Stop Hiring Paper Experts: What a Proof-of-Work Technical Interview Actually Looks Like
Written by: Matt Reaney | Reading time: 6 mins
Are you filtering out the people you want?
Remember the classroom scene in Ferris Bueller? Ben Stein droning at a room of teenagers staring into the middle distance. "Bueller? Bueller? Anyone? Anyone?"
That's a fair description of technical hiring in 2026. Logic puzzles, whiteboard riddles and AI-written CVs that mean nothing in production. A lot of companies aren't defending this process — they've just forgotten there's an alternative.
The real damage isn't that the voodoo lets bad people through, though it does. It's that it filters out the builders. You lose excellent engineers because they're too busy shipping real code to spend evenings practising brain teasers.
Why brain teasers and take-home tests no longer work
Both now measure preparation rather than capability.
Puzzle-style questions were always a proxy, and proxies get gamed the moment the answers are searchable. Take-home tests are routinely generated or outsourced before submission. And any structured question set is by definition a script — which means it can be rehearsed, or fed to a candidate live by an overlay tool during the call.
The 2026 problem in one line: the things that were easy to standardise are exactly the things AI made easy to fake.
What proof of work actually means
It's simple. Instead of asking a candidate to solve an artificial problem, you ask them to defend a real one they've already solved.
Something they actually built. In production. With constraints, deadlines, a budget and colleagues who disagreed with them.
You cannot rehearse the answers because the questions come out of what they just said. And you can't fake it because the follow-ups go three layers deeper than any prep material anticipates.
The questions that separate builders from paper experts
The specifics change by discipline but the shape is consistent.
Trade-offs — why this architecture and not the obvious alternative? What did you give up? Failure — what broke in production, how did you find out, how long were you down? Cost — what did inference cost you at peak and what did you do about it? Constraint — what would you have built with twice the budget or half the timeline? Regret — what decision would you reverse now and what changed your mind? Attribution — which part of this was actually yours, and who did the rest?
That last one matters more than people expect. Fluent candidates tend to describe team achievements in the first person singular. A peer notices immediately. A recruiter does not.
The uncomfortable bit for my own industry
You can hand that question list to any recruiter and it will not work.
Not because they aren't trying. Because they can't evaluate the answer. If you've never owned a latency budget, you cannot tell whether "we moved to a smaller model and quantised it" was a smart call or a disaster, and you certainly can't judge the sentence that follows it.
Recruiters failing at technical hiring is an understanding problem, not an effort problem.
Which is why we don't do the assessing at reaney.ai. Our Industry Advisory Panel does — active CTOs, CISOs and engineering leaders, currently building in the domain, who put every shortlisted candidate through the System Check before a client sees a single CV. Peer to peer.
Real questions, in real time, from someone who has made the same decisions and paid for the wrong ones.
The two-stage process
The Reality Check comes first. We sanity-check the work, the logic and the decisions on paper — filtering out test-passers and AI-polished CVs before anyone gets near the System Check.
The System Check is the candidate sitting with an active technical leader, physical or virtual, defending the architecture of a real system they've built. If they can't, they don't pass. There's no partial credit for a good CV.
The deliverable is the 30-Day Technical Shortlist — verified outliers, checked by the panel, inside thirty days.
The thing this process finds that others miss
Most vetting conversations are about exclusion. This one works in both directions.
Standard processes converge to the average — they reward the candidate whose profile best resembles the profiles that came before. Proof of work does the opposite. It gives the unconventional candidate somewhere to put their evidence.
The physicist who moved into perception. The engineer with a two-line CV because she spent four years shipping one very hard thing. The lead whose company had no name recognition and got acquired eighteen months later. None of them rank well against a template. All of them do extremely well in a room with a CTO asking what broke.
That's the whole argument for doing it this way. Not just that it catches the fakes — it surfaces the outliers everyone else's process quietly discards.
I'd like a meritocracy for my two daughters. To be judged on what they can do, not on how well they navigate a process designed to be gamed.
Life moves pretty fast. If you don't stop and look at the work once in a while, you could miss the best talent in the market.
FAQ
What is a proof-of-work interview?
A technical assessment in which a candidate defends a real system they've built, under unscripted questioning from a practising technical leader, rather than solving an artificial test problem.
Are take-home tests still worth using?
Rarely, at senior level. They can be generated or outsourced, they impose significant unpaid work on candidates, and strong builders often decline them. A live proof-of-work conversation gives better signal in less time.
Who runs the System Check at reaney.ai?
Active CTOs, CISOs and engineering leaders from our Industry Advisory Panel — practitioners currently building in the relevant domain, not recruiters.
Matt Reaney has been building AI, computer vision and cyber security teams since 2013. Book a search briefing today.