Training Data
Sequoia Capital
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.
Episodes
Mentioned books
46 snips
Oct 28, 2026 • 44min
European Defense Sovereignty Starts in the Factory: Stark’s Uwe Horstmann and Johannes Schaback
Uwe Horstmann, Stark’s co-founder and CEO, and Johannes Schaback, its engineering chief, explore how Europe can build defense technology at startup speed. They discuss soldier-first design, mass-manufacturable strike drones, venture-style programs, a failed trial turned breakthrough, and guardrails for autonomous weapons. Horstmann also proposes cloud-like contracts that make surge production part of deterrence.
57 snips
Oct 20, 2026 • 45min
Inside the Robot Data Factory: Mecka's Josh Gao on Sensorizing Humans to Train Robots
Josh Gao, founder and CEO of Mecka AI, builds the data infrastructure powering physical intelligence. He discusses egocentric human recordings, tactile gloves, custom hardware, massive data pipelines, and quality control in a global robot data factory. He also explores why home robots remain so difficult, how industrial automation may arrive first, and the long road from behavior cloning to self-learning machines and asteroid-mining robots.
184 snips
Oct 13, 2026 • 1h 8min
David Silver and Lasse Espeholt of Ineffable Intelligence: A Superlearner for the Age of Experience
David Silver, the AlphaGo pioneer, and Lasse Espeholt, architect of scalable reinforcement-learning systems, reveal Ineffable Intelligence’s bold plan for a “superlearner.” They discuss agents that learn from experience rather than human data, crucibles for open-ended discovery, whether language limits intelligence, and why simulation could make superintelligence safer to study. They also share the GPU infrastructure behind the effort and striking timelines for results.
166 snips
Oct 6, 2026 • 1h 4min
Google's AI Infrastructure Chief, Amin Vahdat, on the Physics & Economics of Frontier AI
Amin Vahdat, Google’s chief technologist for AI infrastructure and a computer networking researcher, explores the machinery powering frontier AI. He discusses why goodput matters more than FLOPS, the evolution of Google’s TPU strategy, and close co-design with DeepMind. They also examine agent-driven demand for CPUs and storage, optical switching, gigawatt power constraints, orbital data centers, and the multi-megawatt supercomputer rack of 2036.
229 snips
Sep 15, 2026 • 1h 5min
Box's Aaron Levie: On Reinventing Yourself in the AI Age and Enterprise Diffusion
Aaron Levie, co-founder and CEO of Box, discusses reinventing a public company for the AI era. He explores why enterprise AI needs workflow-specific applications, how agents can transform document-heavy work, and why permissions and accuracy matter. Levie also examines the rapid spread of coding agents, the rise of background automation, falling inference costs, and what it takes to make companies legible to AI.
34 snips
Sep 1, 2026 • 52min
Making Cities Awesome: Peregrine’s Nick Noone & Ben Rudolph
Ben Rudolph, Peregrine co-founder and former humanitarian-tech builder, joins Nick Noone, co-founder and former Palantir national-security leader. They discuss earning trust after 24 rejections, connecting city data without collecting more, and balancing public safety with privacy and sovereignty. Along the way: cold-case AI, massive evidence searches, synagogue threat detection, weather-related rescue patterns, and the future of civic technology.
91 snips
Aug 25, 2026 • 55min
Parallel’s Parag Agrawal: Building a New Web for AI Agents
Parag Agrawal, former Twitter CEO and now founder and CEO of Parallel Web Systems, is rebuilding search for agent-driven web use. He argues agents will query the web 1,000x more than humans and that human click data is the wrong signal. He explains launching an agent-first product, cutting agentic search latency to 200 ms, and a Shapley-based plan to pay content creators for agent value.
229 snips
Aug 18, 2026 • 54min
Rich Sutton and Khurram Javed: Why AI Models Stop Learning, and How to Start It Again
Rich Sutton, reinforcement learning pioneer and cofounder of Oak Lab, and Khurram Javed, his former student and continual-learning researcher, discuss why models stop learning and the need for agents that learn from experience. They cover the Bitter Lesson, limits of synthetic data and simulators, the big world hypothesis, catastrophic forgetting, and ambitions for energy‑efficient, continually learning minds.
214 snips
Aug 4, 2026 • 47min
Chai Discovery's Bitter Lesson: Drug Design Is Another Scaling Problem
Matt McPartlon, co-founder and protein design researcher, and Josh Meier, co-founder and former OpenAI researcher, explain treating drug design as an engineering scaling problem. They discuss scaling data, models, and compute to boost antibody hit rates, diffusion models for protein generation, faster lab-design loops, and a Molecular CAD vision that could shrink discovery timelines from months to days.
98 snips
Jul 29, 2026 • 49min
Building the Automated AGI Lab: Core Automation's Jerry Tworek and Rohan Anil
Rohan Anil, researcher and Core Automation co-founder who built the Shampoo optimizer and led Gemini pre-training. Jerry Tworek, AI researcher and Core Automation co-founder who led reasoning at OpenAI and scaling RL. They debate transformer limits, arguing architectures must enable continual learning and test-time adaptation. They discuss inefficiencies in computation, uniting pretraining with RL, and automating kernel generation to unlock new algorithms.
