Authentic Taste vs. AI Taste

Art doesn’t just give you something to like. It changes what you like.¶

The first time you heard the song that would later become your favorite song, you probably didn’t love it. You noticed it. Something in it caught you. Then you heard it again, and then again, and something in you rearranged around it. A month later, songs you used to like sounded worse. A year later, you couldn’t remember why you’d ever liked them.¶

That’s what art does. It shapes taste. Taste isn’t something you have. It’s something you build, and the thing that builds it is exposure to work that stretches you.¶

Recommendation platforms don’t do this. They can’t.¶

Every recommendation platform is optimized for the taste you already have. Netflix. Spotify. TikTok. Amazon. Instagram. All of them show you things adjacent to what you already clicked. They run the average of what people like you clicked. Their whole architecture assumes your taste is a fixed target they’re trying to hit. So they feed you what fits.¶

That works for the platform. Every next click is more clicks. But it never changes what you like.¶

It can’t help you build real taste.¶

Brian Eno’s friend Jon Hassell used to ask what he called the most important question a person could ask: What is it that I really like? Eno cited it in a recent interview with Ezra Klein, and then he named the environment we’re all in:¶

We are being told probably about 10,000 or 12,000 times a day what we ought to like. It’s very easy in that tsunami of suggestions to forget what we actually do like.¶

The platforms tell you what you ought to like based on what you’ve already clicked. Art tells you what you actually love by revealing something you didn’t know was in you.¶

Suggestion machines average. Art reveals.¶

This isn’t anti-AI.¶

AI can be pointed at averaging what you already like. AI can also be pointed at helping you find what you actually love. Those are different products. The second one is possible.¶

The same computers Eno used to make Music for Airports can be used to build tools that catch the moment something reveals you to yourself, remember it, and let you build outward from it. AI can hold more of your history than any human curator ever could. It can put a piece of work in front of you that no engagement-optimized algorithm would surface, because the engagement metric would flag it as too weird for the version of you it’s been tracking.¶

The technology can do the taste-building work. It just isn’t being pointed there.¶

That’s not an AI problem. It’s a builder problem.¶

If you optimize for engagement, you get more engagement. If you optimize for retention, you get more retention. If you optimize for the small feeling that says this is me, you get something else. Something rarer. Something that compounds.¶

That’s the direction I want to build in.¶

Not more suggestion machines. Tools that help people figure out what they actually like, and then build outward from that revelation.¶

Taste is made, not recommended. I want to build the tools that make it.¶