My process is that I run top Claude model on the small set of changes to reflect it in the docs. I get like 80% from there and then I polish it by hand. Remove bad wordings, change examples, cut whole paraghraphs of text, etc.

I tried different “skills” for writting, but I never got any good results from them :(

score 1 · canonical

Our AI policy is to allow AI-generated code, but only after several rounds of my manual review. My own AI usage is limited to docs, changelogs, some tests for corner cases.

Nothing too crazy.

But, we develop a lot of AI tooling for end users: skills / MCPs / evals / contexts / etc. That’s why these ignores are needed :)

score 1 · canonical