My diagnosis is that almost nobody knows how to hire programmers. Recruiters don't know how to hire programmers. MBAs certainly don't know how to hire programmers. Even many programmers don't know how to hire programmers. Good programming sense is an art, and it's really hard to evaluate it in someone else without truly grasping it yourself.
So in the absence of real criteria, like a poorly-trained ML model everyone just latches on to superficial signifiers that have worked out in the past, and that local maximum probably gives them slightly-better-than-random outcomes, even if it's far below what it could be.
So in the absence of real criteria, like a poorly-trained ML model everyone just latches on to superficial signifiers that have worked out in the past, and that local maximum probably gives them slightly-better-than-random outcomes, even if it's far below what it could be.