Most of What You're Building Will Be Replaced by a Better Model. Here Are the Five Layers Between You and Irrelevance.

If your name isn’t Anthropic or OpenAI or Google, you have a problem. You need to find a space to build in where a better model won’t just erase everything you’ve made. That problem isn’t hypothetical. It’s happening right now to a dozen companies worth billions of dollars.

The AI app builder category looked like one of the safest bets in tech as recently as last year. Lovable just raised $330M at a $6.6 billion valuation. They crossed $400M in annual recurring revenue in February, up from $100M just eight months earlier. 100,000 new projects a day. And the space is still collapsing in on itself, because most of these companies — Lovable, Bolt, Replit, Shipper, and a long tail of smaller players — are thin wrappers around the same foundation models, and their moat is about a week deep.

What’s interesting isn’t who dies. It’s what the survivors tell us about where durable value actually lives on the web. There are exactly five things AI cannot replace, and they’re going to organize the entire future of the internet.

Here’s what’s inside:

• The collapse of the build layer. Why a dozen companies racing to be your AI app builder are mostly trapped.

• The escape hatch. What separates Replit, Vercel, and Notion from the wrapper companies that will die.

• The five durable verticals. Trust, context, distribution, taste, and liability: the things AI structurally cannot provide on its own.

• The real map. How the future web organizes around these five layers, and what it means if you’re building something.

• Your positioning audit and agent-readiness stress test. Two prompts that tell you which vertical you should be building in and whether an AI agent can actually use what you’ve built.

Let me show you how the pieces fit together.