For work I often have to review multi thousand file changes in PR mostly generated by AI/Llama/claude code. It's mainly rust with many unit tests, so there is a lot tooling helps verify the code already to help a lot. But I struggle with these reviews, finding me to just glance over at times unable to comprehend the changes in any higher level capacity (e.g. architecture). Which I think is mainly due to so many changes of these thousands of lines happening that I lack any coherent model of what things look like in the first place. And the LLM generated descriptive commit messages or the PR description and documentation of what they do or even graphite doesn't really help. I wouldn't even be sure I could verify over thousands of lines in multiple commits whether these changes end up doing the architecture described. So, I wonder, how do other people experience this? How do you review all that code that these generated and what are you looking for? Especially when CI/tooling takes care of all the lints and compile problems? Or do you just yolo that code now, accepting the occasional hallucination?
Comments URL: [https://news.ycombinator.com/item?id=48538778](https://news.ycombinator.com/item?id=48538778)
Points: 2