At 12:47 on a random day(night), I found myself comparing two lamps.
They were, to the untrained eye, the same lamp. Both were small. Both were vaguely Italian-looking. Both gave off the emotional energy of a dentist’s waiting room that had recently discovered jazz. One cost a reasonable amount of money. The other cost enough that I briefly wondered whether it came with a minor European title.
I sent screenshots to a friend.
“The second one has more taste,” she replied.
Not “it looks better.” Not “the shape is nicer.” Not even “buy that one.” It had more taste, as though taste were a rare mineral embedded in the ceramic, detectable only by people who own at least one chair nobody is allowed to sit on.
I stared at the lamps for another twenty minutes, which is how I lost custody of my time.
The next morning, the word was everywhere. A founder was talking about taste in product design. A writer was explaining why taste would survive artificial intelligence. A creative director was posting about “training taste.” Someone on a podcast said execution was becoming cheap and taste was becoming valuable. I opened another app and was informed that a beige kitchen had “timeless taste,” despite looking like a place where even the salt had signed an NDA.
Apparently we have entered the Age of Taste.
This is surprising because, until recently, taste was mostly something your aunt accused you of lacking when you wore sneakers to a wedding. Now it is a career strategy, a technological moat, a personality, a leadership principle, a content category, and possibly the last defense of the human soul.
Naturally, I became obsessed with why everyone else was obsessed.
The Word That Ate the Internet
The current taste fever did not appear from nowhere, although it did arrive with the usual confidence of a person entering a conversation halfway through and immediately correcting everyone.
In early 2026, tech people began arguing publicly that when artificial intelligence makes production easy, the scarce skill will be deciding what deserves to exist. Paul Graham revived an argument he had made years earlier about taste among makers; OpenAI president Greg Brockman called taste a new core skill; and the discussion ricocheted through the usual online corridors where a thought begins as a reasonable observation and ends twelve hours later as a personal brand.
Kyle Chayka, writing in The New Yorker, noticed that “taste” had become a Silicon Valley cliché on the scale that “disruption” once was. Companies and founders were using it to describe everything from product judgment to recommendation tools, design software, and the ability to pick profitable ideas before everyone else. He called part of this phenomenon “taste-washing”: wrapping automation in the language of human sensitivity so the machine feels less like a machine and more like a thoughtful person who buys olive oil in an opaque bottle.
This explains the suddenness. AI did not invent our interest in taste. It created a panic around it.
For years, the internet rewarded output. Post more. Ship faster. Publish daily. Make a video, then make the same video vertically, then crop it into a square, then turn the square into a newsletter, then stand in a field and explain the newsletter into a tiny microphone.
Generative AI looked at this exhausting arrangement and said: I can produce infinity before lunch.
The moment supply becomes infinite, selection becomes theatrical. The person who can say “not that, this” begins to look powerful.
Taste is what we call that pointing finger when we want it to sound profound.
We Have Been Here Before, in Worse Hats
The funny thing about taste is that it always pretends to be private while behaving like a group project.
I like this painting. I hate that sofa. This album changed my life. Those shoes are a crime. We describe these reactions as if they emerged from a sealed chamber inside the self, untouched by class, money, geography, friends, advertising, childhood, envy, convenience, and the fact that an attractive person once wore something similar near us.
History is less romantic.
In 1899, Thorstein Veblen wrote about conspicuous consumption: people buying and displaying things partly because those things communicate status. He was especially interested in objects whose uselessness or expense was the point. A difficult-to-maintain possession tells the world that you can afford difficulty. This remains the governing philosophy of white sofas, tiny handbags, and any sink that requires special instructions.
Decades later, Pierre Bourdieu’s Distinction argued that taste helps organize social class. Preferences are not merely preferences; they become signals of education, upbringing, access, and belonging. Knowing which wine, book, neighborhood, composer, restaurant, or deliberately ugly jacket to approve of can function like a password. Cultural capital is wealth that has learned to speak softly.
This does not mean nobody genuinely loves difficult jazz. Some people do. I admire them and fear their car journeys.
It means taste has always done two jobs at once. It gives us pleasure, and it places us socially.
The internet simply put the placement system on a dashboard.
You no longer need to invite someone into your home to reveal your preferences. Your bookshelf is photographed. Your meals are archived. Your music arrives in an annual graphic report. Your clothes appear in tagged posts. Your saved folders contain enough evidence to reconstruct you after a minor disappearance.
Taste used to leak out accidentally. Now it has a content calendar.
The Factory Can Make Anything Except a Decision
There is a basic economic reason taste suddenly feels important: production is getting cheaper while attention remains stubbornly mortal.
A model can generate twenty logo directions while I am still deciding whether “direction” means anything. It can create a hundred images, draft ten headlines, suggest fifty names, redesign the landing page, and produce a presentation in which every slide appears to have been raised by the same highly competent aquarium.
This is genuinely useful. It is also a problem.
When you have two options, choosing is easy. When you have two thousand, choosing becomes the work.
The Industrial Revolution mechanized muscle. Software mechanized calculation. Generative AI is mechanizing variation. It can keep rolling the cultural dice without becoming tired, insecure, hungry, or distracted by a lamp.
But abundance does not remove judgment. It increases the number of moments in which judgment is required.
This is why “taste” has become such a convenient word. It bundles several abilities together: noticing quality, understanding context, rejecting almost-right ideas, sensing when something is overworked, predicting what people may care about, and recognizing the difference between surprising and merely strange.
That last distinction is important. A swan wearing office trousers is strange. A swan calmly waiting in the lobby for an interview is surprising. Context is doing the heavy lifting.
Machines are excellent at producing candidates. Humans are now being told their special role is to stand beside the conveyor belt and look disappointed.
This is not nothing. Editing is real work. Direction is real work. The ability to kill an idea you spent three days making is real work, although it feels less like work and more like burying a small horse.
Still, I notice how flattering the new story is. We were afraid AI might make creative skill less valuable, so we renamed the remaining part “taste” and gave it a velvet rope.
Curation Became a Full-Time Personality
Before the AI boom, social media had already spent years turning taste into performance.
The old internet asked what you liked. The newer internet asks whether what you like can be arranged into a recognizable aesthetic with a searchable name.
A person once owned books, a coat, and a coffee mug. Now these objects may belong to dark academia, coastal grandmother, quiet luxury, cluttercore, clean girl, office siren, or some other temporary republic whose constitution is mostly lighting.
The aesthetic label performs a useful service. It converts the terrifying complexity of being a person into a shopping filter.
You do not need to determine what kind of home feels meaningful to you. You can type two words and receive seventeen lamps, six linen cushions, a playlist, a skincare routine, and a woman whispering that your current cutting board is blocking abundance.
TikTok and similar platforms explicitly personalize feeds based on what users watch and engage with; Spotify advertises recommendations tailored to individual taste. These systems are designed to learn preference quickly and keep supplying adjacent material.
That is convenient. It is also how taste can begin to feel like a room with invisible walls.
You click one video about Japanese stationery. Three days later, the internet has decided you are a minimalist architect who eats pears beside a concrete window. You listen to one melancholic folk song and your recommendations begin preparing you for a divorce you are not having.
The feed does not exactly tell you what to like. It notices what held you for 1.8 seconds longer and constructs a tunnel.
Then we mistake the tunnel for a personality.
The Algorithm Knows Me, Which Is Rude
Personalization creates an odd psychological loop.
The system recommends something because it resembles what I liked before. I encounter it repeatedly. Familiarity lowers the friction. I like it. The system records this as proof that it understood me. Soon both of us are congratulating the algorithm for predicting a preference it helped manufacture.
This is not mind control. It is much more mundane, which makes it harder to notice.
Psychologists use the term “mere exposure effect” for the tendency to develop more positive feelings toward things as they become familiar. Repetition is not all-powerful, but it can gently sand down resistance. The fifth time you see a shoe, it has stopped looking absurd and started looking “interesting.” By the twelfth appearance, you are checking whether it comes in your size.
Recommendation systems complicate this further because they optimize for behavior, not for the full, mysterious health of your inner life.
A field experiment on Spotify found that personalized podcast recommendations increased listening but reduced the diversity of each individual user’s podcast consumption, even while aggregate diversity across users rose. In other words, the whole platform could become more varied while each person’s private corridor became narrower. We are all unique together, in separate little tubes.
Researchers studying cultural recommendation have argued that these systems shape not only individual choices but shared cultural experience: what populations encounter in common, what becomes legible, and what disappears into private personalization.
This matters because taste requires friction.
You need to encounter the book with the ugly cover, the song your friend plays too loudly, the movie that starts badly, the restaurant you would never have selected, the painting you resent for six months before it suddenly opens.
A perfectly personalized world protects you from irrelevant things. Unfortunately, irrelevant things are where half of a person comes from.
Taste Is Status Wearing Better Lighting
There is another reason everyone wants taste now: wealth has become visually confusing.
Luxury used to announce itself with material excess. Gold. Marble. A carriage requiring several horses and one emotionally exhausted footman.
Modern status often works through restraint. The expensive object looks plain. The logo disappears. The restaurant has no sign. The hotel lobby contains one rock. The website uses a font so understated it appears embarrassed to be there.
When products become easy to copy, knowledge becomes the signal.
The impressive thing is not owning an object. It is knowing which object to own before it becomes obvious.
This is where taste turns into social timing.
Too early and you look deranged. Too late and you look influenced. The ideal moment is approximately six minutes before everyone else, which is why contemporary culture feels like a permanent airport sprint.
Social media intensified this because it made preference measurable. Followers, saves, reposts, waitlists, resale prices, and sold-out notices turn aesthetic judgment into visible proof. A person with taste is not merely someone who likes good things. They are someone whose liking appears to precede demand.
Silicon Valley understands this instinctively. Venture capital is partly a profession of placing large bets on tomorrow and then explaining that the bet was obvious. “Taste” gives financial prediction a romantic glow. It makes picking companies sound less like spreadsheet work and more like standing in a gallery with one hand behind your back.
Tech’s version of taste often slides toward profitability: discernment becomes the ability to identify what will win.
That is useful, but it is not the whole thing.
Some of the best taste is commercially disastrous.
It loves an unmarketable film. It keeps a chipped cup. It wears the coat everyone hates. It follows an artist through the bad period. It refuses the frictionless, optimized choice because something inconvenient feels alive.
Real taste sometimes makes you less legible.
Personal branding hates this.
Can a Machine Have Taste, or Just Excellent Manners?
The AI industry has already begun moving beyond the claim that humans will supply taste.
Now some companies suggest the machines may have it too.
In April 2026, Axios reported that Anthropic described a new model as more “tasteful and creative” in professional outputs such as interfaces, slides, and documents. Other companies were also pursuing AI systems framed around aesthetic judgment and cultural fluency. The debate immediately became philosophical, commercial, and slightly silly, because “Can a machine have taste?” sounds like a question asked by a French robot while smoking indoors.
There are two different questions hiding here.
Can a machine predict what humans will judge as tasteful? Obviously, increasingly, yes. It can learn patterns in highly rated designs, successful products, acclaimed writing, popular images, and expert preferences. A 2026 research paper even framed “scientific taste” as the ability to identify and propose potentially high-impact research ideas, training models with community signals such as citation patterns.
But can a machine care?
Taste, as humans experience it, is not only ranking. It is biography.
I like a song because my father played it in the car. I hate a fabric because it resembles a sofa from a childhood waiting room. I trust a typeface because it reminds me of library books. I reject a perfectly designed café because the chairs seem hostile to the lower back.
These are not necessarily noble reasons. They are barely reasons. They are sediment.
A model can simulate the result without possessing the sediment.
Of course, the market may not care. If the machine consistently chooses what people love, companies will call that taste before philosophers have finished clearing their throats.
And then the situation becomes beautifully circular: AI learns from human culture, humans consume AI-curated culture, AI learns from the resulting behavior, and eventually everyone’s living room contains the same emotionally supportive mushroom lamp.
The Secret Ingredient Is Rejection
Whenever taste becomes a fashionable skill, people begin asking how to acquire it, usually in the same tone used to ask how much protein is in an almond.
The disappointing answer is exposure, attention, comparison, and making things badly for longer than feels dignified.
Paul Graham’s 2002 essay “Taste for Makers” treated taste as something that can improve through practice rather than a mystical gift. He connected good work with qualities such as simplicity, durability, solving the right problem, and learning from strong examples.
I think the most important part is rejection.
Taste is not just the ability to recognize what you like. Toddlers can do that with alarming clarity. Taste is the ability to explain why one choice serves the thing better than another and then remove the weaker choice even when it is attractive, expensive, trendy, or yours.
This is why experienced creative people often look less magical up close. They are simply less easily impressed.
They have seen enough to know when an idea is pretending. They recognize borrowed gestures. They can feel when polish is compensating for emptiness. They know that adding another feature, color, metaphor, ingredient, camera angle, or distressed serif will not save the central problem.
Good taste often looks like restraint because restraint leaves fewer places to hide.
But restraint is not the same as minimalism. A maximalist can have extraordinary taste. A Baroque church is not suffering from a lack of editing. It has simply edited toward thunder.
Taste means coherence, not beige.
This is where the internet’s version goes wrong. It often treats taste as a visible collection of approved objects. But the objects are downstream. Taste is the invisible pattern of decisions: what you notice, what you connect, what you refuse, and what you are willing to love before it has social proof.
You cannot buy the pattern directly.
Many companies are devastated by this information.
The Plot Twist Nobody Ordered
So why is everyone obsessed with taste?
Because we are drowning in competent options. Because algorithms have made our preferences easier to predict and harder to locate. Because social media turned liking into signaling. Because AI threatens the pride we once attached to execution. Because “taste” sounds more glamorous than “being the person who deletes nineteen generated drafts.” And because, underneath all the branding, we are genuinely afraid of sameness.
That fear is not imaginary. Generative systems learn from existing culture and can produce polished averages at enormous scale. We all know that AI-generated design is already developing recognizable clichés, just as earlier algorithmic platforms encouraged cultural flattening.
The reaction has been predictable and human. As slick machine-made imagery spreads, some artists have moved toward visible imperfection: handwriting, rough surfaces, awkwardness, evidence of the hand.
We always run toward scarcity.
When objects were scarce, possession signaled status. When information became abundant, curation did. When production becomes abundant, judgment does. And if judgment becomes automated, perhaps the next luxury will be inconsistency.
A badly composed photograph. A recommendation with no data behind it. A friend saying, “I don’t know, this reminded me of you.” The future of taste may not be perfect discernment. It may be an attachment that cannot be efficiently explained.
Which brings me back to the lamps.
The Lamp, Reconsidered
I bought the cheaper one.
This was not an act of rebellion. I would love to tell you I rejected status performance, escaped algorithmic influence, and reclaimed my sovereign aesthetic identity.
In reality, the expensive lamp sold out while I was thinking.
When the cheaper lamp arrived, I placed it on the desk and waited to discover whether I had taste.
Nothing happened.
The room did not become more culturally literate. No invisible committee approved me. The lamp simply made a warm circle on the wall. It looked slightly strange. Then familiar. Then, after a week, like it had always been there.
And that, I think, is the part missing from most conversations about taste.
Taste is not only the dramatic moment of choosing. It is the slower process of living with a choice, noticing what it does to your attention, and allowing affection to grow without immediately converting it into proof of who you are.
The internet wants taste to be visible. Business wants it to be predictive. AI companies want it to be programmable. Status wants it to be exclusive.
But personal taste may be most honest when nobody is watching.
It is the lamp you keep after the trend passes. The song you do not post. The ugly cup your hand reaches for first.
The thing you love after it has stopped helping you explain yourself.
