Building with AI, making sense of what comes next

I explore AI by building with it and trying to understand it. The work here shares one question: once a system can show its work, how do you check that the showing is honest?

October 2026

  • Running an AI-Built Project Nablaframe

    When AI agents write the code, the hard part is knowing it is right. Notes from Nablaframe on contracts before code, tests that are themselves tested, a decision log, independent review, and one machine whose reports are the only results.

  • Programming in a Glass House Glasshouse

    Drag a house on the screen and the program that drew it rewrites itself, honestly. What Bret Victor demonstrated, what he left open, why we built a new language to make it real — and what a first step can and cannot establish. Includes a live demo you can drag.