What makes great AI researchers

Why AI is going vertical (again) | Dianne Penn (Anthropic)

Topic 6 of 18

Published Jul 26, 2026 · Clip 31:36–35:20 (3:43)

Diane observes that Anthropic's most successful researchers and research leaders are strong first-principles thinkers who are passionate about their area and hold bold descriptions of what's possible. Crucially, they stay close to the details: the chief scientist and heads of fine-tuning and RL actually inspect training runs, evals, and underlying data, which builds taste. She cites Dario's 'we can transform software engineering' ambition and the 'be more ambitious' meme, describing the practical version as being stubborn about the area while staying loose about the exact approach, and ensuring what you build is forward compatible. She grounds this operationally with her Claude 8 thought experiment: ask what changes in what users do and what that means for how you're building today.