I wouldn't call 25% an outlier. Also, that's only people he knows. The numbers from interviewing.io (Aline Lerner) suggest that interview performance is arbitrary[0] and that engineers are bad at gauging their own interview performance[1]. The numbers suggest it's far more common than just 25%.
> I'm one of four people I know of who fails interviews at a high enough rate that I could fail 20 in a row (and I believe I have).
My reading of that is that out of all the people he knows, there are four that could plausibly "fail" 20 interviews in a row, not that 25% of (engineers?) he knows would do so.
>"I'd never thought about it this way before (look at who gets rejected and see what they have in common), but I guess there's one thing interviews are pretty good at filtering for: people who are a certain type of nervous in interviews."
Sounds about right. I can pass "take home quiz" style interviews at nearly 100%, but put me in front of a whiteboard and everything goes blank. Not sure why people insist on doing that.
Same experience here, I always chalked it up to anxiety. Tons of interviews, they generally love the hand-in stuff, but then whiteboarding kills me. Definitely jaded about it by now.
