Jeff Dean: The 1% Rule for Building in AI

Y Combinator

Summary

Jeff Dean, Google's Chief Scientist, explores the evolution of AI systems, emphasizing the shift toward agent-based architectures, specialized hardware, and context engineering. He reflects on his career, from building MapReduce and the TPU to current work on Gemini and AI-driven scientific discovery, while offering practical advice for founders on how to apply napkin math, cultivate professional taste, and tackle complex problems through iterative, self-improving loops.

Chapters

  1. 1:44Predicting the Future of AI
  2. 1:54Automated Machine Learning
  3. 3:21Lessons from System Engineering
  4. 4:03Specialized Inference Hardware
  5. 5:12Agent Capability Growth
  6. 7:37The TPU Origin Story
  7. 12:21The Cost of Data Movement
  8. 14:08Efficiency through Batching
  9. 16:37Context Engineering and Tool Use
  10. 16:42Beyond Model Size
  11. 19:16Improving Model Performance
  12. 21:43Internalizing Performance Hints
  13. 22:20Building Reliable Agent Workflows
  14. 23:09Handling Out-of-Distribution Errors
  15. 23:40Multi-Agent Orchestration
  16. 24:15Inference Time Compute
  17. 25:49Founder Strategy: Finding Your Niche
  18. 28:45Identifying Opportunity
  19. 29:08The Power of Proprietary Data
  20. 39:28Challenging First Principles
  21. 42:39The Future of Scientific Discovery
  22. 45:22Accelerating Validation
  23. 46:30Recursive AI Improvement
  24. 52:53Designing for Impact

Topics covered

  • Agent-Based Systems
  • Napkin Math

+4 more topics (full content when you follow the channel)

Transcript excerpt

All right. Should we go Should we get started, Jeff? >> Sure. Sounds great. >> All right. Jeff, welcome. And again, thank you so much for being here. Especially I just got a cold and thank you for being here. >> Yeah, I'm afraid I've lost my voice. I don't normally sound quite like this, but we'll we'll do what we can. >> So, um, you built map reduce, big table, tensorflow, the TPU, Gemini. We could spend a whole hour on all the things you've done, but what I love is that you're still making bold predictions in public. Last year, yes, last year in May 2025 at AI Ascent, you said that AI is at the level of a junior engineer. That was about a year ago. It's been How close are we to that prediction? Yeah, I mean I feel like uh the models have been getting a lot better at sort of agent-based longer running coding tasks and it seems pretty clear that they are now actually pretty capable and d…