> The most likely scenario is that a non-technical person is being expected to gauge technical answers.

I'm not disagreeing with this - this certainly appears to be the case.

But I think a qualified technical person should be able to understand the question that the non-technical person is asking and respond in a useful way. Although, yes, if they immediately respond with "Wrong, it's <keyword>", it's hard to do that. But I feel like a good interviewer (good is orthogonal to technical!) is likely to say "OK, so what is the runtime?", and "constant" and "O(1)" should both be on their list of keywords.

My charitable mis-remembering would be that the transcript here skipped these sorts of prompts, or that the interviewer was actually upset at the interviewee's demeanor/attitude already and wanted to cut the interview short by that point and was just trying to finish their block of questions. (Which I think is legitimate. As a technical interviewer, if you start condescending to me during the interview, I'm much less likely to give you the benefit of the doubt and help you along with Socratic hints.)

"As a technical interviewer, if you start condescending to me during the interview, I'm much less likely to give you the benefit of the doubt and help you along with Socratic hints."

As a technical interviewee, if a screener acts like all they are there for is to prove that they are smarter than me and start condescending to me, I'm not going to give a damn about their questions because I wouldn't want to work with them.

When a technical question has more than one answer, (like binary, decimal or mnemonic values for changing file permissions), the screener has to know that, or they make fools out of themselves. The whole thing with "attributes" vs "metadata" was just stupid. (Technically the metadata is comprised of the attributes. So they are both correct.)

>But I feel like a good interviewer (good is orthogonal to technical!) is likely to say "OK, so what is the runtime?", and "constant" and "O(1)" should both be on their list of keywords.

In that case, he was the CTO, though, and should know that "constant" is the same as "does not depend on the input size". (Though I'll admit I could have been unclear by trying to pre-emptively show how I knew what the next binding constraint would be.)

Agree with your points otherwise, but I still think the protocol should be (even if you're skipping through), to say something more like "let me note that down and I'll pass on your response" rather than imply it's not wrong because it's not in your list.

It's not the derivative of distance vs time that matters but the integral of acceleration.