AI handles the finish, but designer judgment still has no substitute

A UX Collective essay lists the skills that separate those who think from those who only produce screens. For Brazilian design, used to competing on execution, it's time to take this seriously.

A UX Collective essay lists the skills that separate those who think from those who only produce screens. For Brazilian design, used to competing on execution, it's time to take this seriously.

Nil Thyrion has been reviewing portfolios, hiring designers, and leading design teams for years. In an essay published this Saturday (11/7) on UX Collective, he sums up in one paragraph the discomfort I've also felt reviewing discovery and prototype work these past few weeks.

But something has been bothering me. Recently, reviewing work with my own team, I found myself thinking: I would rather see a great idea drawn on a piece of paper than a beautifully vibe-coded prototype solving the wrong problem.

Nil Thyrion, design leader and author at UX Collective

Vibe coding, for those who haven't kept up with the term, is generating a functional interface just by describing what you want to an AI, without writing code line by line. Thyrion's point isn't against the tool. It's that it solves the "how" and doesn't touch the "should we." And that changes what's worth hiring for, promoting, and expecting from a design team.

The craft became a commodity, judgment didn't

Thyrion splits craft into two parts: literacy (knowing theory, principles, references) and skill (knowing how to apply that in a real decision). He cites Anders Ericsson's 1993 study of violinists at the Berlin Academy of Music, which inspired the (now somewhat mythologized) 10,000-hour rule. What the study actually showed wasn't a magic number, but that mastery comes from deliberate practice: play, fail, correct, play again.

This matters now because AI has driven down the cost of literacy. Anyone can generate a screen with correct visual hierarchy, consistent components, decent copy, in minutes. What the tool doesn't give you for free is the skill to know whether that screen solves the right problem. In the Brazilian market, where much of the hiring process still asks for a "pretty portfolio" as the main criterion, this is an uncomfortable inversion: what used to set a senior apart from a junior got cheaper to fake.

Having references isn't having judgment

Industrial designer Raymond Loewy summed up decades of observation in an acronym, MAYA (Most Advanced Yet Acceptable): people want what's new, but they're wary of what's too new. Thyrion calls this taste, and describes how he developed it early on, comparing logos he made for local businesses with the best work he knew, going home unsatisfied, redoing them.

In short: taste isn't knowing what's trending. It's knowing where the line sits, for this product, this audience, this moment, between "we've seen this before" and "this isn't us." AI trained on billions of interfaces tends to converge toward the middle of the funnel of existing visual references. It reproduces what's acceptable. It doesn't risk the advanced.

Bringing a hypothesis before the research

French typographer Olivier Nineuil, according to Thyrion's account, would make a tiny sketch in the corner of his notebook during the client briefing, before any formal process began. Weeks later, the project's final direction would usually land close to that first sketch.

Thyrion uses the example to attack a habit I also see in Brazilian squads: treating "we haven't done the research yet" as an excuse not to have an opinion. Discovery, here, is the process of investigating whether a problem exists and is worth solving before building the solution. It shouldn't be a ceremony to validate an idea that's already settled, but it also shouldn't be used as a shield against having a hypothesis. Good instinct is experience running faster than explanation, not a guess with no track record.

Seeing the cathedral, not just the brick

Thyrion draws on the parable of the three stonecutters, recorded by Peter Drucker in The Practice of Management: asked what they were doing, one answered "I'm earning a living," another "I'm the best stonecutter in the country," the third "I'm building a cathedral." All three were doing the same manual work. What changed was how much of the whole each could see.

On a digital product team, that becomes a practical question: a designer who gets the request "create this icon" should take care making the icon, but should also be able to ask why that icon is needed. Maybe it duplicates something that already exists. Maybe the problem lies in how the experience is organized, not in the lack of a symbol. Asking only for a better icon guarantees a better icon. It doesn't guarantee discovering that it was unnecessary.

The recipe isn't the method

Chef Massimo Bottura has a line that Thyrion uses to talk about design process.

Recipes are not meant to be followed literally.

Massimo Bottura, chef

Double Diamond, Design Thinking, journey map, prioritization workshop: these are tools, not the craft itself. Thyrion describes a pattern I recognize on hiring panels here: dozens of portfolios with the same sequence of slides (persona, pain points, journey, wireframe, final screen) without the candidate being able to explain why that specific process was chosen for that specific problem. AI that generates a journey map in seconds only makes this worse: producing the artifact got easy, deciding whether it was necessary is still hard.

The counterargument you can't ignore

There's a real hole in this reasoning, and it's fair to name it: a good share of designers in Brazil don't sit at the table where strategic decisions get made. On many squads, the PM decides the what, the designer decides the how, and "having an opinion about the problem" doesn't fit the scope someone handed you. Asking that designer to "think more strategically" without changing the structure that hired them is, in practice, demanding work the company neither pays for nor recognizes.

But that doesn't invalidate Thyrion's argument, it only limits where it applies on its own. The way out isn't waiting for permission to have judgment: it's bringing the hypothesis even when it's not asked for, questioning the brief in a critique, documenting the reasoning behind a visual choice instead of just handing over the file. That's the kind of behavior that stands out in a portfolio precisely because most people don't do it. Changing a company's structure is a slower game; changing what you deliver inside it starts tomorrow.

The question that remains

The next time you generate a screen with AI in minutes, it's worth pausing before polishing it: is this the right solution, or just the fastest one to produce? Thyrion closes the essay by comparing the designer to María Lorena Ramírez, the Rarámuri runner who won a 50-kilometer ultramarathon in 2017 wearing huaraches, against athletes with top technical gear. The tool makes the runner better. It doesn't make the runner exist before the race.

Source: Nil Thyrion, "Otherwise you're just a machine", UX Collective (11/7/2026). Available at: https://uxdesign.cc/otherwise-youre-just-a-machine-a9d3b21179b5