Programmers will document for Claude, but not for each other

This article is a short reflection by Mark Dominus on the role of "documentation" in AI-assisted development. The starting point is a complaint he has seen many times: many programmers are now willing to write very detailed CLAUDE.md and PROJECT.md files so that AI agents can understand a project's context, yet in the past they were unwilling to write equally clear documentation for their human colleagues. The author doesn't stop at the irony, but instead turns this phenomenon into an actionable workflow. On larger projects, he has Claude maintain a handoff document recording what was originally intended, what has been completed, and other relevant background; when one Claude session ends and the next Claude takes over, it first reads this document to get up to speed. He later realized that these handoff documents shouldn't be thrown away after a project ends, because if you copy them into the repository and commit them, someone using git grep to find information in the future might land on a useful clue. Going further, he now asks Claude at the end of a project to rewrite a complete but high-level project summary explaining the problem being solved and the changes that were made; this is not a running log, but a structured overview. The author emphasizes that he carefully reviews and edits as needed before committing, because the commit carries his name and the responsibility is his, so the standard should be the same as overseeing a human engineer. The article also warns that AI documentation can still be wrong: Claude once mimicked an old report and copied an approval statement verbatim into a new document. The author's closing advice is very practical: if you're already having Claude write notes, commit the valuable content into the repo; you can also ask Claude to write a project summary and then preserve it after human review. For development teams, the key point is to use AI to lower the cost of producing documentation, but without skipping human understanding and accountability.