When Anyone Can Build Anything

If you could ship a thousand new features tomorrow, each perfectly engineered and well designed, would you?

The answer is no—but not because you should ship less.

Your competitors won’t slow down, and neither can you. The question is: how do you ship relentlessly without making your product worse?

Even shipping one new major feature every day would overwhelm users. They’d never keep up. Your product would become a maze of capabilities nobody asked for and even fewer people understand. You’d build something simultaneously more powerful and less useful.

Yet the ability to build new major features, nearly instantly, feels inevitable.

The increasing functionality of our tools will outpace the people using them. We don’t like learning new things. People adopt slowly, even when change would help. We rely on autopilot. We stick with what we know works, what we’re used to, what’s become muscle memory.

Think about Slack. Over the years, they’ve added dozens of features: Slack Connect, Huddles, Clips, Canvas, AI summaries, workflow automation, and more. The app is objectively more capable than it was before.

Which ones do you actually use?

If you handed me a much older version of Slack right now, I’d work just fine. I’d send messages, create channels, search history, share files, integrate with other tools. The core experience—the thing that made Slack valuable in the first place—would work exactly the same.

It’s not that their major new features aren’t useful. They are. Some people use some of them. But they’re not integral. Once you’ve built muscle memory around the core experience, adopting new patterns introduces friction. We resist changing our habits.

Your response might be: “Then we solve for discovery. Improved onboarding sequences. Intelligent feature suggestions. Contextual help that appears exactly when needed.”

Sure. And AI will likely solve that problem too—personalized tutorials that generate dynamically, context-sensitive help that adapts to individual workflows, smart recommendations tuned to usage patterns. Even with these, tuned to optimize the peak amount of newness one can absorb, users will still be well behind on what a product will eventually be able to do.

There’s a limit to how many new things people will tolerate learning, regardless of how easy you make it. People have a finite appetite for change. Make it as simple as you want—there’s still a hard limit to adaptation.

You might think chat interfaces sidestep this entirely. “Just ask for what you need—the AI handles the rest.” But chat has a hidden assumption: the user knows what’s possible. It’s great for executing intent, terrible for shaping it. Most chat interfaces are already plastering reminders across their UI—“Now with better image generation!” “Try uploading a document!"—because without them, users would never discover these capabilities organically. You can only prompt for what you already have in mind. And even when you learn something new, there’s a second hurdle: remembering to reach for it later. A chat could surface features proactively—mentioning a capability in context, exactly when it’s relevant. But relevance in the moment doesn’t guarantee recall in the future. That might scale to a few features. Not a few thousand. Chat doesn’t solve discovery. It relocates it—from navigating menus to maintaining a mental inventory of what’s even possible.

The limiting factor won’t be what you can build, but what users can actually absorb and reliably reach for.

So what then? What’s the alternative to shipping an infinite number of features each day? I’m still figuring that out; we all are. But I think the question itself is the starting point. If I wouldn’t launch a thousand features tomorrow, what would I do instead? And what happens when my competitors can ship just as fast? How do you design a product today that’s ready for a world where capability is no longer the bottleneck? Working backwards from that question, a few principles start to emerge:

Personalize ruthlessly

• To scale capabilities, you have to shrink surface area. Amazon figured this out years ago. Your Amazon isn’t my Amazon. New features surface naturally because they’re relevant to you specifically. Think about carousels: you learn the interaction once—scroll horizontally, click to explore—but Amazon can fill them with anything. New product categories, seasonal themes, personalized groupings. The UI remains predictable; the possibilities inside it infinite.
• Photoshop pioneered this in productivity tools—try using someone else’s custom workspace. It’s nearly impossible because there are so many settings and customized toolbars.
• The paradox: to offer more, you must show less. Most software is designed for the average user. Soon, every product will need to be designed for each individual, continuously adapting to their data, context, and habits. No two desks are alike, and soon, no two interfaces will be either.

Earn being the default

• Finding features matters less when AI can answer “Can this tool do X?” or users get an agent to do it for them. Discoverability becomes irrelevant—capability becomes the game.
• The real challenge: be the best in your domain. Be the tool users trust enough to keep reaching for. Stripe owns payments. Figma owns collaborative design. Excel owns spreadsheets. When someone needs that thing, they don’t evaluate alternatives—they just open the app they already know.
• But there’s a new dimension: be the default for agents too. As AI handles more tasks on users’ behalf, those agents will choose which tools to use. They’ll pick based on technical factors—reliability, API quality, speed, cost—and practical ones: what data it can access, what permissions are already in place, and which tool dominates the category. The same instinct that makes a user say “just use Stripe” will be encoded into the agents acting for them. You’re no longer just earning a habit—you’re becoming the default answer for software making decisions at scale.
• Your goal: be the obvious choice for that one thing, whether a human or an agent is choosing.

Deepen your moats

• When anyone can build features instantly, the differentiators become the things AI can’t replicate quickly. Your network. Your data. Your customer relationships. This isn’t new but has never been more important.
• Moats can compound over time. They create switching costs that AI can’t eliminate.
• The companies that win won’t win because they build more—they’ll win because users can’t leave even when competitors offer identical features.

We’re entering strange territory. For the first time, the constraint won’t be what’s technically possible. It’ll be what’s humanly manageable.

The companies that master this will build products that feel simultaneously simple and infinitely capable. The ones that don’t will build everything and achieve nothing.

What happens when you can build everything? You learn that building was never really the hard part.