> "The question I use was first used on me and I solved it completely in 20 minutes. I haven't had a single candidate solve it in 45 minutes, and I give them huge amounts of help."
Honestly it sounds like you're proving GP poster's point here: these interview questions are administered by engineers who want to feel better than the candidate.
Speaking from experience, it's much more enjoyable to interview a candidate who does well than one who does not.
An failed interview is awkward, and then I have to turn around and write clear and explicit feedback about exactly why someone shouldn't get hired. It's probably my least favorite job responsibility.
But at least with calibrated questions, I can be fair and impartial and know how much help the candidate needed. Throw me into a room with a question I don't know, and the same general complaints apply, but now I have to figure it out as we go too, so my focus isn't on feedback and helping, but on understanding the problem.
These interview questions are administrated by everyone at Google, so I just picked one used on me after assuring that it wasn't in any public database. The problem is apparently significantly harder than what interviewers typically get, I didn't know that when I picked it. However I don't think it is a bad question, seeing people struggle with hard problems gives a lot of strong signals as well. Also compared with other interviewers I tend to be on the nicer side of grading, so it is not like getting a hard problem hurts them.
Honestly it sounds like you're proving GP poster's point here: these interview questions are administered by engineers who want to feel better than the candidate.