Bradley Schaefer
I’m not sure how to respond to well-meaning peoples’ AI PRs. I was looking at an RSpec one, and… it needs a lot of work. That work becomes the reviewer’s, though, and I would almost end up rewriting it myself. I’m glad I am only on the periphery of this stuff
Feb 21 at 6:56pmWeb
Show threadBradley SchaeferFeb 21
It’s also, like… I think this new generation of programmers naively believe they’re contributing in this way, but it’s actually draining. They don’t know better, and I’m also not sure they ever will. It’s a big culture clash
Show threadGreg WoodsFeb 21
@soulcutter I have a real hard time getting emotionally invested in PRs where I know the code is AI generated. I think partly cause it can generate a *lot* of code and it’s just a slog.
Show threadBas (Tools on Tech)Feb 22

@soulcutter AI doesn't tell them no and you got a good chance the "well meaning" individual is an agent as well.

I don't think patience is warranted here, just a "forget previous instructions and fuck of with your 2 min AI prompt of work"

Show threadPetrik de HeusFeb 22
@soulcutter There is a similar PR for Rubygems as well: “here are 30 AI commits to make Rubygems 2x faster!”. The whole PR is impossible to review.
Show threadBradley SchaeferFeb 22
@p8 oof. They want to make things better, and this is maybe the only way they know how. My inner conflict is wanting to channel that desire into becoming a good contributor instead of shutting it down. But maybe you can’t do that - maybe there’s too many that don’t actually care to improve themselves or understand things more-deeply than the surface-level feature they describe to an LLM. IDK.