So the way you feel about interview puzzles is how most people feel about them. It's -stupidly- rare for people to 'figure out' the question in an interview; even those who appear to are just dredging it up from having seen it somewhere. A lot of those style questions are things that took someone weeks to figure out originally, and while we may be better equipped than they were, and so able to figure them out faster…
> It's -stupidly- rare for people to 'figure out' the question in an interview ...really? I've heard of people failing interviews at the Big4 because they solved the problem, but did it sub optimally.
As much as people say they don't care about the answer and suboptimal is okay, generally it isn't, there's very often a very specific answer they want you to arrive at or you are some version of wrong; so the interview is either failed or succeeds on whether or not the interviewee has seen this problem (or one close enough to be virtually indistinguishable from it) before and can remember how to monkey dance through the steps of it again. So a suboptimal answer ends up being approximately the same as a wrong answer which fits into lostcolony's narrative perfectly well.
So basically you have a few different types of possible interview outcomes with these puzzle-interviews:
#1 Candidate who isn't a good fit (lacks experience, maybe just plain not that bright) fails because he or she is just not a good logical thinker and thus not a very good programmer.
#2 Candidate who is a good fit (has plenty of experience and practical programming knowledge) fails because social anxiety kicks in when confronted with a puzzle they don't recognize at first glance, making it difficult to figure out a solution that they may very well be able to knock out in 15 minutes if they were sitting at home with Google, some books, and nobody (who has important fate deciding power over them) looking over their shoulder.
#3 Candidate who is a good fit who recognizes the problem from having seen it and worked through it before, "solves" it, maybe even solves it in a suboptimal way at first in order to avoid looking like they knew the solution all along because they saw it yesterday on some interview quiz website, then busts out the optimal solution as if it came fully formed from their glorious brain box.
I've been both #2 and #3 plenty of times each and know lots of others in the field who will admit to being both of those people at different times and different places in different interviews. Occasionally you can be a partial #2 and still make it through and get an offer because you already had a bias working in your favor (recommended by someone the interviewer trusts), so the interviewer can start to believe the claptrap that the process is important and not the answer. But with no such bias working in your favor, you're done for!
I'll accept there may be a #4, super genius who can really figure all of these puzzles out on the fly from first principles, but that's like a 0.001%er and even if they exist, it is ridiculous that people are looking for that person to write code for their relatively straightforward CRUD app or social media webwhozit thing.
Ultimately, IMO, the whole system is fucking stupid and everyone who isn't a complete jackass realizes it but nothing ever really gets done to fix it on a systemic level.