[nerfhammer]

Even if he got it slightly wrong, the more fundamental problem is sample size.

CRUMB : BREAD :: (A) ounce : unit (B) splinter : wood

How much confidence would you have in a version of the SAT that was reduced to just one question which determined the entire score?

It's exactly analogous to tech interviewing. It's like saying, oh he couldn't even get that analogy right! He can't do analogies. Same thing with giving people one programming question. Much less nitpicking that one question...

So we are forced to go on lots of interviews and just hope that we happen to quickly grok the one or two problems they give us -- so we'll look "perfect" in that short period of time. which is eventually almost certain given enough interviews. but you can't be super confident to pass any particular interview you might have had your hopes on.