Taste can't be automated

Everyone's talking about what AI can do.

I want to talk about what it can't.

Last month I watched two founders build the exact same product. Same idea. Same market. Same AI tools. Same weekend timeline. One of them shipped something that felt like a real product. People signed up, poked around, came back the next day. The other shipped something that looked almost identical on the surface but felt like nothing. People opened it, clicked twice, closed the tab.

Same tools. Same speed. Same capabilities. Completely different outcomes.

The difference was taste. And taste is the one thing AI can't generate for you.

Everyone can build now. So what?

Let's start with the obvious. The barrier to building just hit the floor.

A PM with no engineering background can ship a working product in a weekend. A designer can go from sketch to deployed app without writing a single line of code. A founder can test ten ideas in the time it used to take to build one. This is genuinely incredible. I'm not being sarcastic. The democratization of building is one of the most important shifts in a generation.

But here's what nobody wants to say out loud.

When everyone can build, building isn't the advantage anymore. The thing that was hard became easy. And when a hard thing becomes easy, the value migrates somewhere else. It migrates to the thing that's still hard.

Building is easy now. Knowing what to build is still impossibly hard. And knowing what NOT to build might be even harder.

That's taste. That's the whole game.

What taste actually is

People throw the word around loosely. "That product has great taste." "She has good taste in design." It sounds subjective. Almost fluffy. Like it's an innate gift some people have and others don't.

It's not. Taste is a skill. A trainable, developable, sharpening-over-time skill. But it's a weird one because you can't learn it from a course or a book. You develop it through accumulation. Through years of noticing, comparing, absorbing, and caring about why things feel the way they feel.

Let me try to define it concretely.

Taste is the ability to make the right call when there's no data to tell you what the right call is.

That's it. That's the whole thing.

Should the onboarding be three steps or five? There's no A/B test yet. You have to feel it. Should the pricing page show three tiers or two? You have to make a judgment call based on how you want the product to be perceived. Should the notification be a banner or a toast? There's no user research for a product that doesn't exist yet. You decide. And the quality of that decision is taste.

AI can generate both options. It can build three-step onboarding and five-step onboarding in ten minutes. It can design both pricing pages. It can implement both notification patterns. What it absolutely cannot do is tell you which one is right for your product, your users, your positioning, your moment.

That judgment is yours. And it's worth more than it's ever been worth because now it's the only bottleneck left.

The taste gap is widening

Here's what I think is actually happening and most people aren't seeing it.

AI isn't eliminating the gap between people with taste and people without it. It's widening it.

Before AI, a founder with great taste but no technical skills was stuck. They could see exactly what they wanted but couldn't build it. Meanwhile, a founder with mediocre taste but strong engineering skills could ship a working product that was functional but forgettable. The technical founder had the advantage because execution was the bottleneck.

Now? The founder with great taste can execute just as fast. They describe what they want, AI builds it, they refine it based on their instinct for what feels right. Every cycle gets the product closer to their vision. Their taste compounds with every iteration because AI gives them instant feedback loops.

The founder with mediocre taste? AI made them faster at building mediocre things. They ship more, but everything they ship feels generic. Same rounded corners. Same gradient backgrounds. Same component library defaults. Same "clean and modern" aesthetic that means nothing and says nothing. Their speed went up. Their quality didn't.

AI is an amplifier. It amplifies whatever you bring to it. Bring sharp taste and you get a sharp product fast. Bring no taste and you get a generic product fast. The fast part is equal. The sharp part never was.

What AI-generated "taste" looks like

You can spot AI-default aesthetic from across the room now. It has a look. A feel. An uncanny sameness.

Every AI-generated landing page uses the same hero section pattern. Big headline, short subtitle, one or two buttons, gradient background. The colors are always pleasant but never memorable. The spacing is always correct but never intentional. Everything is fine. Nothing is felt.

This is what happens when you let the model make the taste decisions for you. The model optimizes for "not bad." And "not bad" is the enemy of "great." Not bad is invisible. Not bad is forgettable. Not bad is every SaaS product you've signed up for, used twice, and never thought about again.

The products that break through are the ones where a human made a call that the model wouldn't have made.

Stripe's website feels different from every other payments company because someone decided it should feel like an editorial magazine, not a fintech dashboard. That's a taste call. No model would suggest that.

Notion chose a nearly monochrome interface when everyone else was adding color. Someone decided that restraint was the statement. A model trained on existing products would never recommend less color. It would recommend the average amount of color. Average is what models do.

Arc browser put the URL bar at the bottom and made tabs disappear. Someone thought "what if everything the industry assumes about browsers is wrong?" An AI trained on existing browsers would have given you Chrome with slightly different icons.

Taste is pattern-breaking by definition. AI is pattern-matching by design. They're structurally opposed.

The three layers of taste

Here's how I think about it. Taste operates on three layers, and most people only develop the first one.

Layer 1: Visual taste. Does this look good? Is the spacing right? Do the colors work together? This is the layer most people think of when they hear "taste." It matters, but it's the shallowest layer. And frankly, AI is getting decent at this. It can make things look fine. Not distinctive, but fine.

Layer 2: Product taste. What should this product do? More importantly, what should it NOT do? Which features serve the user and which features serve your anxiety about missing something? This layer is where most products live or die, and it's entirely human. It requires understanding people, understanding markets, and having the conviction to make hard cuts.

Layer 3: Narrative taste. What story is this product telling about the world? What does it believe? How does it make the user feel about themselves when they use it? This is the deepest layer. Superhuman didn't just build a fast email client. They told a story about what it means to be someone who values speed and focus. The product made you feel like a certain kind of person. That's narrative taste. No model generates that.

Most builders stay at Layer 1. They make things look nice. The builders who create products people love operate on all three. They make things that look right, work right, and mean something.

How to develop taste (it's not what you think)

You don't develop taste by consuming more content or following more design accounts or reading more product teardowns.

You develop taste by giving a damn. Relentlessly. About everything.

That sounds vague so let me be specific.

Use things and pay attention. Not just products. Restaurants, hotels, packaging, store layouts, the way a good bartender sequences their attention across the bar. Taste transfers across domains. The person who notices why a particular restaurant feels welcoming understands something about onboarding. The person who notices why one book cover makes them pick it up and another doesn't understands something about landing pages.

Make things and be honest about them. Build stuff. Look at what you built. Ask yourself: would I use this if someone else made it? Not "is this functional" but "does this feel like something?" Be brutal. The gap between what you made and what you wish you'd made is where taste lives. That gap is painful but it's the signal. Chase it.

Study the decisions, not the outputs. Don't just look at products you admire. Ask why they made specific choices. Why did Linear use that particular shade of purple? Why did Notion make their sidebar collapsible but not removable? Why did Figma put multiplayer at the center instead of treating it as a feature? Every product is a trail of decisions. Reading those decisions teaches you more than looking at the final result.

Develop strong opinions through building. You don't know what you think about dropdown menus until you've built five different versions and watched users interact with each one. Opinions born from experience are the raw material of taste. People who have "good taste" are really just people who've formed a dense web of opinions through years of making and observing.

Spend time with people who have better taste than you. This is the fastest accelerator. Work with someone whose judgment you admire. Watch what they notice. Listen to what bothers them. Ask them why they'd change something you thought was fine. The gap between what bothers them and what bothers you is your growth edge.

The anxiety is misplaced

I talk to a lot of product people and designers who are anxious about AI. They worry they're going to be replaced. That their skills don't matter anymore. That anyone with a prompt can do what they spent years learning to do.

I get the feeling. I don't share it.

Here's why.

The people who are going to get displaced by AI are the ones whose entire value was execution. If the only thing you brought to the table was the ability to turn a spec into working code, or turn a wireframe into a polished mockup, yes, AI can do that now. That's real. That's happening.

But if what you bring is judgment? If you're the person who looks at a product and knows what's wrong before anyone else sees it? If you're the person who can feel when an interaction is a millisecond too slow or a screen has one element too many? If you're the person who can look at a feature request and know it would make the product worse even though the customer wants it?

You just became the most valuable person in the room. Because everyone around you can now build at the speed of thought, and your thoughts are what determine whether that speed produces something great or something forgettable.

AI made execution free. That made taste priceless.

The products that win from here

Ten years from now, every product will be well-built. AI will handle the engineering. The code will be clean. The interfaces will be functional. Everything will work.

The question is: will it matter?

Some products will matter because a human with taste and conviction shaped every decision. They'll feel intentional. Specific. Like someone cared about every detail not because they had to, but because they couldn't help it. Those products will have users who love them in a way that's hard to explain and impossible to manufacture.

Other products will be technically flawless and emotionally empty. Built fast, built well, built by nobody in particular. They'll work. They'll be fine. And they'll be forgotten the moment something with a soul shows up.

The gap between those two outcomes is taste.

It's the last human moat. And the best part is that unlike every other competitive advantage, you can't buy it, you can't automate it, and there's no shortcut to developing it.

You just have to care more than the next person. For longer. About things that most people don't even notice.

That's the whole game now. And honestly, I think that's beautiful.