Anthropic: product management on the AI exponential

Cat Wu, Head of Product for Claude Code at Anthropic, published a piece that directly addresses the challenge every PM building with AI faces: the technology improves faster than your roadmap can account for.

The article describes how Anthropic’s own product team has adapted. Features that were technically impossible three months ago become feasible mid-sprint. Roadmaps that assumed stable model capabilities need constant revision. The team has shifted from fixed quarterly plans to shorter planning cycles with explicit checkpoints where they re-evaluate what the latest model improvements make possible.

Wu argues that PMs working with AI need to develop a new instinct — planning for capability improvements that have not happened yet, rather than only building on what exists today. This requires closer collaboration with research teams and a willingness to prototype aggressively, knowing that some prototypes will become viable products when the next model version ships.

For PMs outside Anthropic, the takeaway is practical: build flexibility into your AI product roadmap, because the technology foundation is moving under your feet.