OpenAI embraces the 'platform paradox' by selling API access to startups that compete directly with its own apps like ChatGPT. The strategy is to foster a broad ecosystem, believing that enabling competitors is necessary to avoid losing the platform race entirely.

Instead of building a walled-garden AI, the Zed IDE created the Agent Client Protocol (ACP), allowing any coding agent to integrate. This 'Switzerland' strategy, modeled after the Language Server Protocol, lets Zed benefit from all AI innovation rather than competing against it, even attracting competitors like JetBrains to adopt the standard.

Superhuman Go is not just another AI assistant; it's a platform designed to be the "mass transit" for third-party AI agents. By providing the underlying infrastructure, they enable partners like Radical Candor to embed their unique knowledge directly into users' workflows across any application, a powerful distribution strategy.

The current ecosystem of insecure, community-submitted AI agent skills is unsustainable. The likely monetization path is a trusted, centralized "app store" that vets skills for security, offers them via subscription, and takes a revenue share from developers.

Dreamer's AI "Sidekick" builds apps using the same command-line interface available to human developers. This forced the team to create excellent documentation and a clear API surface, which not only enables the agent but also significantly improves the developer experience for humans, creating a virtuous cycle.

The nascent AI agent ecosystem lacks effective discovery mechanisms for third-party tools ('skills'). This creates an opportunity for curated marketplaces that help users find, vet, and even pay for high-quality, trustworthy agent capabilities, solving a key bottleneck to adoption.

The crypto community often criticizes platforms like Solana for paying partners like Western Union. However, this "pay-to-play" model is a standard business development strategy used by giants like Amazon (for Alexa) and Facebook to bootstrap their ecosystems and kickstart the flywheel with marquee partners.

David Singleton, CEO of Dreamer, reveals that the complex platform—encompassing an agent studio, tool ecosystem, and OS-like architecture—was built by a core team of only about six people. This highlights the incredible productivity and leverage that small, high-talent-density teams can achieve in the AI era.

VLLM thrives by creating a multi-sided ecosystem where stakeholders contribute for their own self-interest. Model providers contribute to ensure their models run well. Silicon providers (NVIDIA, AMD) contribute to support their hardware. This flywheel effect establishes the platform as a de facto standard, benefiting the entire ecosystem.