Very cool, if true

Can confirm. I've done a lot of interviewing people for jobs over the years, skewing very techie. I don't use FizzBuzz itself, but very similar, simple coding tests. They are useful as conversation starters for actually able candidates - getting in to performance and space/speed and maintainability tradeoffs as the requirements might grow and change over time. Indeed, good candidates often want to know about the context before they write a single line of code.

But the main point is to screen *out* the ones who sound utterly plausible in interview, and have an entirely plausible CV, often with coding jobs behind them with good references, but literally cannot write *any* working code for the simplest of toy problems. I am not exaggerating here.

Every time we've had a shortlist of 4-6 coders to interview - and I've never been short of candidates so these are 4-6 good ones on paper - at least one has flunked this test utterly. They'll do absolutely anything to try to avoid actually doing it, and when they do write something down, it's gibberish and can't possibly work. You get them to talk you through it and it's bewildering nonsense that sounds like they're talking about a program but doesn't come close to solving the problem.

I once, near the start of my interviewing career, appointed someone who would have failed that test had we had one, and I was the most junior person on the panel so it wasn't entirely my fault, although I was the only actually technical person so it was to a large degree. It was a miserable, draining experience for everyone for the entire year it took to get rid of them.

The only tip I have for candidates regarding this (assuming you can actually code!) is to be aware that while asking for more context before cracking code can be a very positive sign, stalling on actually writing any code is a showstopping bad sign and you don't want to run close to raising that flag.