AI transparency
Some posts on this site are drafted with help from Claude (Anthropic's large language model). Where that happened, the post header shows a small AI-assisted pill that links back to this page.
What that means in practice:
- The argument, the angle, and the opinion are mine before any LLM is involved.
- Claude helps draft prose, find candidate citations, and tighten paragraphs that read awkwardly.
- I read every cited source. If I cannot verify a claim, I remove it from the post.
- I edit the final prose by hand, including the parts the model drafted. The opinions, errors, and bad jokes are mine.
Posts without the AI-assisted pill are written by me without significant LLM involvement.
Why bother saying so? Because opinion writing is a different game from generated text, and conflating the two is bad for both. The IEEE, ACM, and most serious publications now require disclosure when generative AI plays a substantial role in a piece. This is the lightest-weight version of that policy I can think of that still respects the reader.
Corrections, disagreements, or "you got this wrong" emails: mart@traagel.dev.