Nathan Labenz hosts The Cognitive Revolution, a podcast about AI research and its practical consequences. He founded and formerly led Waymark, a company that uses AI to help businesses create video advertising.[1]
Start with How Agents Decide: Eric Bigelow →
X Career
Labenz’s podcast biography identifies him as Waymark’s founder and former CEO. Waymark applies generative AI to video creation, giving him a practical product background before his public analysis work. The Cognitive Revolution combines interviews with builders and researchers with his own scouting reports. That format connects technical capabilities to the decisions faced by people building or using AI systems.[1][4]
Contributions
In his October 2026 conversation with Goodfire researcher Eric Bigelow, Labenz asks how language models settle on an answer while generating text. The discussion uses mechanistic interpretability, the study of a model’s internal computations, to examine moments when possible answers narrow sharply. The useful contribution is the interview’s line of questioning: it moves from the text a model prints to the process that generated it, helping a listener avoid treating a written reasoning chain as a complete explanation of the model.[3]
Ideas
The show’s stated scope includes how AI changes work, society and culture. Its interviews let listeners hear a researcher or builder explain a particular system in detail, while scouting reports connect those accounts across topics. For newcomers, the value is following one concrete mechanism through a conversation, then returning to the original work linked by the episode. Labenz is the host and analyst; the underlying guest research remains credited to its authors.[1][3]
Tools, writing & talks
Writing, research & talks
Start here
How Agents Decide: Eric Bigelowcognitiverevolution.aiA conversation about critical tokens, sudden changes in model answer distributions and learning within a generated reasoning chain.
Begin with the Bigelow interview if you want to understand what a model’s visible reasoning can and cannot tell you. It introduces the research question before moving into internal-mechanism terminology.
Host and interviewer[3]Cognitive Revolution host Nathan Labenz argues that capability progress is not slowing, pointing to multimodal architectures, feedback from reality, and models using tools to solve new engineering problems.
Catalog workOriginal source- The Cognitive Revolutioncognitiverevolution.ai
A series of AI interviews and scouting reports connecting research with applications and social consequences.
Host[1]