Every piece of startup advice eventually collapses into the same mantra: focus. One feature. One product. One type of customer. One marketing channel. Do one thing and do it well. It sounds so clean. So obvious. And yet, I keep catching myself drifting from it, especially now, when building has never been cheaper or faster.

The advice is everywhere

Peter Thiel argues in Zero to One that startups should dominate a small market before expanding outward in concentric circles. Paul Graham says the most important thing for startups is to focus, because there are so many things you could be doing, and one of them is the most important. The Founder Institute calls it the "Rule of One": one problem, one customer, one solution. Kevin Kelly's "1,000 True Fans" essay makes the case that you don't need millions of customers, just a thousand who deeply care. The logic is airtight. If you spread yourself across multiple customer segments, multiple features, and multiple channels, you end up mediocre at everything. You dilute your message. You confuse your audience. You burn resources chasing breadth instead of depth. I know this. I've read the books. I've nodded along to the podcasts. And I still struggle with it.

Why it's harder now

Here's what's changed: the cost of shipping has collapsed. AI tools have made it possible for a solo founder to build in a weekend what used to take a team months. Cursor hit $100 million in ARR with 20 people. Midjourney built a $200 million business with 40. The barrier between "idea" and "feature" has never been thinner. And that's exactly the problem. When building is cheap, the temptation to build more is overwhelming. You think, "Well, it'll only take a day to add this." Then another day for the next thing. Then another. Before you know it, you're maintaining six features for three different types of users across two channels, and none of them are working particularly well. I'm a victim of this myself. The voice in my head says: one feature isn't enough anymore. If shipping is so easy, surely the bar has risen. Surely competitors will catch up overnight. Surely you need to offer more to stand out. That voice is persuasive. It's also mostly wrong.

The cost of building isn't the cost that matters

There's an important distinction that gets lost in the "AI makes everything cheap" narrative. Yes, the cost of building has dropped. But the cost of distribution, of actually getting your product in front of the right people and earning their trust, has not. If anything, it's going up. When everyone can ship a product, attention becomes the bottleneck. As one analysis from Dockyard Capital put it: "When everyone can build the product, only distribution differentiates." The winners aren't the ones with the most features. They're the ones who understood one customer so well that their product felt inevitable. This is where the "just one" philosophy actually becomes more important, not less, in the age of AI. You can build anything, so the question is no longer "can I?" but "should I?" And "should I" requires a level of focus that cheap shipping actively undermines.

One is a constraint, not a limitation

The Unix philosophy, dating back to the 1970s, says it plainly: write programs that do one thing and do it well. Half a century later, the principle still holds, not because technology hasn't advanced, but because human attention hasn't. Your customers don't want another product that does twelve things adequately. They want one that solves their specific problem so well they don't have to think about it. The "Rule of One" in marketing captures the same insight from the demand side: one audience, one message, one call to action. When you make ideas or messaging complex, people tune out. The discipline of "just one" isn't about doing less for the sake of minimalism. It's a constraint that forces clarity. One feature means you have to pick the right one. One customer means you have to actually understand someone. One channel means you have to get genuinely good at it instead of spreading thin across five platforms with half-hearted efforts.

The MVP trap in the AI era

There's also a subtler trap worth naming. The traditional advice says: start with the smallest MVP, ship one core feature, learn, iterate. That's still sound. But in 2026, MVPs ship so fast that founders often skip the "learn" step entirely. They ship, see mediocre traction, and instead of going deeper on the problem, they add another feature. Then another. Some are now arguing that we've moved past MVPs entirely, that the bar has risen to what's called a "Minimum Lovable Product." Users in 2026 expect polish from day one. But polish doesn't mean more features. It means the one thing you do works beautifully. It means the experience is considered, not just functional. The instinct to add more is a symptom of not going deep enough on the one thing.

What I keep reminding myself

I don't have this figured out. I catch myself reaching for the "just one more feature" lever constantly. But every time I zoom out, the pattern is the same: the things that work are the things where I committed to one bet and executed it with conviction. The products I admire most, the ones I keep coming back to, are the ones that feel opinionated. They chose a customer. They chose a problem. They said no to everything else. That clarity is what makes them feel solid, like someone actually thought about this instead of just shipping whatever was possible. So here's the reminder, mostly to myself: the cost of shipping is low, but the cost of scattered attention is as high as ever. Just because you can build it doesn't mean you should. One feature. One product. One customer. One channel. Easier said than done. But that's the whole point. If it were easy, everyone would do it, and it wouldn't be a competitive advantage.

References