Comment by wizzledonker

I think the real issue might be that how “good” the code is matters less than being able to form a mental model for what the human who wrote the code was “thinking”. If written by a machine, this contract is broken and we get more confused, even if our traditional methods of evaluating the code come out equal.