3 ms·
> > LLM’s never provide code that pass my sniff test > This is ego speaking. That's been my experience of LLM-generated code that people have submitted to ope
by nicoburns 2y ago
> > LLM’s never provide code that pass my sniff test
> This is ego speaking.
That's been my experience of LLM-generated code that people have submitted to open source projects I work on. It's all been crap. Some of it didn't even compile. Some of it changed comments that were previously correct to say something untrue. I've yet to see a single PR that implemented something useful.
- constantcrying 2y agoTerrible software engineers sent you terrible code. A good software engineer will obviously send you good quality code, whether he used AI for that is impossible to tell on your end.
- mh- 2y agoIsn't this a kind of survivor bias? You wouldn't know if you approved (undisclosed) LLM-generated code that was good..
- talldayo 2y agoGood LLM-generated code comes from good developers that can provide detailed requests, filter through wrong/misleading/cruft results, and test a final working copy. If your personal bar for PRs is high enough then it won't matter if you use AI or not because you won't contribute shitty results. I think the problem has become people with no bar for quality submitting PRs that ruin a repo, and then getting mad when it's not merged because ChatGPT reviewed the patch already.
- cies 2y ago> LLM-generated code that people have submitted to open source projects I work on Are you sure it was people? Maybe the AI learned how to make PRs, or is learning how to do so by using your project as a test bed.