Anthropic’s first technical PM on token maxing, the jagged edge, and living in the future | Dianne Penn

2026-07-26 · Source: Lenny's Podcast: Product | Career | Growth · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Project & Product Management, Emerging Technologies & Innovation · Depth: Intermediate, extended

Summary

Dianne Penn, Anthropic's first technical product manager and current Head of Product for AI Research and Labs, details the company's rapid evolution since joining in 2023. She discusses shipping models from Claude 2 through Fable, incubating initiatives like Claude Code, MCP, and Skills, and pioneering an "eval-driven development" loop. Penn highlights key inflection points, including Opus 3 and Opus 4.5, emphasizing the crucial synergy between frontier models and product experiences. The discussion also covers the concept of "token maxing," the unique, bottoms-up culture of Anthropic Labs, and the critical role of human judgment and adaptability in navigating AI's exponential growth. She stresses the importance of hands-on leadership and continuous experimentation to keep pace with technological advancements.

Key takeaway

For AI Product Managers navigating rapid model advancements, prioritize "eval-driven development" by translating nuanced user feedback into measurable evaluation sets. This approach, coupled with hands-on engagement with frontier models, ensures product experiences keep pace with AI capabilities. Cultivate adaptability and a "first principles" mindset to identify discontinuous opportunities, and foster team collaboration to avoid burnout in this fast-paced environment.

Key insights

Adaptability and eval-driven development are crucial for navigating AI's rapid, discontinuous capability jumps.

Principles

• Adaptability is paramount in AI development.
• "Evals are the new PRDs" for user value.
• Hands-on leadership is essential for PMs.

Method

Identify user pain points, reproduce failures, standardize into eval sets, and use these to measure model improvements for actionable research and product development.

In practice

• Use AI for personalized coaching and brainstorming.
• Pair with others for AI experimentation.
• Go deep on 1-2 AI applications for impact.

Topics

• Anthropic
• AI Product Management
• Frontier Models
• Eval-Driven Development
• AI Research
• Claude Code
• Token Maxing

Best for: AI Product Manager, Director of AI/ML, AI Scientist

Related on AIssential

• Inside Anthropic Labs — Anthropic's Labs team builds AI products for future model capabilities, accepting initial low utility for later exponential gains.
• AI News: Anthropic Went Crazy This Week! — AI platforms are rapidly integrating multimodal capabilities and expanding into direct computer interaction and content generation.
• Anthropic messed up — Anthropic's past compute underinvestment limits its current market competitiveness despite having a superior model.
• Inside How Anthropic Is Building the Next Claude | Alex Albert — Anthropic develops Claude as a product, integrating user feedback, defining capabilities, and using AI tools to refine model character and autonomy.
• Why the rise of open source AI isn’t hurting Anthropic … yet — Frontier AI models drive discovery, while open-source models increasingly own production in a two-tiered AI economy.
• We just took the DARK timeline — Ideological rigidity and strategic miscalculation led to Anthropic's corporate downfall and market marginalization.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Lenny's Podcast: Product | Career | Growth.