Pre-LLM, it was much easier for reviewers to discern that. Now, the AI-generated code can look like it was well thought out by somebody competent, when it wasn't.
replyHave you ever reviewed an AI-generated commit from someone with insufficient competence that was more compelling than their work would be if it was done unassisted? In my experience it’s exactly the opposite. AI-generation aggravates existing blindspots. This is because, excluding malicious incompetence, devs will generally try to understand what they’re doing if they’re doing it without AI
replyI think the issue is not that the patches are more compelling but that they're significantly larger and more frequent
replyI try to understand what the llm is doing when it generates code. I understand that I'm still responsible for the code I commit even if it's llm generated so I may as well own it.
replyI have. It's always more compelling in a web diff. These guys are the first coworkers for which it became absolutely necessary for me to review their work by pulling down all their code and inspecting every line myself in the context of the full codebase.
replyYes and if they copy and paste code they don’t understand then they should disclose that in the commit message too!
reply