Prefer to read this newsletter in your browser? Click here.

💡 The Big Idea: In the Age of AI, Taste is the Moat

I was listening to a fantastic recent episode of the Nathan Barry Show with David Perell (creator of Write of Passage and host of the How I Write podcast) last week when he named three reasons people struggle with writing:

1. They don’t have any good ideas.
2. They have good ideas, but lack the courage to say them.
3. They don’t know how to write.

The first one is an input problem, and according to Perell, it’s the most common one. A lot of people, as he put it, “want to be heard more than they have something to say.”

That may sound a little harsh, but I think it’s fair.

Good ideas don’t show up on demand. They don’t appear because you really want them to. They come from being curious about the world around you, capturing the things that resonate, and letting those things collide with other notes and ideas over time before combining them into something new and original.

Nothing interesting going in means nothing interesting coming out.

In my opinion, the second problem isn’t a writing problem at all. It’s a conviction problem. Maybe you doubt whether you have anything meaningful to contribute. Maybe (like me) you think you’re just not creative. Maybe you think you have something worth sharing with the world, but saying it feels risky. So you sand it down until it’s agreeable, safe, and completely indistinguishable from everything else in the feed.

The third one is a craft problem. Structure, clarity, rhythm. The mechanics of moving an idea from your head onto the page in a way that actually lands.

Three very different problems. And if we want to use AI effectively to do our best creative work, we need to understand the difference.

AI Is Coming for #3 (and Only #3)

AI is genuinely good at the mechanics of communication. Give it a clear idea, and it can help you structure, draft, tighten, and polish it. The craft of turning a thought into competent prose — the thing that used to take years to develop — is rapidly becoming a commodity.

And for people who tie the value they bring to their technical ability to write good prose, this can be pretty threatening.

Which is why we should get good at the thinking part.

Here’s how David said it in that podcast episode:

Ask AI for ideas, and you get the most probable, generic ones. That’s not a bug; it’s how LLMs work. A language model compresses toward the representative by design.

But a good idea is, almost by definition, non-average. The thing you noticed and the connection you made with it is what makes a good idea unique.

This is something Austin Kleon unpacks in Steal Like an Artist (the book that finally convinced me I actually was creative). Nothing is completely original. Every new idea is a mashup or a remix of the ideas that came before it. Creativity isn’t conjuring something out of nothing; it’s connecting existing dots in new and interesting ways. Nobody else has your dots — which means nobody else can make the same connections you would make.

And courage? There’s no prompt for conviction. AI can generate ten takes on any topic, but it can’t make you believe one of them enough to put your name on it. If anything, it makes hiding behind forgettable writing easier than ever as you can now publish endless vanilla content without the risk of ever having an actual opinion.

Taste Is the Moat

Look at Perell’s first two problems again, because underneath, they’re related.

“I don’t have any good ideas” is rarely an idea shortage. Ideas are everywhere. Good ideas are rare. The real bottleneck is the ability to recognize which ones are worth pursuing. To do that, you need to know which idea resonates. That’s how you decide which connections actually mean something.

That’s taste.

And the courage problem? Courage is downstream of taste. You’ll discover the courage to say the risky thing only when you trust the judgment behind it.

You have to own it. Conviction is just taste that you’re willing to attach your name to.

So when the communication part becomes a commodity, taste becomes the entire game. Not the ability to write or communicate well. The moat is knowing what’s good, knowing what’s you, and knowing what’s worth saying at all.

And taste is the one thing AI can’t generate. An averaging machine is terrible at finding the non-average. It can hand you more options than any human ever could, but it can’t choose what's valuable or useful for you. Choosing is the job that makes you you. It’s how you develop the taste that sets great writing apart from the LLM-generated average.

To do that, you have to be willing to have an opinion.

David Perell says:

Strong opinions are required for creative tastemaking. In fact, Perell himself says he wasn’t born with good taste. He built it through years of deliberate reps. He would go on weekly New York City museum visits, and he’d walk out asking himself what he absolutely loved, what he absolutely hated, and why.

Taste isn’t a gift you're born with. It’s a skill you develop the same way you develop any other: on purpose, over time.

Which brings us back to the practical question: how do you actually use AI without flattening the very thing that makes your work worth reading?

By treating it like a dial, not a switch

AI is a Dial, Not a Switch

A lot of creators I know are apprehensive about using AI in the creative process.

You can hear the disdain in their voice when they talk about it. They’re not only skeptical about the value it can add, but they’re also usually proud of their completely analog process.

(BTW, I love analog tools. I have a display case full of fancy fountain pens sitting on my desk. But I use analog tools tactically in my workflows to eliminate distractions when I really need to focus.)

On the other hand are the AI bros posting about how their agents are hiring and managing other agents and bragging about how AI does their work for them.

Neither of them are right. But they’re not exactly wrong, either.

The problem is they’re viewing AI as either on or off. Black or white, there’s no room for gray.

But AI isn’t a switch. It’s a dial. And the Perell framework tells you exactly where to set it:

Turn it up on #3. Outlines, rough drafts, structural feedback. Hire AI for the job it’s really good at, then go through the result line by line and own every word. Your voice is what makes your writing worth reading in the first place.

Turn it down for #1 and #2. Not because using AI here is cheating, but because asking it to just generate something from nothing will result in low-quality work. Your taste and your convictions come from a practice of working with and developing your ideas, not a magical prompt you get when you comment below a social post. It comes from capturing what resonates, connecting what you’re learning to what you already know, and developing a point of view you believe enough to defend.

One warning: most people misdiagnose themselves. They think their problem is #3 (“I just can’t write,” or for me, "I guess I'm just not creative") so they throw AI at the mechanics and wonder why the output still feels empty. Perell’s bet is that most people are actually decent on the writing part already; it’s the first two problems they struggle with. So if the drafts come easy now and the work still isn’t landing, your problem was never the writing. It’s your taste that needs the reps.

And no amount of AI can give you those.

The Bottom Line: Taste Has to Be Developed, Not Generated

Here’s something concrete you can do this week: before you write anything, ask which of the three problems is actually stopping you.

If it’s genuinely #3, turn the dial up guilt-free. Then spend the saved time and energy on the hard part: the thinking.

But if you’re honest and it’s #1 or #2 — the ideas aren’t landing, or they’re there and you realize you’re afraid of sharing them — ask yourself why. Here, the dial won’t save you. What you really need is a practice that consistently builds your taste: better inputs, real connections between them, and a point of view you’re willing to own.

(Want a starter rep? Steal Perell’s: the next time you consume anything — a museum, a bookstore, your own feed — note one thing you loved and one thing you hated, and make yourself articulate why. That’s taste training in action.)

AI made saying what you want to say easy.

But having something to say is another thing entirely.

Craft was the moat. Now it’s a commodity. What’s left is taste. That part is, and always will be, yours to develop.

— Mike

P.S. If you want to hear what some of the world's best writers think, I highly recommend the How I Write episode on What Experts Really Think About Writing with AI.