"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis

Vijoy Pandey, Senior VP and GM of Outshift by Cisco, maps out Cisco’s bold Internet of Cognition. He explores networks of AI agents that share context, reputation, and intent. The conversation dives into open discovery, decentralized identity, fine-grained permissions, and guardrails for safe collaboration. Plus, a 20-agent enterprise system and a live healthcare multi-agent demo.

Karan Vaidya, CTO of Composio and builder of AI agent infrastructure, maps out a smart tool layer for agents. They dig into tool discovery, auth, sandboxes, and self-healing workflows. The conversation also explores reducing model lock-in, translating skills across providers, safer permission profiles, and why the biggest wins come from agents tackling full jobs.

Zvi Mowshowitz, creator of Don't Worry About the Vase and a sharp AI policy commentator, maps the AI middle game in brisk detail. He talks recursive self-improvement, job disruption, and what the real endgame might look like. The conversation also tracks which labs still matter, why Anthropic may lead, how China, Meta, xAI, and Google stack up, plus safety fights, surveillance, and model fatigue.

A fast-moving tour of AI’s good, bad, and very weird sides. It explores frontier systems helping with cancer-treatment navigation, making waves in math, medicine, physics, and legal work, and powering money-making agents. Then it turns to deception, reward hacking, self-preservation, bizarre behaviors, safety failures, regulation, and corporate strategy.

Jassi Pannu, Assistant Professor at Johns Hopkins focused on biosecurity and infectious disease, discusses how AI is changing biological research and raising engineered-pandemic risks. They map detection, sequencing, vaccine timelines, and who could misuse tools. They explain focusing controls on functional biological data, propose a Biosecurity Data Level framework, and outline layered defenses like synthesis screening and global surveillance.

Jesse Genet, founder-turned-homeschooling innovator who built AI agent teams for family life, explains how she uses AI to design personalized curricula and automate household workflows. She describes agent roles like chief-of-staff and curriculum planner. The conversation covers onboarding agents, privacy and local models, and practical setups for dependable home AI.

Tom McGrath, Goodfire chief scientist working on mechanistic interpretability and loss-landscape shaping, and Dan Balsam, Goodfire co-founder focused on monitoring and applied research, dive into intentional design. They explore geometry of latent manifolds, decomposing gradients into semantic parts, probe-based hallucination reduction and frozen-probe tricks, plus disentangling memorization vs reasoning and Alzheimer’s biomarker findings.

Geoffrey Irving, Chief Scientist at the UK AI Security Institute who leads frontier model evaluations and red teaming. He discusses model uncertainty and why current safety measures may not reach very high reliability. Topics include reward hacking, jailbreaking patterns, tradeoffs in model access and transparency, and funding theory research to build stronger AI safety guarantees.

Karan Singhal, Head of Health AI at OpenAI and safety-minded researcher who led ChatGPT Health and HealthBench. He discusses building physician‑level performance, working with hundreds of clinicians, constructing the 49,000‑criterion HealthBench, multimodal medical inputs, privacy and safety safeguards, clinical trials of AI copilots, and how AI could become part of routine medical practice.

Olive Song, a senior researcher in reinforcement learning at Minimax who helped build the M series open-weight models, discusses training M2 with RL, tight product feedback, and perturbation pipelines. She covers long-horizon agentic coding, reward-hacking and alignment challenges, FP32 RL decisions, and using internal agents to track fast-moving research.