scripod.com

Jeff Dean: The 1% Rule for Building in AI

Shownote

In 2001, Jeff Dean and Sanjay Ghemawat did the math and realized Google’s entire search index would fit in RAM — then shipped it in a few days, and search got fast. In 2013, another napkin calculation showed that three minutes of daily speech recognition p...

Highlights

In this insightful conversation, Google's Chief Scientist Jeff Dean sits down with YC's Diana Hu at Startup School 2026 to explore the evolution of AI, from the foundational decisions that made Google search fast to the cutting-edge hardware and strategies shaping the future of machine learning. Dean shares the thought experiments behind the TPU, introduces his '1% Rule' for founders, and discusses the importance of context engineering, persistence, and developing a keen sense of taste in a rapidly changing field.
07:23
Identify bottlenecks and solve problems from first principles.
14:34
Building systems around models is accessible to everyone.
31:19
The scarce skill is having good taste in choosing problems.
38:34
Questioning assumptions can lead to entirely new design methodologies.
49:12
Persistence matters despite rejections.

Chapters

From Napkin Math to the TPU: How a 2013 Calculation Shaped AI's Hardware Future
00:00
The New Frontier: Why Context Engineering is the Key Skill for Building with AI
14:34
The 1% Rule: A Founder's Guide to Choosing Problems That Matter in AI
21:25
Cultivating AI Taste and Automating the Scientific Method with Faster Loops
35:08
Persistence, Meaningful Work, and the Next Big Challenges in AI Research
45:19

Transcript

Diana Hu: All right, should we get started, Jeff? Jeff Dean: Sure, sounds great. Diana Hu: 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. Jeff Dean: Yeah, I'm afraid I'v...