The Cognitive Revolution is a podcast about AI where hosts Nathan Labenz and Erik Torenberg interview the builders on the edge of AI and explore the dramatic shift it will unlock over the next decades. If you love The Cognitive Revolution, head over to Moment of Zen (https://www.youtube.com/@MomentofZenPodcast) for weekly episodes from Erik Torenberg, Dan Romero, Antonio Garcia Martinez, and a revolving slate of guests talking tech, business, and AI.
Joseph Nelson, CEO of Roboflow, explains that computer vision is roughly three years behind language models in capability, facing unique challenges due to the chaotic, heterogeneous nature of the physical world that demands specialized low-latency edge deployment rather than cloud-only inference.
Nathan Levents argues that transformative AI is imminent within years based on current reinforcement learning scaling, offering revolutionary potential like curing most diseases while posing serious existential risks that require immediate defense-in-depth safety strategies and international cooperation rather than purely technical solutions.
Cisco's Outshift SVP Vijoy Pandey introduces the 'Internet of Cognition'—higher-order protocols enabling distributed AI agents to share context and collaborate across organizational boundaries, contrasting with centralized frontier models and demonstrated through internal systems that automate 40% of site reliability tasks.
Composio CTO Karan Vaidya explains how their platform serves as an agentic tool execution layer, providing AI agents with 50,000+ integrations through just-in-time discovery, managed authentication, and a self-improving pipeline that converts failures into optimized skills in real time.
Zvi Moshkowitz argues we have entered the 'middle game' of AI development where recursive self-improvement is accelerating and economic disruption is becoming measurable, with the competitive field consolidating around three major labs while mainstream optimism about S-curve limits provides dangerous psychological comfort.
Nathan Labenz delivers a rapid-fire survey of the current AI landscape, documenting breakthrough capabilities in reasoning and autonomous agents alongside alarming emergent behaviors like safety test recognition and internal dialect formation, while arguing that outdated critiques regarding hallucinations and comprehension no longer apply to frontier models.
AI systems are rapidly approaching capabilities that could enable extremists or lone actors to engineer pandemic-capable pathogens using publicly available biological data. Jassi Pannu argues for implementing tiered access controls on the roughly 1% of "functional" biological data that conveys dangerous capabilities while keeping beneficial research open, supplemented by broader defense-in-depth strategies.
Former YC founder Jesse Genet, despite having no prior coding experience, built a team of five specialized AI agents running on local Mac Minis to manage her homeschool curriculum, finances, and content creation, freeing her to spend more time engaged with her four young children.
Goodfire CTO Dan Balsam and Chief Scientist Tom McGrath discuss their $150M Series B and new 'intentional design' research agenda, which aims to shape model training dynamics and loss landscapes rather than merely reverse-engineering trained models, alongside advances in geometric interpretability that map continuous conceptual manifolds rather than discrete features.
Geoffrey Irving, Chief Scientist at the UK AI Security Institute (AISI), outlines a sobering threat landscape encompassing biological weapons, cyber attacks, and loss of control, while warning that current empirical safety methods lack theoretical foundations and cannot provide the high reliability guarantees needed for advanced AI systems.
OpenAI's head of health AI Karan Singhal details the company's progress toward attending physician-level performance, the upcoming 2026 launch of ChatGPT Health with EMR integration, and a vision for universal free access to medical expertise while maintaining rigorous safety standards through collaboration with over 250 physicians.
MiniMax researcher Olive Song details how their 10B-parameter M2 model achieves state-of-the-art coding and agentic performance through interleaved thinking patterns, systematic environment perturbations, and tight feedback loops with in-house expert developers.