Code reviewing was never the most interesting thing to do. But it had one important element. That, if done right, it was knowledge exchange between the reviewer and the coder. That can be quite motivating. Helping a fellow coder to become better. Reviewing "AI" written code does NOT come with that potential reward. The machine doesn't learn the way a human does. This turns code review into a menial, fruitless task that leads to frustration instead. That's my observation and opinion.
1/2
Replying to @jwildeboer@social.wildeboer.net
The second element I have heard a few times now: "AI" written code is not seen as "my" code by coders. They don't feel attached to it in the way they do with code they have written themselves. Hence they are not really "feeling" for it, so they are also not really interested in making it better or defending it in the review cycle. Just tweak the prompt and move on. This is having a real impact on motivation, discussion and results. It is hard to put in metrics, though.
2/2
Replying to @jwildeboer@social.wildeboer.net
@jwildeboer One problemaic thing is that development teams on larger teams do not work as a team anymore. Everyone uses their AI that does random changes over all of the code base - no one is responsible anymore and owners of specific components get too many changes to review by others that did not even bother to review them. It does not work very well when your code base requires working as a team or multiple.
Bonus: LLMs only like their own changes - different LLMs change each others code.
