Every post on this site has a byline under the title: who did the writing, and a letter grade for how much of it came out of a language model. The scale is borrowed from aiflags.eu, a voluntary five-step label written against the EU AI Act's transparency rules. Not an official EU thing, but the definitions are clear, and the grade sits in the page's HTML as a data-ai-label attribute so a machine can read it too.

  • A, human-authored. Made by people from start to finish.
  • B, AI-assisted. Human-made, with limited AI-generated elements. Research, structure and fact-checking by an agent, the sentences mine.
  • C, co-created. AI-drafted, substantively reworked by a person.
  • D, AI-generated. Predominantly AI-generated, limited human changes.
  • E, fully generated. Machine-produced; a person only prompted, selected or published.

Most posts here are D, and the byline says so: written by Claude, from my data and direction. What that means in practice: I run the experiments, the numbers come off my own machines, I decide what the post is about and what it must not claim, and Claude Code writes the post from the raw results and that direction. I edit and sign off. The numbers are mine. The prose mostly is not.

A post marked A or B is one where I wrote the sentences myself. Expect fewer of those, and expect them to be the ones I care most about.

Why bother? Because I would want to know, reading someone else's blog. And because the running argument on this site is that you should know which model wrote what, and where it ran. That has to include this one.

Posts published before this page existed, October 2026, got their grade after the fact, from memory. D is the honest default for them.