> I frequently ask at least one or two questions that I really don't expect anybody to know the answer to. Just to expand on this: It seems really optimal, in

> I frequently ask at least one or two questions that I really don't expect anybody to know the answer to. Just to expand on this: It seems really optimal, in terms of extracting information, if the candidate can answer about half the questions (maximal entropy, if you will). If the questions are too easy and the candidate nails them all, you're wasting time. If the questions are too hard and the candidate fails them all, you're wasting time. You want to quickly get to the limit of the candidate's knowledge, and spend some time there. This is particularly true if you aim to cover several areas of interest quickly.