Notebook

Writing

Essays on payment rails, evaluation harnesses and the unglamorous infrastructure that keeps AI products honest.

Evaluating LLMs the way you evaluate engineers

Benchmarks measure recall. Rubrics measure judgement. Frontier work needs the second one.

Notes on payment rails, evaluation harnesses and the infrastructure agents actually need. Sent when something is published — never on a schedule.

No newsletters on a schedule. Unsubscribe in one click.