I work for a technical screening firm, and we still do a live coding round on a small, self-contained problem. The interview isn't testing what engineers can produce - it's testing how they think. If we can't watch someone reason through a small problem on their own, we can't judge whether we'd trust them to use AI wisely on much bigger ones later.

In the end, hiring comes down to trust that the engineer will make good decisions, including decisions about when and how to use AI.

P.S. Small plug: we're currently hiring engineers who fit that description https://ats.vistulo.com/roles/hn/

Well its been my experience that developers using AI are ahead, it obviously tool you want to use, also I don’t think it make sense to let engineers to try to write code in interview. I’ve been using ai for a while and I would not pass any any coding interview, I don’t even know how to write code anymore by hand, but I know very well what I want and how I want it. So my cookbook on how to hire is as follows:

1) look for opensource software, hunt ppl who been doing software for a while ideally before the era of AI, you are looking for ppl who had expirience before the ai tools 2) once you find someone look for skill, if you consider work good get in touch 3) if engineer replies get the technical stuff out of him, you want to know he has theoretical base and it matches what you saw on his past work 4) the important part, get out of him what tools he uses. That is for me single most telling thing. If answers opus, gpt55 high or whatever is currently best you have candidate 5) last step is to understand if he wants to sell you ai slop or willing to actually work, services such as hubstuff helps with that

- you have to value architecture skills, strong foundation in software, ego is not always bad as long as it is healthy amount, and open minded person, keeping up with latest trends is winner

Ugig