AI markets are currently in a state of chaos, characterized by numerous players across categories like code generation and customer support, yet lacking clear differentiation. This environment doesn't resemble established software markets, which have defined boundaries and stable technology. Instead, these are proto-markets—early stages where demand exists but no clear product specifications or competitive advantages have formed. Despite generating significant revenue, these markets are structurally incomplete and require a different approach for success. Winning in these proto-markets hinges on product plasticity and learning velocity rather than the traditional SaaS strategy of establishing moats and margins first. Companies need to be adaptable, using real user behavior to inform product development. For example, Anthropic's Claude Code started as a simple utility, evolving through user feedback into a more complex product with features like Plan Mode and Agent Skills. This iterative process not only enhances the product but also builds a foundational understanding of user needs, which will be critical as the market matures. Key characteristics of AI proto-markets include blurred market boundaries, continuous product evolution, unsettled economics, competition against existing solutions rather than direct rivals, and a bottoms-up adoption model. Many AI applications are still priced below cost, and users often compare AI tools to non-existent alternatives. These factors create an opportunity for early adopters to shape products based on actual usage patterns rather than preconceived notions, fostering a competitive edge that can evolve into durable market advantages as the landscape stabilizes.