“Sorry, the autogenerated api documentation was wrong because the ai hallucinated the docstring”

You can't read?

Please don't say you commit AI-generated stuff without checking it first?

I don’t commit ai-generated stuff. Do you?

Of course, but not without review.

It’s exactly like working with another human. PR review is there for a purpose.

>writing docstrings and type hints

Disagree. Claude makes the same garbage worthless comments as a Freshman CS student. Things like:

// Frobbing the bazz

res = util.frob(bazz);

Or

// If bif is True here then blorg

if (bif){ blorg; }

Like wow, so insightful

And it will ceaselessly try to auto complete your comments with utter nonsense that is mostly grammatically correct.

The most success I have had is using claude to help with Spring Boot annotations and config processing (Because documentation is just not direct enough IMO) and to rubber duck debug with, where claude just barely edges out the rubber duck.

I intentionally said docstrings instead of comments. Comments by default can be verbose on agents but a line in the AGENTS.md does indeed wrangle modern agents to only comment on high signal code blocks that are not tautological.