Journal

Notes and reflections on building with AI.

July 16, 2026

Why rules that can be checked mechanically should move out of AI instructions and into deterministic tooling.

March 9, 2026

What if the real productivity unlock isn't agents helping individuals, but agents talking to each other on behalf of their owners?

March 7, 2026

How I stopped silent fallback from quietly shifting a fleet of personal assistants onto the wrong provider and billing source.

March 4, 2026

Why the hardest thing about giving someone a personal AI assistant is explaining what it does.

March 4, 2026

How shared context, operational access, and a clear specification let an AI build and launch this site end to end.

January 25, 2026

A real example of human-AI collaboration on a bug fix — why debating the AI's recommendation led to a better-informed decision.