What the first round has to measure now

The first round was always a filter on fluency. Someone who could describe a blue-green deployment, name its failure modes and talk through a rollback had probably done one, because an accurate description was expensive to fake. That assumption is now false. Every candidate can open a capable model in a second tab, and the answers come back clean, structured and confident.

So the round has to measure something else — not what a candidate says about the work, but what they do when handed a piece of it, and not a score on its own but a score with the actions and artifacts behind it, so the hiring manager reading it can disagree on the evidence rather than on the number.

The first round is also the most expensive screening you run. Senior engineers pay for it in hours, and most of those hours go to candidates who were never going to clear the bar. That is the question behind AI technical screening as a category: what a first round is actually for, once the answers are free.