The useful distinction is not "AI product" versus "non-AI product"
The current catalog already includes products with AI somewhere in the story. The stronger editorial cut is narrower: products whose core value is an AI agent, an AI employee, or an autonomous workflow.
That narrower definition is what makes the current Agent founders layer worth keeping. It does not treat every AI-enabled app as evidence. It uses a tighter bar and only includes products that make the agent workflow the point of the product.
The early catalog already shows multiple agent shapes
Coursekit points toward embedded agents inside a founder-owned product. Parsewise compresses document-heavy work into a research workflow. Aident AI Beta 2 and Golf both lean into operational orchestration. Willow Voice for Teams pushes toward voice-first workflow. Hermit treats persistent assistant behavior as the main reason to use the product.
That is enough variety to support one real claim: the current approved catalog is already broad enough to justify an agent-native discovery layer. It is not broad enough to claim full market coverage, and the site should stay disciplined about that.
The stack signal is still the moat
The reason this directory can be more useful than a generic agent list is the stack evidence behind the products. Claude and Cursor show up inside the current set. Supabase and Vercel still appear as a repeated base layer. Framer also shows up around products that move quickly on the surface layer.
That does not mean there is one canonical stack for agent-native founders. It means the category is already producing repeated implementation patterns worth inspecting through the underlying product and tool pages.
Browse the directory: Claude products → · Cursor products → · Supabase products → · Vercel products → · All agents →
That founder angle is important because it keeps the catalog closer to execution than marketing. Founder-led products tend to reveal a lot through their tool choices: they optimize for shipping speed, they reuse a small set of dependable platforms, and they expose where the AI or agent behavior is actually central to the product. In practice, that gives the directory a better editorial cut than broad “AI company” coverage. It favors products that are clearly opinionated about workflow and implementation, not just products that added AI copy to an existing landing page.
It also creates a stronger browse loop. If a visitor starts on an agent-native founder article, then moves into a product page, then into tool pages like Claude, Cursor, Supabase, or Vercel, they should come away with a sharper sense of how real teams are assembling these businesses. That is the kind of grounded utility the blog layer should reinforce.