Every time I read these posts I find them very depressing. Crunch through an enormous bunch of algorithm problems in order to perform well on some unrealistic whiteboard programming (I think the system engineering problems are actually more useful though) - as usual we've focused on the easy to measure metric rather than anything actually useful. Maybe these things are meant to be a proxy for how good you are at othe…
After 20 years in software, I had a phone interview with an Amazon guy who sounded half my age (at most). Asked me to write some kind of function to do something, sort some stuff or some such. I described verbally how I would do it, almost line by line, in pseudo-code. He said "I need you to write it out in [some language]". I said "But why? I just told you how I'd do it." But he needed me to literally write the code…
The point is that when you write out the code, there will be several other people who will end up looking at it as part of your “packet”. This can actually be beneficial for the interviewee: if you use some technique or library that the inexperienced interviewer has never seen, they might be confused and write negative feedback, but someone down the line will be able to see what you wrote and say, “Oh, this is actually really good.”