I feel like one coding benchmark should be just telling it to double check or fix something that's actually perfectly fine repeatedly and watch how bad it deep

I feel like one coding benchmark should be just telling it to double check or fix something that's actually perfectly fine repeatedly and watch how bad it deep fries your code base.