Mike Joined Anthropic to Shape AI's Future

At Anthropic, Mike Krieger has witnessed a dramatic evolution in AI capabilities that has fundamentally shifted his perspective on what's possible and how quickly it's arriving. His experience leading product at one of AI's frontier companies offers valuable insights for both leaders and individual contributors navigating this rapidly changing landscape.

Mike has radically changed his view on AI's creative independence. Initially skeptical about AI having original perspectives, his experience with Claude Opus 4 transformed his thinking when it provided genuinely novel angles on his product strategy documents. This wasn't just incremental feedback but truly new thinking that changed his approach. This shift happened much faster than he anticipated, reinforcing his belief that AI timelines are accelerating beyond most people's expectations.

The pace of development has profound implications for how teams should operate. At Anthropic, approximately 95% of Claude Code is written by Claude itself, creating unprecedented productivity but also new bottlenecks. While code generation has become dramatically faster, the team found themselves constrained by infrastructure like merge queues that weren't designed for this volume of output. This suggests teams should prepare for AI acceleration by strengthening systems that surround the actual creation process.

For product development, Mike observes that AI shifts bottlenecks toward the beginning and end of the process. The constraints become: deciding what to build, aligning stakeholders, and packaging work into coherent releases that users can understand. The middle part—implementation—becomes dramatically faster. This means product teams should invest more in strategic thinking and user education rather than implementation details.

Mike's approach to working with researchers offers a crucial lesson for cross-functional collaboration. Rather than separating product people between those working on the model and those working on user experience, he found dramatically more leverage by embedding product people directly with researchers. The most valuable work happens at the intersection of research capabilities and product thinking, not by treating them as separate domains.

For individual contributors, this suggests seeking opportunities to work across traditional boundaries rather than staying in specialized lanes. The most impactful work happens when you can translate between technical capabilities and user needs, not when you focus exclusively on one domain.

As AI capabilities grow, Mike believes product teams will continue to provide critical value in three areas: making AI comprehensible to users, developing winning strategies for where to play in the market, and opening people's eyes to new possibilities. The gap between what AI can do and how people currently use it remains enormous—creating opportunities for those who can bridge this divide.

This perspective challenges both leaders and ICs to reconsider where they focus their energy. Rather than optimizing for implementation speed, the highest leverage work involves defining problems worth solving and helping users understand how to leverage AI's capabilities in their specific context.