Except ai is actually good at improving performance and finding bugs. Literally. And every time an author denies this I feel like they or I are smoking crack.
Some combination of ego or loss of self seems to make people totally unable to admit that an llm can produce code as good as (i would say better) than them.