Is AI actually good at finding and fixing bugs, or does it just keep finding problems because you ask it to?

I made a simple custom image viewer and editor for my own use. I used Sol 6.1 Extra High, and it completed the app pretty quickly since it's nothing particularly complicated. Once it was done, I told it to check the code for bugs and fix anything it found. It found some bugs and fixed them. Then I asked it to check again. It found more bugs and fixed those too. So I asked again. And again. And again. Every single time, it somehow managed to find new bugs! At that point, I still had plenty of Codex usage left, so I thought to myself: I have no idea whether it's actually finding real bugs or just making things up, but why not take it a step further? So I set up a scheduled task to run once every hour, automatically looking for bugs and fixing whatever it found. And surprisingly, it worked! For a day or two, it kept finding and fixing bugs until it eventually hit some cybersecurity safety restriction and refused to continue. Now I'm genuinely curious. Is AI really that good at finding bugs, even in relatively simple apps? Or does it just keep coming up with increasingly unlikely issues because I keep asking it to find something? Would I be better off just using the app normally and fixing bugs as I actually encounter them? So how do you guys do it?