AI reviewing AI
Code generation is quickly becoming the default way we write software. We generate code, submit it for review. On the other end, the reviewer asks an agent to review it and posts comments. Then the author asks Claude to fix the PR and re-submit. It works… but for how long? If we continue down this path, we risk becoming increasingly detached from what’s actually happening in our codebases. That gap might not matter — until it suddenly does. ...