Good taste wins in a world of AI. How do you develop it?
Taste. SO HOT right now.
Good taste exists. Some things are better designed, or more beautiful, than others.
I’m just going to assert that without rigorous defense. You might not agree—“How can you say that Bach is better than Matchbox Twenty? Isn’t that just, like, your opinion, man?”—but you already know that taste exists in practice. Some movies win audiences; some don’t. Some music speaks for centuries, and some is lost. The iPod resonated in a way the Nomad didn’t. Slack succeeded and Hipchat failed, in part because of design.
Beethoven’s late String Quartets will blow your ever-loving mind, if you let them.
If they don’t, I have good news for you: you have the opportunity to discover something incredible. Maybe start with No. 15.
The late quartets actually have an interesting story when it comes to taste. They’re pretty much the last things Beethoven composed, and most people hated them. They’re unusual, sometimes dissonant, exploring musical spaces and modes that had never been explored before. They were called “incomprehensible” and “indecipherable horrors” at the time. Even today, audiences struggle with them.
To others, they are some of the greatest music ever composed. Actually, that’s too small: they are amongst the greatest works of art in human history. Superlatives abound. Schubert: “After this, what is left for us to write.” Schumann: “on the extreme boundary of all that has hitherto been attained by human art and imagination.” Stravinsky: “…will be contemporary forever.”
I had to force myself to listen to the late quartets at first, which is the kind of thing you do when you’re 19 and a philosophy major and you like TS Eliot. Eliot’s “Four Quartets” were supposedly named after, or inspired by, the five late quartets of Beethoven. So I bought a CD and listened. I didn’t really understand them at first, but I listened again, and again.
I started as a dumb kid who thought it would be cool to read Eliot and listen to Difficult Music. A few months later, I was a dumb kid who had somehow been gifted an encounter with the sublime.
I repeat: good taste exists. Something doesn’t have to be objective, or unanimous, to be true.
Good taste wins in a world of AI.
This has become a truism lately. In this case it’s a true truism, because the statement “Good taste wins” has also been true for a long time, regardless of AI. So let’s explore what taste really is and how it’s developed.
The idea here is that AI is going to get really good at writing code, writing prose, designing interfaces, and making movies…but it won’t be good at knowing what to build or write, or what separates a good interface or a good movie from a bad one. We’re all looking for the “safe space” that AI isn’t going to disrupt. "Taste” might be that safe space.
Of course, there’s a pretty important assumption buried in this little truism: that AI will struggle to develop good taste. Let’s Parking Lot that one for the purpose of this post, if you know what I mean. (If you don’t know what I mean, you probably don’t spend hours a day in tech company meetings, and I envy you.)
Anyway.
How do I develop taste?
I’m going to tell you in a few minutes, and there is an answer, and it’s not that hard to understand. But before that, I need to make a short detour to this soapbox over here.
Developing good taste was considered a worthwhile thing to do for a long time. Humans used to think that art had value, and that you build a cathedral at least partly because building a cathedral was a good thing to do, because it was beautiful and impressive.
As recently as the 1960s, classical music was actually quite mainstream and quite popular in the United States, not a niche preference for wealthy octogenarians. Leonard Bernstein was a household name. Aaron Copland wrote movie scores. Families listened to symphonies, live, on the radio.
In the 90s, in a public high school, in North Dakota of all places, I read Milton and Dante and Shakespeare and Aeschylus and Austen and Steinbeck and poetry and history and philosophy and music theory, and it wasn’t just my high school, that’s just what high schools did.
Any guesses where this is going?
It turns out that reading literature isn’t profitable. It won’t teach you a trade, and it certainly won’t make the global financial system spin faster.
Capitalism coexisted with other values systems for quite a while, but it was just too darn successful. Markets have boomed for forty-something years. The average person works longer hours, the average family needs two incomes, and yet financial stability hasn’t increased. Capitalism has been slowly, steadily squeezing out culture. Why waste time reading Shakespeare when what we really care about is the GDP?
Look—I’m a capitalist, in the sense that I would much prefer to live in a capitalist system than a non-capitalist system (and yes, Ocasio-Cortez and Denmark are both capitalist, whatever Fox News might tell you). I think economic development is the most efficient way to combat extreme poverty, and I think innovators should take risks and do interesting things and be financially rewarded for it. But that doesn’t mean one can’t be clear-eyed and honest about its flaws.
The story I’m trying to tell you looks like this.
The System turns art into an elective and humanities into a waste of time, and higher education optimizes for alumni income as its success metric.
Markets spin faster, technology advances, and no one tells Zuck that he’s making choices that are bad for humanity, because they haven’t read history; or if they tell him, no one cares because it’s profitable to not care.
AI portends doom or at least disruption, and human taste is interesting once again.
Well, well, well. Look who’s come crawling back.
I said: how do I develop taste?
OK, this time, for real. Theory first.
Gadamer (20th century German philosopher; awesome) recognized that taste is not just preference. Taste is a way of knowing.
Someone with good taste knows things that others don’t. Think of design; how does a great designer do what they do? It’s not a technique or a method; they aren’t a good designer because they know how to use Figma, or because they have memorized what colors go with what. It’s that they have honed their imagination to see and understand things that the rest of us don’t know.
This is taste, and it turns out taste is actually everywhere, not just in art and design. “Will this essay, or menu, or company, or policy resonate with others?” is a taste question.
Gadamer says that taste is developed through deep engagement with tradition, and in the process, openness to being called into question.
Taste is developed through deep engagement with tradition, history, culture. I like Beethoven at least in part because of tradition. This goes in two different ways. On the one hand, my tradition tells me to like Beethoven, because two centuries of humans have liked his music. So far, so culturally relativist.
But that’s not all. Beethoven is himself responding to a longer tradition and history and culture that you and I also share: the music, art, history, politics, religion, philosophy, habits, practices, and daily life that preceded him. He was formed by the same things that form me, and he turned that formation into something that transformed the centuries after him. I can ignore the subjective preferences of music critics; I can’t ignore the tradition that shapes every thought and value and preference I have, which I share with Beethoven, and which he mastered.
So taste is developed by reading, listening, practicing, and experiencing things, engaging both broadly and deeply.
How do you decide what things to actually read, listen to, taste, practice, and experience? I can’t tell you which way to go, because there isn’t a single answer. Engage with something and your taste-as-a-way-of-knowing will get smarter in that direction. You can improve your taste in comic books or classic literature or Indian cooking or math by engaging deeply with those things, and by engaging in the cultures and traditions behind them. What I can recommend is: find things that are good, and then go deeper.
Developing taste also requires openness to being called into question.
Beethoven called me into question. I thought I knew what good music was, and it was Weezer, Smashing Pumpkins, Ani, Rush (don’t judge me), and some of the more accessible parts of the classical music tradition. String Quartet 15 didn’t fit my categories, and that’s the whole point.
Every time you encounter a new person or a new skill or a new work of art, your horizon has the potential to expand. Your little world gets just a little bigger. Or at least: your horizon expands if you’re open to it.
If you read (say) the book of Genesis through a fixed lens, convinced that you already know what it means, then you aren’t letting the text call you into question and your horizon isn’t going to grow.1
What if I just want to get into Y Combinator and start a cool AI company? I just want to get rich; I don’t care about Steinbeck or Beethoven.
First, you should, because Steinbeck and Beethoven are good.
Second, try not to be so narrowly transactional and utilitarian. If you want to develop taste for the purpose of making money, you’re not actually open to new experiences calling you into question, and you aren’t going to grow.
Third, if you’re annoyed with me for rambling for N thousand words about music and philosophy, this is what you’re waiting for. Develop taste in software, taste in your product domain…and also taste in our shared human tradition.
I’m belaboring the point about “our shared human tradition” because that is, mostly, the point. Who do you think will have better taste: the founder who cares about nothing but software, who works 996 on software, who studied software, who thinks everything that isn’t software is a waste of time? Or the founder who cares deeply about software…but also reads books, has an artistic outlet, studies history, and has friends outside of tech? Taste is developed by going wide, not just by going deep.
If you’re building for users in (say) North America or Europe, then every thought and belief and value and mental model of your users has been deeply conditioned by the same tradition of which Steinbeck and Beethoven are master practitioners, and which they in turn influenced.2 There isn’t a simple, direct link between art and literature and your YC startup, but going deep in the culture and history and art and traditions of your users will give you better taste, and maybe that can help your YC startup.
How do I develop taste in a domain like software?
The answer is the same as in other domains. Dwell deeply in the world of software and in the world of your users. Approach these worlds with openness and a desire to learn. Don’t be a narrow specialist. Find the things that challenge you and engage with them. Let them call you into question.
This is why developers make good startup founders. It’s not because you want your CEO writing code—it’s that you want your CEO to have good taste about software. Product people make good founders for a similar reason.
I taught myself programming in my early 20s. It turns out The System was right, and being a philosophy major doesn’t maximize your income or job opportunities right out of school.3 I cut my teeth on Perl, Ruby, and Lisp. Then I built a bunch of bad apps. Then I built a few ok apps. Then I built something good.
The first thing that I built that was good was a cloud video API called Zencoder. I knew this was a good idea, and it worked. How did I know? Taste. (What you might not know from my LinkedIn profile is I also had a few failed startups before Zencoder. Why did they fail? Deficient taste.)
Part of why I had the right taste at Zencoder was that I’d been in the shoes of our users. I’d been a developer reaching for APIs, video and non-video. I’d used, and built, video systems. And I’d gone wide, studying everything from UX to functional programming to business to humanities.
Let’s be honest: on a scale of L3 to L8, as an engineer I was nothing special. But I’d developed some taste, and it worked.
My taste in software is stale at this point. Software in 2026 has some things in common with software in 2012, but man, a lot has changed. Thank God for co-founders.
So yes: taste can atrophy. Some kinds of taste are relatively timeless, and I think I understand some things better now than I did when I was younger. Other domains change quickly, like software. I mostly stopped shipping software before Kubernetes was a thing. Maybe I’m like the critics who didn’t understand late Beethoven: it was something new, and they didn’t take the time to engage, or they remained closed to it.
I didn’t say it would be easy.
All I think about is AI, all the time. Can you say it in a way I’ll understand?
Understanding grows by assimilating a wider and more diverse range of inputs, especially inputs that challenge prior understanding.4
This is too long; I didn’t read it.
Taste is a form of knowledge.
It is developed by engaging deeply with tradition, while remaining open to being called into question by new experiences.
Tradition is both the narrow thing (say, taste in software or taste in wine) and also the broad thing (the history, culture, philosophy, and art of the humans involved in the software or the wine, including neighboring traditions).
Don’t be transactional about this; if you try to develop taste for the mercenary purpose of building a unicorn, you’re not really going to develop taste.
Note that this isn’t cultural chauvinism, as if “engage with your tradition” means “only read the Western canon.” Reading too narrowly stunts taste, and reading with a closed mind stunts taste. However you define the tradition you have engaged with, there’s always something outside it that can expand your horizon. And in a deeper sense, “my tradition” is never a closed and discrete thing; I actually have a lot in common with (say) a 5th century Mayan peasant, and I can expand my horizon by learning from hers.
If your users are elsewhere—say, rural Africa—you will have better taste if you go deep into those traditions.
I genuinely believe humanities can pay off in the long run, though!
Here is one more way of articulating this idea in a different paradigm. It’s a bit reductive, but I think it works.
Q: What is the thing that actually makes taste-judgments, from a material perspective?
A: A human brain.
Q: How does the brain make those judgments?
A: Neurons fire or something like that. I’m not a neurologist, but I believe this to be true.
Q: Where do those neural pathways come from?
A: Some connections are developed and strengthened based on experience, while others weaken or disappear.
Q: How do you assemble a collection of neurons that are good at (say) understanding art or designing products?
A: Train those neurons on art and products and history in a way that reinforces good connections and weakens bad connections.