
Hardware Agent Runtime connects external coding agents to embedded devices through observable, verifiable hardware-in-the-loop workflows.
All field notes
2026
Building an Agent That Learns From Its Work
OpenKyrozen is an experiment in making an AI agent improve through the work it already performs.
Teaching a Card Game Agent to Think in States
Kards AI turns a complex card game into a deterministic environment for simulation, self-play, and reinforcement-learning research.
Giving AI Agents a Safe Path to Real Hardware
Hardware Agent Runtime connects external coding agents to embedded devices through observable, verifiable hardware-in-the-loop workflows.