Since I consider this website my personal notebook, I try to minimize the usage of AI. I don't have any interest in pretending that tokens produced by LLM are my words and my thinking; I believe writing1 Using my own biological neurons! is thinking, hence I think there's no benefit of delegating the actual thinking process to AI.

But AI is still very helpful and I frequently use it while building this website. Almost every line of code on this website's infrastructure (source) is written using Claude. It also maintains the test page.

For transparency, I mention any usage of AI. For example, test page includes Claude Opus 5.5 (as of 14 October 2026) as one of the authors. I also keep Claude Code's git attribution enabled, so that it is clear which commits are co-authored by Claude or any other AI.