Some things are complex.
You could've been curious and ask why it would take 1-2 days, and I would've happily told you.
I wanted to add audit logging for all endpoints we call, all places we call the DB, etc. across areas I haven't touched before. It would have taken me a while to track down all of the touchpoints.
Granted, I am not 100% certain that Claude didn't miss anything. I feel fairly confident that it is correct given that I had it research upfront, had multiple agents review, and it made the correct changes in the areas that I knew.
Also I'm realizing I didn't mention it included an API + UI for viewing events w/ pretty deltas
Besides, do you think that a LLM would be better at deciding what to log than a human that has even just a little experience with the actual system in question?