Hacker News Reader

Comment by the_arun | Hacker News Reader

From jobseeker point : We need better interview process. I don’t understand leetcode style of interviews when we have ChatGPT to generate code. Even OpenAI has tough coding rounds.

View on HN

Friend of mine interviewed there. Left the coding part of the interview with the strong impression that they were using him to train some kind of code LLM.

Google interviews always felt like this, even a decade ago.

That's genius. Instead of dumping resumes in the trash, dump them into an AI training program (and then dump them in the trash)

Heh I did a coding interview for one the Ai companies that post here all the time(Imbue i think it's spelt) and when I got rejected I was wondering if it was just some AI training thing...

Yes, but nobody is lobbying for worse interview processes and if anyone has an incentive to make it better, it's employers (for whom it is disproportionally more expensive).

People love to poke fun at it, but employers are not trying to make it bad. It's just a hard problem.

Even a single person can be hypocritical or self-sabotaging sometimes. And "an employer" is often not a single person but an organization with a bunch of people pulling in different directions. Some people making hiring decisions or deciding on hiring procedures may not necessarily care about the procedure being "bad". Maybe they e.g. care about maintaining "standards" or whatever justification they came up with.

Unless you have a novel insight, let's assume that market forces apply. Yes, we all had bad experiences — and they go into the calculus. A company that creates consistently horrible experiences for their candidates will be taxed by the market.

I always understood it was to filter out dudes who’d succeed at it: people who have a lot of spare time who would use that spare time to grind really hard for some arbitrary coding test. As an employer, I’d look at that dude and think: this dude could defo do overtime and he likes coding! We’ll take him. But that’s my two cents.

ChatGPT isn't very good for knowing how well or correctly code will run

Reply on news.ycombinator.com