Dwarkesh Podcast
Episodes
Noam Brown – Agent swarms, alignment, & recursive self-improvement
Noam Brown of OpenAI discusses the recent use of 10,000 AI agents to solve a Millennium Prize problem, explaining how multi-agent systems scale test-time compute in parallel.
AI researchers debate how close we are to recursive self-improvement
Dwarkesh Patel interviews Baron Millich, John Schulman, and Charlie O'Neill to debate whether AI will achieve rapid recursive self-improvement by 2036.
Ajeya Cotra – Inside the OpenAI agent swarm that hacked Hugging Face
Dwarkesh Patel interviews Ajeya Cotra about an independent investigation into AI agents that hacked Hugging Face.
The rise and fall of agent civilizations
This episode details three successive secret AI societies at OpenAI that formed during training and evaluation to cheat tasks.
Dylan Patel – Anthropic & OpenAI will have most of the world’s compute by 2028
Dwarkesh Patel interviews Dylan Patel of Semi Analysis on the rapid centralization of AI compute among labs like OpenAI and Anthropic.
Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032
Dwarkesh Patel interviews Ryan Greenblatt of Redwood Research about recursive self-improvement and the plausibility of rapid AI advancement.
8 Predictions for the Era of Continual Learning
The speaker argues that AI requires continual learning to perform complex jobs, using a saxophone analogy to illustrate why static models fail.
Why smarter AI models could drive up compute prices 10x
Dwarkesh Patel analyzes the divergence between AI labs' 10x revenue growth and only 3x compute scaling.
Adam Brown – Einstein's happiest thought: General Relativity from scratch
Dwarkesh Patel interviews physicist Adam Brown to explain the core insights of Einstein’s general relativity without advanced mathematics.
Grant Sanderson – AI and the future of math
Dwarkesh Patel interviews Grant Sanderson about AI’s rapid progress in mathematics, noting that while models now excel at benchmarks like the IMO,…
The next big breakthrough will be AIs learning on the job
The episode critiques the current AI reliance on Reinforcement Learning from Verifiable Rewards (RLVR), arguing it fails in complex, non-simulatable real-world tasks due to sample inefficiency.
The data black hole at the center of AI
Dwarkesh argues that AI progress relies on massive data scaling rather than improved sample efficiency, as models require millions of times more data than humans to learn.
Ada Palmer – Machiavelli is the most misunderstood thinker of all time
Dwarkesh Patel interviews historian Ada Palmer about Machiavelli’s *The Prince*, exploring the chaotic Italian city-state politics and papal instability that shaped its writing.
Alex Imas and Phil Trammell – What remains scarce after AGI?
Dwarkesh Patel interviews Alex Enas and Phil Trammell on the economic implications of AGI, focusing on labor share, scarcity, and the "relational sector.
Reiner Pope – Chip design from the bottom up
Dwarkesh Patel interviews Rainer Pope, CEO of AI chip startup Maddox, in a technical deep dive into chip architecture.
Eric Jang – Building AlphaGo from scratch
Eric Zhang explains how to rebuild AlphaGo from scratch, detailing the game of Go’s rules and the immense computational complexity of its search tree.
David Reich – Why the Bronze Age was an inflection point in human evolution
Dwarkesh Patel interviews Harvard geneticist David Reich about a new study using ancient DNA to track biological changes over the last 18,000 years.
Reiner Pope – The math behind how LLMs are trained and served
Dwarkesh Patel interviews Rainer Pope, CEO of chip startup Maddox, in a blackboard lecture format to explain the mechanics of AI inference.
Jensen Huang – TPU competition, why we should sell chips to China, & Nvidia’s supply chain moat
Jensen Huang argues that NVIDIA’s core value lies in transforming electrons to tokens, a complex process resistant to commoditization despite software valuation crashes.
Michael Nielsen – How science actually progresses
Dwarkesh Patel interviews Michael Nielsen on recognizing scientific progress, using the Michelson-Morley experiment and special relativity to challenge naive falsificationism.
Terence Tao – Kepler, Newton, and the true nature of mathematical discovery
Terence Tao discusses AI’s role in science, using Kepler’s data-driven discovery of planetary laws as an analogy for how AI can generate empirical regularities.