One question,
answered properly.
Notes on interviewing engineers in a world where every one of them has an assistant. No predictions, no maturity curves — the arguments we have had in public with hiring managers, written out in full.
Can you detect AI use in a coding interview?
Not reliably. And a detector that worked perfectly would still answer the wrong question — here is what the false-positive maths does to honest candidates.
When is an AI-free assessment the right call?
We sell the opposite, and we still think some rounds should exclude AI. The three cases where it works, the one condition it needs, and when it is a trap.
Who checks the evidence behind an AI score?
A score that cites its evidence beats a feeling. But a citation can be invented, so something has to re-check it — and withdraw the score when it fails.
How do you assess an engineer who uses AI?
Four things to grade once the finished code stops being the signal, what each one looks like in a real session, and why a score has to point at something.
Should candidates use AI in a technical interview?
Yes — and the company that builds the model reached the same answer for its own hiring. The argument, and what changes about the interview once you allow it.
// rss