A note on "The worst programming interview question"

Like all interview questions, if they are not structured and scored, they're unreliable as predictive tools. I'm always amazed that data-driven companies ignore reliability and validity when hiring and evaluating performance.

His reasoning for using puzzle questions may have been face valid, but there's no evidence that they're related to resilience (or anything else they're supposed to measure) and the little research that has been done has shown only a moderate relationship with cognitive ability(http://www.researchgate.net/profile/Chris_Sablynski/publicat...).

I agree that a trial would be best if not for the lack of time. A good alternative is an assessment center, which is a structured simulation of the job. They're difficult and costly to develop but once they're complete they can be reused year to year.