Ask HN: How do you interview devs in a post-AI world?
TL;DR
A hiring manager says about 80% of developer candidates now direct AI agents instead of writing code. The manager asks whether leetcode and design rounds still make sense. Most HN answers say to test judgment rather than typing. They suggest having candidates review flawed code, extend their own submission without AI, or explain every part of what they built.
The discussion
A small, practical thread of interviewers swapping formats that worked for them. Most keep AI out of the room to see how candidates think; a smaller group hires only AI-first developers and grades how they prompt. A few bitter replies show how many developers dislike where hiring is heading.
Test reasoning with AI out of the loop
One interviewer gives senior Java candidates a small CRUD API seeded with bugs like missing auth, which they explain to the interviewer playing a junior. Another has candidates submit a tiny CSV exercise however they like, then extend it live without AI. Every candidate who had AI write the submission failed there, unable to break the problem down even out loud. A third company got better signal from a no-AI take-home than from an AI one, where many pasted the whole ticket into Claude.
Every candidate that has submitted an ai submission thus far has failed because they have literally no idea where to go. I've interviewed dozens at this point. I'm not saying "they're unfamiliar with the structure", I'm saying "they cannot actually break down the problem even verbally".
When they point out errors, I get them to explain why they are errors by playing the junior engineer to see how they explain technical issues to someone that might not have the same understanding
Let them use AI, then make them explain it
Several commenters accept that hires will write most code with AI, so the interview should check whether they can tell good output from bad. They let candidates build with any tool and then question them on the design, the alternatives and the tradeoffs. One candidate called a no-code take-home, a color-extraction tool built with any tools, their most fun interview because they had to defend it. Another interviewer suggests planting a nonsensical requirement that a candidate should push back on and an AI would happily build.
However, the question becomes: Can they determine if the output is correct/well formed/production quality.
Something nonsensical that the candidate should give a bit of pushback on, and a coding assistant will probably happily implement for you.
Hire AI-first and grade the prompting
A minority says to stop testing hand coding at all. One commenter's company hires only AI-first developers: candidates get an unfamiliar repo and an internal AI setup with rules and quality gates. For two hours they may not write any code by hand, and the company still rejects about 80% for weak AI interaction patterns. Replies called this a developer's hell, and others suggested grading the first two prompts or a Claude Code clone take-home.
We still reject about 80% of the people who interview with us, because even though they know how to code with AI, their AI interaction patterns just aren't on par with what we need.
Give them a time-constrained challenging problem that is wide in scope and see how well a candidate can decompose it before feeding it to AI. You can get pretty high signal within the first 2 prompts.
Nothing changed, just talk to them
Some interviewers say AI changed little because they already skipped leetcode. They ask candidates to walk through projects on their CV, what they chose and why. They judge curiosity, product sense and how clearly candidates explain things, and one called it an extended Turing test. Another dropped whiteboard architecture after sitting on the candidate side and finding it stressful and uninformative.
Basically the same as always--an extended interactive Turing test. Can they speak coherently and in-depth about the things they claim to have done on their CV/resume?
Paid work trials
One commenter runs a conventional technical interview and then a paid one- or two-day trial, and says no candidate has refused. Replies doubted this works for people who already have jobs, citing scheduling, legal limits on side work, and onboarding in so short a window. One argued it shrinks the pool to the unemployed. The commenter answered that on-site interviews already cost days and that trialists are handled like any contractor.
Not sure it benefits the hiring company when they are artificially reducing the pool of applicants to only people who don't already have jobs.
Consensus
Leetcode tells you little now. The useful signal is whether a candidate can break a problem down, judge code whoever wrote it, and explain their decisions. Most formats people shared test that through review, live extension or questioning.
Dissent
A minority wants AI-first hires and grades prompting skill directly, and one company says it works well even for juniors. Others came from the candidate side. One developer has shipped a year of Python without writing a line by hand, and asked how to interview at all. Another asked how to find employers that don't force AI work.