Post

Atila
@atila.io

I would've liked the disclaimer if at least they had tested it themselves - they could've used AI to create them a test scenario even the "potentially fixing" above the disclaimer is weird, sounds like they're just trying to bait you into doing the work

Atila
@atila.io

yeah, frankly i’m fine with vibecoding and the occasional fumble from the AI. I’m not ok with “here’s my untested code, figure it out” - ai or not

Ragura
@ragura.bsky.social

Yes, at that point just start automating PRs the moment an issue is created. Maybe one in X will be good (and that percentage could go up as the tech improves), but it will take vastly more time to check than to write it manually (for people like Daniel at least).

Jonathan Schneider
@jonathanschndr.de

I also think it's good that there is a comment about this. At least that makes it transparent. But I fundamentally don't think it's a good idea to try to solve issues with AI code without checking the code yourself, but instead demanding that someone else look at it.

Adam Thompson
@thesonofthomp.com

Genuinely curious how you handle that though. Appreciate the disclaimer, but are you really gonna spend the hour+ to determine whether their code works? And if it doesn’t, how do you provide that feedback since it doesn’t seem like they really understand their code in the first place?

daniel r(ooooooh) in nyc 🇺🇸
@danielroe.dev

in this case I checked out the pr locally, looked at the test and determined it was incorrect then I reverted the pr and logged to see if it did anything, and verified that it did not in fact do anything I have now asked for a reproduction

daniel r(ooooooh) in nyc 🇺🇸
@danielroe.dev

I would far rather have someone try to fix something at worst case, it's a bug report at best, it's the gateway to someone becoming a valued contributor! (and the disclaimer signals that the person isn't trying to impress - humility is very important in oss!)

Doug Parker
@develwithoutacause.dwac.dev

I'm not sure I feel the same way. I can appreciate the honesty, but if it's "blindly authored" by Claude with no human understanding or validation of what it did, then I'm not sure this would provide enough value to be worth taking a reviewer's time on a (presumably unfamiliar) OSS project.