Ever since the AI stuff started rolling around on coding i've seen MORE documentation, theres a big incentive to properly document your API endpoints so LLMs can figure it out from specs, and even when not documented the llms can also just read the code and figure it out directly (for libraries and similar). And at least in my experience they tend to document or write it down for future sessions too!

I know you're being facetious but there may well be docs. It's just that the same AI most likely wrote _them_, too.

Did anyone (person or competing LLM) bother to verify that they're correct, though? Who knows! Let the next generation of models worry about that.

I've heard this is now most of some CS jobs now. Just writing documentation for AI.

on the contrary, theres more of an incentive for apis to have docs for agent discovery. the docs / interfaces themselves can be auto-gened (stainless / mintlify)