Writing
Notes: explanations, interactive work and reading. This is the writing collection; the wider discovery page also includes concepts, examples and animations.
Published here first.
Essays, notes and working papers on measuring how AI systems behave over time. Each piece is published on this site as the canonical version, then syndicated elsewhere. Where a piece has been posted to another platform, the link is listed with it.
- Safety Is Retrospective
A working paper mapping six architectural failures from aviation, oil and gas, nuclear and drug safety onto current AI governance, and proposing measurement-based alternatives.
Shorter commentary, elsewhere.
Long-form work is published here. Shorter commentary and threads appear on the profiles below. Citable outputs, including the peer-reviewed record and public software, are listed under publications and outputs.