Back in 2019 I wrote that recruiting people is effing hard, and then more or less left it there, because I had a stack of CVs to wade through and you had a day to be getting on with. That post was written from the slush pile, the candidate's side of the glass: here is how you write a CV that gets an actual human to read it. It mostly still holds, though these days a robot reads it before I ever do, which is a separate headache.
This is the view from the other chair. Once the CV has done its job and there is a real person sitting across from me, what am I actually trying to find out? Because here is the thing it took me an embarrassing number of years to say out loud: it was almost never the thing I was pretending to test.
Passion, trust, and maturity (and no, this isn't a dating profile)
Three things, roughly in that order. Passion, trust, maturity. I know exactly how that reads. Give me a minute.
The passion I want has been self-directed: someone who took their own learning into their own hands and built themselves into the engineer they wanted to be, and then didn't stop, because they can already see the engineer they could become. It is a bit of a mirror, and we will come back to how dangerous that is. But passion on its own is a loose cannon. It has to be grounded, balanced by someone who can take a proper kicking over a decision and turn it into fuel rather than sulking off to lick their wounds.
Trust is the one I will not compromise on. I am not a manager who hovers, never have been and never will be. I need to know that if I hand someone a problem they will see it through, and, just as importantly, that they will know the exact moment to put their hand up and shout for help, from me or from whoever can get them unstuck. There is no shame in "I don't know." There is a great deal of shame in pretending. Which is why, in an interview, I will happily walk someone clean out of their comfort zone, because what I want to watch is what they do when they reach the edge of what they know. Do they bluff? Or do they say "I'm not sure", and then, and this is the bit that wins me over, have a proper go at it anyway, usually with the breadcrumbs I left lying about earlier in the conversation on purpose.
Maturity is the slipperiest of the three, so I am going to make you wait for it.
The artefact is a pretext
Years ago, leading the technical side of interviews at a PHP-focused agency, I had a candidate in called Alex to talk through some code he had submitted, a tidy little homebrew framework. And I absolutely ripped into him over one choice: he had reached for XSLT to do his output templating, which is a peculiar abstraction in PHP, a language plenty of people consider a templating engine in its own right. Here is the thing though... his code was genuinely good, and the grilling was pure theatre, a tactic to see how he would wear it. He looked me dead in the eye, cool as a cucumber, and said, "So? It works, and it got the job done quickly. What would you rather I used?" I would have hired him on the spot. He took everything I threw at him and turned it straight back into a question of his own. The code was never the thing I was testing. The way he stood behind it was.
That instinct runs through the formal stuff too. Here is the confession the title already gave away. Most of the take-home challenges people send me go straight in the bin.
The brief is deliberately open-ended, with my direct line and email address sitting right at the top, and instructions threaded through it to get in touch with any questions at all. The submission is rarely what I am marking. What I am marking is whether they pick up the phone. Do they question the brief, or do they assume they already know what I want, even though I have told them in writing to ask? Ask, and they are through to the next round. Ask something genuinely sharp, something that makes me rethink my own brief, and they have earned themselves a note and a good few brownie points. I even have the recruiters, internal and external, nudging candidates to ask. Most still don't.
Whiteboards I gave up on years ago. All a whiteboard ever told me was how gregarious someone is in front of an audience, how sure of themselves they are while waving a marker at wavy boxes that could mean almost anything. That is a personality test wearing a lanyard, not an engineering one.
The one I genuinely love is the code golf. For a good while code.golf was a real part of my process. At the end of a technical interview, completely unprepared, I would ask the candidate whether they fancied a go at one of the simple ones, fibonacci, the top however-many places. I set it up like a kata: ignore the character count, just show me how you would solve it, in absolutely any of the sixty-odd languages it supports, five or ten minutes depending on how much I liked them and how the hour had gone. They can say no. I have hired people who said no, there is no shame in it... but it helps. And to be clear, I do not care whether they finish. I do not care whether they score well, though a properly elegant solve will absolutely impress me, I am only human. What I care about is that they have a go, and that when I start poking at it, "why reach for recursion there and not a comprehension?", they can stand behind a choice they made sixty seconds ago under a stopwatch. Cool, calm and collected, or panic and buckle. That is what I am actually reading.
None of this is me being clever for the sake of it. The rituals all have their place. It is that companies reach for them to measure the wrong thing, and that, more than anything else in hiring, is what gets my goat.
It slides with the level
What I weigh changes with who I am hiring. For a junior or a mid, knowledge is the least of it. They have the formulas and patterns they picked up at university and not a lot else yet, so passion is the only forward-looking signal I have to go on. A good junior is already circling the ring, looking for the next bout of close-quarters combat with whatever language they have fallen for. I love nothing more than a junior who is contributing to open source! That is someone chasing the dream rather than clocking the nine-to-five, and they can carry a real conversation about something they genuinely care about. That is where the trust starts.
For a senior or a principal, I want the scars. You do not get to senior without going through the mill a few times, and the experience earned in there has to shine as brightly as the passion that drove them into it in the first place. Put it plainly: an experienced engineer is not someone who knows how to do it right. It is someone who got it wrong enough times to know how not to do it. The scars are the knowledge. Which means even at the top of the ladder I am not really hiring for correct answers, I am hiring for the wrong turns that taught someone better.
I will own a bad habit here. For years I have over-hired seniors. I fell into the old mantra of hiring people better than myself, and, imposter syndrome aside, I am good at what I do, so seniors have tended to be flavour of the month. They are also simply easier to hire: the battle-hardened keyboard warrior with a scar for every mistake can carry a conversation forward all on their own. I have leant on that more than I probably should have.
A team of Matts
Which brings me to the most uncomfortable thing I know about the way I hire. Left to my own devices, I hire myself. Likeminded, passionate, the same instincts, the same itch. A team of Matts.
The catch is that Matt may hire himself, but Matt does not especially like himself. So there is a whole chunk of my own head I have to box off in the room so it cannot run the interview from the back seat, and, more reliably than my own discipline, I make sure there is a second person sitting in whose judgement I trust, someone who will turn to me afterwards and tell me plainly when I have just spent an hour admiring my own reflection. It does not disqualify anyone. A mirror can still be a cracking hire. But it means I walk in with my eyes open, knowing I will have to work to lead a room of people who all think the way I do before it fragments and falls apart. The teams that actually held together over the years were the diverse ones: different strengths, different blind spots, covering for each other. That is not a poster on the breakroom wall, it is just the truth of it.
What about the machine
You cannot write about hiring in 2026 without the obvious question: doesn't AI break all of this? A model will ace your take-home now. It will write a tidy fibonacci in any of those sixty-odd languages before the kettle has even boiled.
And the plain answer is, no, not really, and the reason is the whole argument in miniature. I was binning the take-home anyway. I was never marking the artefact, so it makes very little difference who or what produced it. The thing I am testing, whether you rang me, whether you asked, how you carried yourself when I leant on you, is not something you can hand to a model.
There is a rather lovely trap in it, too. To use AI to get past me, you first have to do the human part. You have to ask the questions, because half the answers I give are made up on the spot, little fictions I have been handing overconfident candidates for years, and without them the model has nothing to work with. Skip the asking and lean on the AI, and you have not cheated the test, you have failed the only part of it that counted.
And knowing the trick will not save you, even with this published. Read it, spot it, and call it out in the room, and good, you have passed, and I will cheerfully change the game on the fly. Read it, say nothing, and try to game it, and you will find my answers getting steadily more ludicrous as the hour wears on, especially once I start cross-referencing your story against itself... with none of your own code to anchor you, just your memory of a submission you never really read, because a model wrote it and I dropped it in the bin. Knowing it almost puts you at a disadvantage.
None of which is me being down on the tools. I will give genuine kudos to the candidate who aces the human part and then uses AI to wring the very best out of the answers I give them. That is not cheating, that is precisely the engineer I want: someone who asks the right questions and then wields the tooling well. A junior who can do both, and gets past my technical interview and my code-golf fetish into the bargain, has earned the right to be forged into something more, and to help me work out which way we steer the boat on these AI seas. That, now, is what the gauntlet should look like.
There is one thing in all this that does keep me up. If the model papers over every small mistake a junior would once have made for themselves, the junior never gets burned, and you do not get a scarred senior in 2032 without a singed junior today. We have already started sawing at that rung. It is why the self-forger, the open-source tinkerer, the one who went through the mill on something because they wanted to and not because a sprint told them to, is fast becoming the most reliable signal I have that anyone has been through it at all.
As for the filter, the robot reading the CV before I get to it, that is exactly what it is: a filter. It cannot tell the wheat from the chaff. I have had CVs the algorithm raved about land on my desk and known, three minutes into the conversation, that it was a no. I have never quite had the heart to end an interview at the three-minute mark, mind, so I sit and suffer through it as graciously as I can manage. The conversation is the one part of the whole business I still trust without reservation.
The bet I can't hedge
Which leaves maturity, the one I made you wait for, because it is the one I cannot really test for at all.
It is abstract, and it shows up almost uniquely in every person I have had the pleasure of interviewing. It is a gut feeling more than anything: how they carry themselves, how they talk about what they have done, how they pick their tools to their own advantage. It does not mean boring, or clinical, or slow and dull. It does not mean self-assured and certain they know everything either. It means being able to own the consequences of your own actions, to apply good judgement in a bad situation, and to find the silver lining that keeps you motivated and moving on to the next big thing. And here is the cruel part: you cannot see it across a table in an hour. It takes months, sometimes, for that maturity to mature. So in the room it comes down to instinct, a bet I place and only get to settle a long way down the line. I would like to think I have been right rather more often than I have been wrong.
So that is the job, when you boil it down. Read a stranger in an hour and wager on who they are going to become. And if you have got this far and you find you rather want to argue with me about some of it, the XSLT, the binned take-home, whether I have just cheerfully talked myself into a team of mirrors, then pick up the phone and tell me so. You will have started the interview without either of us quite meaning to.
Originally published at phpboyscout.uk on 10 July 2026.
