Why PMs are suddenly valuable again | Tulsee Doshi (Head of Product, Gemini)
Nov 8, 2026 · 58m
More Technology podcasts · More Business podcasts
Summary
Tulsi, Head of Product for Google’s Frontier Models, discusses the unpredictable nature of AI development, comparing it to Snape’s potions class where outcomes involve "magic." She outlines Google’s full-stack strategy to compete with OpenAI and Anthropic, emphasizing safety through monitoring and the importance of product managers who understand model "vibes." Tulsi also shares practical advice on writing evals, the 60-20-20 rule for building ahead of model capabilities, and the resurgence of PM value in the AI era.
Topics discussed
Introduction: The unpredictable nature of AI models
Overview of Google's Frontier Models and Gemini
The PM role in model development and 'magic'
Google's strategy in the AI race and full-stack advantage
AI safety, responsible deployment, and cyber defense
Optimism on existential AI risks and societal debate
Discoverability challenges and UX experimentation
Model personality as UX and PM-research collaboration
The resurgence of PMs and the value of taste
Hiring PMs: Agility and understanding model vibes
Practical tips for evaluating AI model improvements
Case study: Improving skin tone evaluation scales
Stanford research on crowdsourcing and agent swarms
The 60-20-20 rule: How far ahead to build
Defining the AI harness: Wrapper and context
Thriving in the new AI world: Skills and confidence
Google's upcoming AI assistant and product strategy
AI, productivity, and the risk of burnout
Managing AI slop and the evolving PM job
Rapid fire: Books, habits, dance, and listener tips
More from Lenny's Podcast: Product | Career | Growth
Listen ad-free on Castria