Notes from the field.

Essays and engineering notes on AI, robotics, and production systems — written like white papers, readable like conversations. Keep scrolling; the feed loads as you go.

July 13, 20265 min read

graphmem: one shared brain for every Claude and Codex session I run

Claude Code and Codex are absurdly capable. The mess is everything around them: secrets, deployments, and context scattered across MongoDB, GCP, Vercel, Tailscale, and four machines. So I built a knowledge-graph memory on Oracle's free tier, and now every session everywhere shares one brain.

— End of feed —