Join us as we train our neural nets on the theme of the century: AI. Sonya Huang, Pat Grady and more Sequoia Capital partners host conversations with leading AI builders and researchers to ask critical questions and develop a deeper understanding of the evolving technologies—and their implications for technology, business and society. The content of this podcast does not constitute investment advice, an offer to provide investment advisory services, or an offer to sell or solicitation of an offer to buy an interest in any investment fund.
What this show covers
reinforcement learningcontinual learningAI agentscomputeAI alignmentsuperintelligenceAlphaGoAI infrastructureTPUsdata centersnetworkingpower economicsenterprise AIapplication layerAI adoption
What listeners will find here
Can AI learn everything from scratch? David Silver’s new bet says yes
David Silver and Lasse Espeholt lay out a plan to build a “superlearner” that discovers language, math, and science through direct experience rather than pre-training on human data. The episode pairs a bold technical vision with concrete discussion of RL systems, compute, alignment, and timelines.
Why Google’s AI infra chief says FLOPS is the wrong metric for frontier AI
Google’s AI infrastructure strategy is the focus here, from why “goodput” matters more than FLOPS to how TPU design, networking, power, and storage are changing at frontier scale. Amin Vahdat also looks ahead to long-horizon agents, orbital data centers, and the supercomputer of 2036.
Aaron Levie on why the AI application layer may matter more than the model
This episode focuses on how Box is reinventing itself for AI while operating as a public company. Aaron Levie lays out why enterprise value comes from connecting models to real workflows, why model-agnostic platforms matter, and where AI adoption will spread fast or stall.
Why Core Automation thinks AGI needs a post-transformer architecture
Jerry Tworek and Rohan Anil argue that scaling reinforcement learning and bigger transformers won’t be enough for AGI. They make the case that continual learning, test-time adaptation, and end-to-end optimization are the real bottlenecks—and explain why they’re building an automated lab around that bet.
Can drug design be scaled like software? Chai Discovery says yes
Drug design is the focus here, framed not as bespoke biology but as a scaling problem in data, models, and compute. Chai Discovery’s founders argue that bigger systems can dramatically improve antibody hit rates and compress discovery timelines while working with pharma, not against it.
Episodes
More like Training Data
The a16z Show
Andreessen Horowitz
All-In with Chamath, Jason, Sacks & Friedberg
All-In Podcast, LLC
The Other Hand
Jim Power & Chris Johns
Decoder with Nilay Patel
The Verge
Big Technology Podcast
Alex Kantrowitz
Times Tech
The Sunday Times
Limitless: An AI Podcast
Limitless
Sources with Alex Heath
Alex Heath
Les Echos de l'IA
Les Echos
Search Engine
PJ Vogt