Isn't this quite paradoxical that those kind of companies have hard interviews where you have to almost prove that P=NP in front of 5 people that judges everything you do, yet they fire engineer because "ratio"?
Could this lead to a manager that tries to have the best team for himself will always include average employees that they can easily fire?
The firing is not motivated by the ratio. It's not a handcuff saying you must find x% of the workforce to fire.
It's setting a bar that when evaluated on a curve a certain percentage of your employees wind up being below that bar. You derive the ratio from how aggressive you expect your bar to be.
But no one is saying you need to fire people. They're saying they expect a certain amount of lower performers and if they don't see it, they want to know reason. But it also only manifests at much higher populations than an individual team. Totally possible a small enough team has everyone >= meeting expectations.
> Isn't this quite paradoxical that those kind of companies have hard interviews where you have to almost prove that P=NP in front of 5 people that judges everything you do
Interview process at Meta isn't that hard. Mostly medium leetcode questions (in front of single interviewer), not particularly tricky. With enough preparation, it's doable by any reasonably good undergrad. Google was slightly harder though. Similarly, system design can be prepared.
The hardest part is to get the interview in the first place. But if you do, it's just a matter of preparation. You want to be able to nail these questions.
That's quite the qualifier. I could not regularly pass these interviews when I had done 200+ LC problems. It's only when I got to 1000+ LC problems and had done hundreds of technical interviews that I was able to regularly pass a lot of these interviews. Even now, it still requires me to prepare a lot because how often am I really thinking about suffix tries? It's an arms race. It has happened in universities as well. An arms race between the professors trying to make sure a lot of students still fail their classes while every student needs a 4.0 to get an internship/job now.
It will get higher and higher as time goes on because we're mostly focused on H1B candidates who come from a similarly culture of grinding for exams. You can see this lifestyle is very normalized on the LC, 1point3acres, etc. Cheating is also seen as completely fine too. At this point, I feel like it's almost unfair to not be cheating due to how many people are. Your competition doesn't care about some sense of ethics (nor does the hiring manager - they cheated to get in too!).
I've met with multiple folks at FAANG and such where these practices of grinding + a little "magic" get you in. You talk to someone and they'll say, "oh yeah, I have like three guys I went to high school with in [country] in my house. Once one of us got accepted, it was easy for the rest of us."
It's a bit like a sport, you can have the skills and yet still have an off day or perform worse under pressure, so it's not just a matter of preparation.
From my experience on the Meta side of the interview process, the questions are designed to account for "a bad day".
Also, the follow ups are designed to eventually lock the candidate, since is not enough to pass the low bar, they want to know what your high bar is too.
These interview practices are probably in part because of the huge number of applicants trying to get into Meta, which allows them to rachet up the difficulty to the highest level.
My assumption based on some interview experience articles[1][2] is that there's an internal arms race between the recruiting and engineering at Meta where candidates are not forwarded for the interview process until they feel sufficiently prepared, often giving them months (in any other company, I feel this would be a big red flag), and interviewers expecting the equivalent of Djikstra or Knuth to join their team, regardless of what the team actually does. (I assume not everyone writes distributed system or database implementations from scratch at Meta.)
As a former interviewer at Meta who did hundreds of technical interviews, I can tell you that the people that I hardly expected "Djikstra or Knuth". Honestly I gave pretty reasonable questions that were similar to the types of things I would do in my work (with some modifications to work as an interview question). And still the majority of people didn't do great. It's even worse at the "screen" stage, where I've had people that, e.g. didn't know what "function" means or couldn't write a loop to find the max number in an array.
Yup, it will become a self-fulfilling prophecy where you hire extra, possibly not top candidates, just so you have something to shed when the time comes without affecting your existing team.
Could this lead to a manager that tries to have the best team for himself will always include average employees that they can easily fire?