The App Store Moment for AI

Why the real AI competition isn’t about models. It’s about who controls the orchestration layer and the future marketplace.

Anthropic is winning the developer race right now. Claude Code and Cowork have real momentum, not just because the models are good, but because the orchestration layer around them is good. The agentic loops, the tooling, the tight integration. And they can subsidise credits because they own the infrastructure.

But it’s fragile. If a better orchestration interface showed up tomorrow, one that plugged into Claude, GPT and Gemini, most users would switch without thinking twice. There’s little brand loyalty here. People care about one thing: does it get the job done?

And “the job” is different for everyone. A developer needs different orchestration than an accountant or a lawyer. Can LLM providers really be all things to all users? History says no.

The Smartphone Parallel

When Apple launched the iPhone, they had no intention of letting third-party developers build apps. But users kept jailbreaking their phones, signalling unmistakable demand. Apple eventually opened the App Store, and at a 30% cut, it became one of the most profitable platform decisions in tech history.

LLM providers are at that same crossroads. Yes, marketplaces are emerging. MCP servers, plugins, skills. But most are still limited to passive lookups and lightweight integrations. The real unlock is opening orchestration APIs so developers can build capabilities that feel native, not bolted on. More developers mean more users, which means more developers. The flywheel Apple discovered.

“More developers mean more users, which means more developers. The flywheel Apple discovered.”

The Browser Economics

Browsers are free. They’ve been free for decades because Google pays Apple billions a year to be the default search engine in Safari. The browser isn’t the product. It’s the distribution channel.

Map that onto AI. An orchestration layer that millions use daily becomes incredibly valuable real estate. LLM providers will incentivise orchestrators to default to their model through subsidised credits, revenue sharing, and preferential API access. The orchestrator becomes the new browser. The LLM becomes the new search engine. The economics of “free” start to make a lot of sense. LLM owning the orchestrators is a strong business model.

“The orchestrator becomes the new browser. The LLM becomes the new search engine.”

The Lock-In Question

But will consumers actually want choice? Or will they pick an ecosystem and stay?

Most people chose iOS or Android years ago and never looked back. The switching costs aren’t just technical. They’re emotional. Your workflows, your habits, your muscle memory. Once an orchestration layer knows how you work, leaving feels expensive.

So maybe the future looks more like mobile: two or three dominant ecosystems, each with its own orchestration, marketplace, and loyal base. That’s the real strategic play. Not just building the best model, but building an ecosystem sticky enough that users don’t want to leave.

Could a New “Browser” Win?

What if someone built a third-party orchestrator from scratch? The way in would almost certainly be vertical, not horizontal. Don’t try to out-general-purpose Anthropic or OpenAI. Build the best orchestration layer for lawyers, accountants, or supply chain managers. Own a niche so completely that the generalists can’t compete on depth. Then expand.

The other path is the aggregator play: model-agnostic, compete purely on orchestration quality. Harder without the subsidy advantage of owning your own model, but if the experience is meaningfully better, users might come.

“Don’t try to out-general-purpose Anthropic or OpenAI. Own a niche so completely that the generalists can’t compete on depth. Then expand.”

Where This Is Heading

The endgame isn’t one winner. It’s layers. LLM providers, orchestration interfaces, and marketplaces, each competing on its own merits, sometimes bundled, sometimes independent. You’ll pick the orchestrator that fits your workflow and the model that fits your needs. Sometimes they’ll come from the same company. Maybe they won’t. Hard to tell yet.

The providers who build their own App Store moment will own the platform. The ones who try to control everything will watch someone else build it for them.