If you hire engineers, the interview you ran three years ago is quietly becoming obsolete. Not because coding stopped mattering, but because the thing that best predicts a good hire has shifted — from whether a candidate can produce code to whether they can decide well. The technical interview is being rebuilt around judgment in real time, and the organisations that notice will hire better than the ones still optimising the old signal.
Why the old signal degraded
The classic technical interview leaned heavily on algorithmic coding problems. For years that was a defensible proxy: someone who could implement the tricky thing under pressure probably had the raw ability to do the job.
Two forces broke the proxy. First, AI assistants made producing standard code cheap and instant, so testing whether a candidate can reproduce a known algorithm from memory measures less than it used to. Second, the coding signal was always noisier than it looked. interviewing.io’s analysis of more than 1,300 interviews found only about 20% of candidates perform consistently — even genuinely strong engineers fail a technical screen roughly one time in five. A single coding round was never the precise instrument it appeared to be, and AI has made its limits impossible to ignore.
Where the signal is moving
The response from serious hiring organisations has taken two visible forms, and both point the same direction: toward judgment.
Some companies locked the assistant out to see unaided reasoning. According to Gartner research reported by Computerworld, 72.4% of recruiting leaders now conduct interviews in person, and firms including Google have restricted AI tools during virtual interviews — not to preserve the coding puzzle, but to observe how a candidate actually thinks.
Others handed the assistant to the candidate on purpose and raised the bar past what it can do alone. Reporting on Google’s 2026 interview loop describes a new “code comprehension” round where candidates read, debug, and optimise real code with an AI assistant available, while interviewers explicitly score “AI fluency” — prompt engineering, output validation, and debugging. The evaluation is no longer “can you write it?” but “can you judge it?”
Underneath both, the leveling logic has become explicit. At senior levels and above — as the pattern is often described at companies like Meta — coding decides “should we hire this person?” while system and product design decides “how senior are they?” The rounds that actually determine level are the ones testing design and judgment, not syntax.
What “judgment” means in an interview
For hiring leaders, the useful question is what to actually assess. Judgment in this context is concrete, not vague. It is the ability to take an ambiguous problem, choose a defensible approach, and articulate the trade-offs — to know what to build, what to leave out, and where a design will quietly fail. It shows up most clearly in the system-design signal, where a candidate reveals whether they reason like someone who has shipped and operated real systems.
It also shows up in communication, which is why the modern interview looks more like a design discussion with a colleague than an exam. The candidate who states assumptions, names trade-offs, and revises gracefully under pushback is demonstrating exactly the architectural judgment worth hiring for — the quality that predicts whether they’ll make good decisions when no interviewer is watching.
What it means for how you hire
The practical implication is to reweight your loop toward what now predicts performance. Keep a coding round as a capability gate if you like, but stop treating it as the decider — its signal is noisy and increasingly gameable. Invest interviewer attention in design and judgment rounds, and calibrate your team on what a strong reasoning answer sounds like, because assessing judgment is harder and more subjective than checking whether code compiles. For senior roles especially, understanding how senior candidates are now assessed — on scope, judgment, and influence rather than raw output — will separate the organisations that hire well from the ones still measuring the wrong thing.
The bottom line
The technical interview is not getting easier or harder; it is getting more honest. As AI absorbs the mechanical parts of engineering, the parts that remain — architecture, judgment, communication, knowing what not to build — are exactly what interviews are moving to measure. The bar moved from coding to judgment. The hiring teams that move with it will spend the next few years hiring better than the ones that don’t.
Sources: interviewing.io’s 2025 survey and interview analysis, via Interview Copilot; Exponent, “Google’s AI-Assisted Coding Interview (2026 Guide)”; Computerworld, “To counter AI cheating, companies bring back in-person job interviews”.