Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

AI definitely makes take homes and non live coding exercises less viable (and even live ones to an extent).

Not my favorite AI driven change as I think live coding is so high pressure it can give wrong signals.



I don’t think take homes should really be about the code, but about the developer being able to reason about why that was the code they wrote.

Asking developers to explain why they wrote that code mitigates against using LLM coding tools - if the candidate can’t back it up then they’ll do poorly in it.

I recently had a candidate submit an otherwise average exercise that was a big mish-mash of coding styles (inconsistently using var/let/const in js, for example). When asked about it, they weren’t able to explain their choice at all and just stumbled through it.


I agree that live coding is very hard on a lot of candidates. How do you feel about asking candidates to read code and explain it? I had that only once and I thought it was genuinely innovative. Even if I couldn't understand all of it, we can discuss various points about it.


I think if that's prompted well and designed well it might work. But if not careful it might just be a signal that reflects your own taste.


Somewhat less but if you are having them do a monorepo with the latest major releases of the frameworks involved, AI will mess it up because there is a 4-6 month knowledge gap




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: