The Bottleneck Has Shifted, Not Disappeared
In a world where Software 3.0 lets a founder prototype in hours rather than months, the scarcest thing is no longer the ability to build. It is the ability to see a real problem clearly. Graham's core move — live inside a problem domain until a genuine gap becomes obvious — matters more when execution is cheap, because a hundred other founders can now build the same thing you imagined from a whiteboard. The founder who noticed the problem firsthand has an edge that no agent can replicate.
Toys and Edge Cases Are Where Agents First Fit
Graham observes that breakthrough ideas often look like toys or serve a tiny, desperate audience. That pattern holds for agent-native products today. The use cases where agents shine right now are narrow, specific, and often dismissed as too small. A founder who notices one of those overlooked pockets — because they personally felt the friction — is positioned to build something real before the category looks obvious to everyone else.
Taste Is the Editor's Job, Not the Agent's
When the interface is natural language, the bottleneck moves from typing to thinking. An agent can generate a workflow, draft a feature, or sketch a product surface in minutes. What it cannot do is decide whether any of that is worth building or whether the problem it solves is one that real people actually have. That judgment belongs to the founder. Graham's insistence on organic, experienced-from-the-inside ideas is essentially a description of what good editorial taste looks like before a single line of code runs.
Distribution and Trust Do Not Self-Generate
Graham notes that the best early markets are small groups who desperately need the thing. For an AI-native company, that desperation is the signal worth hunting. But reaching even a small group requires distribution, credibility, and enough trust that users will hand their workflows to an autonomous system. None of that emerges from a clever prompt. The problem-spotting work Graham describes is also the trust-building work — founders who live inside a domain earn the relationships that make distribution possible at all.
