The AI Fixed My Bug but Created Two New Ones
by•
I recently asked an AI coding assistant to fix a small issue in my project. It found the problem quickly and suggested a solution that looked perfectly reasonable.
But after applying the fix, two unrelated parts of the application stopped working.
I spent more time understanding the side effects than I would have spent fixing the original bug myself.
AI coding is getting impressive, but I still feel we need to be careful about trusting every suggested change.
How do you review AI generated code before accepting it?
3 views
Replies
The review that catches the most for me is not reading the fix, it is reading the diff for things nobody asked for. Side effects usually live in a file the fix had no reason to touch. A concrete one from my repo, where several agent sessions work in parallel: a bare commit picked up a rename and a deletion another session had staged, and the main branch would not build until the next commit put them back. Since then every commit names its files explicitly. Plus a simple rule: if a change could plausibly not compile, run the analyzer, and if a feature lands, actually run it before calling it done.