Extreme close-up of a single crystalline form mid-collapse — an elegant geometric structure that is simultaneously beautiful and disintegrating, its outer facets crumbling into dust motes while the inner core still glows with fading copper light. The crystal consumes itself from the outside in. Warm gray (#A8A29E) dominant for the dissolving outer shell, copper (#D4A574) as the fading inner glow — a single point of meaning surrounded by expanding emptiness. Dust motes, subtle fog gradients, fine grain. Focal crystal centered in vast negative space. Composition: Macro. Density: Sparse.

The Buzzword That Ate Itself

2026-03-042,263 words · 10 min read

Every few years, the design world finds a word it can't stop saying. The word starts out meaning something. Then it means everything. Then it means nothing, and the people who used it first pretend they never did.

"Design thinking" ran this cycle in about a decade. A real insight about how designers work got packaged into two-day workshops, stripped of the critical evaluation that made it useful, and sold to companies who wanted the label without the rigor. Tim Brown published Change By Design in 2009. Stanford's d.school built a curriculum around it. By 2017, Natasha Jen, a partner at Pentagram, stood on stage at Adobe's 99U conference and said what practitioners had been thinking: "Design thinking is bullshit." The term had become a credential you could claim without producing evidence. It meant everything and therefore nothing.

"Disruptive innovation" followed the same arc. Clayton Christensen coined it in 1997 to describe a specific market phenomenon. By 2014, a historian was dissecting it in The New Yorker as a concept that had ballooned from specific description to all-purpose justification. Christensen himself regretted the prefix. "Empathy" went next, from genuine insight about understanding users to corporate training module to checkbox.

The pattern is predictable. A practitioner identifies something real. It gets packaged and evangelized. Adoption inflates into dilution. The concept gets so broad it becomes a credential signal, and then the backlash arrives.

Right now, "taste" is somewhere between inflation and dilution.

The Sommelier Test

Go to a wine shop and tell the person behind the counter what you're looking for. They'll ask you questions with specific vocabulary. Do you want something fruit-forward or savory? Bone dry or off-dry? Light-bodied or full? What kind of finish? Smooth, spicy, bitter?

Every one of those terms maps to a measurable property. "Dry" means less than one gram of residual sugar per serving. "Full-bodied" means high tannin content and alcohol above 14%. "Smooth" describes a specific tannin texture profile. The wine world has spent centuries building a shared language that connects subjective experience to objective characteristics. Two sommeliers can disagree about whether a wine is good and still agree about what it is. The vocabulary gives subjectivity a structure to push against.

Now try the same exercise with software.

Tell me what "good taste" looks like in a product. You'll get a hand-wave toward intuition ("you know it when you see it"), or a tautology ("it's knowing what's right"). One writer calls it "the uncanny, almost intuitive ability to know what should exist and what should not." The words "uncanny" and "intuitive" are doing all the heavy lifting. They're a confession that the author can't fully explain what they're celebrating.

The tech industry's taste discourse has reached saturation. A VC writes a LinkedIn article titled simply "Taste." A Norwegian engineering firm declares that "Taste Is the New Intelligence." The titles vary. The template doesn't. AI can now do X, but it can't replicate taste, therefore taste is your competitive advantage. A concrete definition of taste never follows.

The closest anyone gets to specifics is when a VC describes taste as something found "in the error messages, the loading states, the features you killed because they were merely good." That's a useful observation. But she also describes taste as something that "must be absorbed through proximity and narration, not documentation." Something you can possess and feel, but never quite teach or explain.

That's the mysticism dodge. Once a concept becomes unfalsifiable, it's completed the buzzword lifecycle. It's become a compliment. And a compliment has never helped anyone make a better decision.

A sommelier who described wine as "uncanny and intuitive" would be out of a job by Tuesday.

One writer saw the trap clearly. Sari Azout observed that the fixation on taste emerged because "we're worried that our specialness will be automated with AI, and taste has emerged as our last defense." Then she pointed at what the rest of the discourse keeps missing: "Taste is a byproduct. An emergent property. Taste comes from hunger."

Everyone is celebrating the output. Nobody is describing the work.

Two Kinds of Verdict

A designer and ethicist named Cennydd Bowles offers a more useful frame. He compares taste to chess mastery. A grandmaster doesn't calculate every possible move. Their pattern recognition, built over thousands of games, tells them which moves are worth analyzing and which ones to discard. Taste, in this framing, is the accumulated ability to render a verdict with reasons.

There are two kinds of verdict in play, and they behave very differently.

The first is aesthetic. Does this look right? Are the proportions balanced? Is the typography consistent? Does the color palette feel cohesive? AI can approximate this. Feed it enough examples of well-designed interfaces and it will generate visually competent work. This is already happening at scale.

The second is contextual. Given who will use this, under what conditions, with what constraints, in what emotional state, for what stakes. Does this specific decision serve them?

That second verdict requires knowing things the interface doesn't show you. The regulatory environment. The literacy level of the user base. The failure mode that matters most. Whether the user chose to be here or was forced.

The difference between these two verdicts is the difference between knowing a wine looks beautiful in the glass and knowing whether it belongs on this table, with this food, for this occasion. The first is pattern recognition. The second is judgment. And the second is where people get hurt when it's missing.

When the Interface Is the Instrument

Between 1985 and 1987, a radiation therapy machine called the Therac-25 delivered lethal overdoses to at least six patients, killing three. The machine's interface was modern and responsive. It used single-keystroke commands for speed. When something went wrong, it displayed a cryptic error code and let the operator restart with one key press.

The interface was designed for efficiency. By every surface-level metric, it was user-friendly.

An AI trained on consumer UI patterns would have approved that interface. It was clean. It was fast. It scored well on conventional usability metrics. The aesthetic verdict was fine. The contextual verdict, the one that asks "what happens when this goes wrong," was catastrophic.

The pattern repeats in quieter ways. In ICU wards, alarm systems generate an average of 771 alerts per bed per day. Each alarm is individually correct. Clinicians stop hearing them. The interface works and the patients are at risk. On the surface, the system is functioning. In context, the system is training the humans around it to ignore warnings.

The stakes don't have to be medical for the contextual verdict to matter. Robinhood built one of the most visually polished trading apps in fintech. Confetti animations celebrated completed trades. Frictionless access to options trading made complex instruments feel simple. By every aesthetic measure, the app was excellent. A twenty-year-old named Alex Kearns saw a decontextualized $730,000 negative balance on that beautiful interface and took his own life. The question of who was using this, and what the stakes were, had never been asked.

Forty-Two Pages

In Michigan, the state benefits application was 42 pages long.

If you needed food assistance, or healthcare, or emergency housing, you sat down with a form that assumed you had a permanent address, consistent documentation, and the literacy to parse bureaucratic language. Many applicants didn't have one of those things. Some didn't have any of them.

A design firm called Civilla spent time with the people who actually filled out the form. They watched applicants struggle. They sat in the offices where caseworkers processed the applications. They learned that the form asked the same questions in multiple ways, that entire sections were irrelevant to most applicants, and that the language assumed familiarity with terms that no one outside the system had ever encountered.

They cut it to 18 pages. Eighty percent fewer words. Processing time dropped nearly in half. Harvard recognized it as a Top Innovation in American Government.

The taste that produced Civilla's work looks different from the taste that produces a polished consumer app. Civilla's users are often hungry, stressed, and have no choice but to use the form. The right design decision in that context means fewer questions, plainer language, and the discipline to resist every instinct toward visual sophistication that a designer trained on consumer products would bring.

The contextual verdict at Civilla looked like restraint. Like knowing that a beautiful interface would be the wrong interface, because beauty in that context communicates "this wasn't made for you."

This is what real evaluative judgment looks like in practice. And it has almost nothing in common with the version of "taste" circulating on LinkedIn.

The Pull-Cord

Naoto Fukasawa, the Japanese industrial designer behind many of MUJI's most recognizable products, ran a workshop called "Without Thought" every year for fifteen years. The premise was training designers to observe unconscious behavior. To watch what people actually did with their hands, their bodies, their attention, before anyone asked them a question.

His MUJI wall-mounted CD player uses a pull-cord to start and stop playback. Japanese homes already have pull-cord ventilator fans. The interaction was already in people's hands, already in their muscle memory, already part of the daily gesture vocabulary of a Japanese household. Fukasawa noticed. He designed a product that met the hand where it already was.

A pull-cord on a CD player makes no sense in the abstract. It makes perfect sense if you've spent time in the rooms where people will use it. The design came from sustained, patient observation of how people actually behave, followed by a decision that is legible only if you understand the context it emerged from.

Fukasawa had a word for this. He called it practice. Fifteen years of the same workshop, the same discipline, the same act of paying attention to what people do when they think nobody's watching.

Dieter Rams spent 40 years at Braun and codified his judgment into ten written principles, each specific enough to evaluate against. Vitruvius specified criteria for architectural judgment two thousand years ago, utility, strength, and beauty, applied differently depending on who would live in the building. The people history remembers for their taste wrote it down. They made it examinable. They treated it as a discipline with standards, the way the wine world treats evaluation as a discipline with vocabulary.

The current taste discourse mystifies. It celebrates the verdict while obscuring the process that produced it.

The Uncomfortable Part

Everything above assumes taste is neutral, that it's a skill anyone can develop if they put in the work. There's an older and more uncomfortable problem with that assumption. Pierre Bourdieu spent his career studying taste, and one of his conclusions complicates everything above.

In Distinction, voted one of the most important sociological works of the 20th century, Bourdieu argued that taste is shaped by class and access. Those with cultural capital shape the dominant aesthetic standards. "Good taste" often reflects the taste of people with power, and celebrating it without examining it reinforces the hierarchy it emerged from.

Taste is real, and it's also incomplete without self-awareness about where it came from. When a VC says "taste is the moat," there's a buried assumption. The people with enough career capital and cultural exposure to have developed "taste" will be the ones who survive. That's a class argument in aesthetic clothing, and ignoring it makes the concept less useful, less honest, and less durable.

Bowles raises the same concern from the design side, insisting that we "let go of the notion that only a designer can produce a tasteful solution" and "revel in the ingenuity of the hack and the quick fix." Taste that can only recognize polish will miss the contextual judgment of a caseworker who redesigns her own intake workflow with a paper form and a Sharpie. That caseworker is rendering verdicts too. She just doesn't have the vocabulary, or the LinkedIn following, to call it taste.

The underlying insight is real. When execution gets cheap, the ability to evaluate what's worth making becomes the scarce resource. AI can generate the interface. Whether that interface serves the person sitting in front of it is a question only a human can answer. That gap is real and it matters.

But if "taste" is going to survive its own hype cycle, if it's going to avoid the fate of design thinking and disruption and empathy, it needs to become a discipline with vocabulary, standards, and the humility to examine its own assumptions. It needs to be something you can practice, teach, and hold accountable. Something a sommelier would recognize.

So what would that vocabulary look like?

Consequence awareness. The ability to ask what happens when this goes wrong.
User context sensitivity. Knowing whether the person in front of the screen chose to be there.
Failure-mode judgment. Understanding which failure matters most and designing for it.
Restraint. The discipline to resist a polished interface when a plain one serves better.

Evaluative axes, not abstract virtues. You can teach them, test against them, and argue about whether a specific decision satisfies them.

A compliment celebrates a decision after it's made. A discipline explains how to make one.

Fukasawa spent fifteen years running the same workshop. Civilla spent months in waiting rooms. Rams wrote ten principles and spent four decades testing them. All of them described what they did the same way. Work.

If your definition of taste can't tell you whether a specific decision was tasteful, you have a compliment. And compliments have never kept a concept alive.