I thought I came across a preprint paper about using LLMs to autogenerate documentation/docstrings and the "accuracy" of those statements relative to person-created ones.
I'm still struggling to understand the intended audience for those comments in an LLM world.