By building a feature that competes directly with startups using its own API, Anthropic demonstrates the "platform risk" inherent in the AI ecosystem. Like Amazon with its Basics line, foundation model companies can observe usage, identify valuable applications, and integrate them, creating a kill-zone for dependent companies.

Building a business entirely on a closed-source API from a major provider like Anthropic or OpenAI is precarious. These platform companies can and do release new capabilities that directly compete with and subsume the functionalities of startups in their ecosystem, effectively erasing their business overnight.

AI model providers like Anthropic analyze usage data from customers to identify lucrative verticals, then launch competing applications (e.g., Claude Design vs. Figma). This commoditizes their partners, posing an existential risk for developers building on these platforms.

The assumption that startups can build on frontier model APIs is temporary. Emad Mostaque predicts that once models are sufficiently capable, labs like OpenAI will cease API access and use their superior internal models to outcompete businesses in every sector, fulfilling their AGI mission.

Startups building on proprietary AI platforms like Anthropic or OpenAI face significant risk. The platform can analyze token usage, identify successful applications, and then launch a competing, integrated feature, as Anthropic did to its partner Cursor with Claude Code.

Gurley notes that major AI model providers like OpenAI and Anthropic are shifting from solely selling API access to building their own applications. This move up the stack signals a fear that being a pure model provider is not a defensible moat and could lead to commoditization.

Cursor once constituted up to 50% of Anthropic's revenue, but Anthropic later competed directly by launching its own coding product. This illustrates the extreme danger for application-layer companies building on foundational models that can easily move up the stack and become competitors.

Unprofitable frontier AI companies are expanding into application-layer verticals like drug development as a defensive strategy. They aim to build defensible, high-margin SaaS revenue streams to prove their business model to investors before their core inference and training services are fully commoditized by cheaper open-source alternatives.

Unlike software bottlenecked by engineering headcount, AI models scale with capital. A frontier model company can raise more than its entire app ecosystem combined, then use that capital to launch competitive first-party apps and subsume third-party developers.