Hiring for eye, not pipeline

The single hardest hiring call we make is whether a candidate has *eye* — the ability to look at a frame and know, without being able to fully articulate it, whether it's working.

Everything else is easier. Pipeline fluency, software skills, prompt literacy, shot language, colour theory — those are teachable. We hire smart people and expect them to get up to speed on our specific toolchain within three to six weeks. That's just training.

Eye is different. You either have it or you don't, and if you don't, no amount of process or tooling will produce work that feels alive. The lab only works because every seat in it has eye. Lose that, and we become a prompt factory, and prompt factories are a commodity.

What "eye" actually means

We get pushback on this word. "Eye" sounds mystical and vague and vaguely gatekeeper-y. Fair enough. Let's define it.

Eye is a cluster of specific, identifiable skills:

• Being able to rank ten frames from most to least interesting, and defend the ranking.
• Being able to identify — in under five seconds — what's wrong with a frame that's almost working.
• Being able to watch a 15-second reel and name the moment that stops it from landing.
• Being able to look at a competitor's campaign and articulate why it's good, or why it's not, *beyond* surface aesthetics.
• Being able to hold a frame against the brand brief and know whether it's on or off without checking a colour chip.

Those are trainable, up to a point. But the *speed* at which someone does them — that's the part that doesn't train well. Candidates with eye do it in seconds. Candidates without eye do it in paragraphs, and usually get it wrong anyway.

The interview questions we threw out

Early on, our interview loop looked like every other creative agency's. We asked about favourite work. We asked about process. We asked about software stack. We asked about dream projects.

We threw all of it out, because none of it discriminated between candidates with eye and candidates without. Everyone has memorised answers to those questions. Everyone has a favourite campaign and a story about their creative process. The signal is zero.

What replaced those questions:

The folder exercise. We show three folders of output — two from us, one from a competitor — and ask: "Which one would you ship to a client, and why?" This question does more work than the previous ten combined, because it forces a real judgment call in real time with real stakes.

The candidates who just point and say "the best lit one" self-select out. The candidates who say "the middle folder, because the first one is technically better but the middle one has more intention" — those are the ones we keep talking to.

The rescue exercise. We show a frame that's almost working but has one clear problem. We ask: "What's wrong, and what would you change?" The candidates with eye spot the problem in under five seconds. The ones without eye spot the wrong problem, or spot no problem, or list five problems of equal weight.

The taste provenance exercise. We ask: "What's the last piece of visual work — any medium, any year — that made you actually stop and look at it for more than a minute, and why?" This is the single most revealing question in the interview. Not because the answer matters on its own, but because people with eye can answer it quickly and specifically. People without eye either name famous work they think will impress us, or they struggle to name anything recent at all.

What we stopped caring about

A lot of things, actually.

• Brand-name portfolios. Having worked at famous places doesn't correlate with eye. Some of the worst portfolios we've seen came from candidates with four Fortune 500 logos on their CV. Some of the best came from people who had spent two years making independent work that nobody saw.
• Software certifications. If you can learn Photoshop to a professional level, you can learn our pipeline. Certifications mostly signal that you're good at passing exams.
• Degree. We have teammates without formal art training and teammates with MFAs. There is zero correlation with eye. We stopped filtering on it.
• Years of experience. We have junior hires with better eye than senior hires. We've hired people straight out of side projects. Experience buys process fluency, not taste.

What we care about a lot more

• Recent output. Not the portfolio of past client work — what have you made in the last 90 days, for yourself, with no brief? If the answer is nothing, we know something about how much eye development is happening in your life right now.
• Specific reactions to specific work. Can you look at a frame and have a specific, defensible opinion in under ten seconds? Not "I like it" or "it's cool." A specific opinion.
• A sense of what *your* work looks like. Not "I can do anything in any style." People with eye have a centre of gravity, a taste, a set of things they gravitate to. Generalists who "can do any style" are usually generalists who can do no style.

The senior/junior trade

Here's the uncomfortable one: for our lab, a 22-year-old with eye beats a 35-year-old without, almost every time.

That's not an ageist statement — we have teammates across the age range and the best eyes in the room are not clustered in any one demographic. But the trade is real. Eye is a rare resource. Experience is a trainable resource. If we have to pick between the two, we pick eye, because experience compounds on eye but not the other way around.

The operational point

We don't need more operators. Tools have absorbed most of the operator work. What we need is people who can *decide*, and decide well, under time pressure, with taste, about images that don't yet exist.

That's what we hire for. It's specific. It's hard to test for in a first-round conversation. It's the only thing we care about.

We don't need operators. We need opinions. The tools do everything else.