"The Cognitive Revolution"

Your Biggest Lever: Designing your AI Career for Maximum Impact, with 80,000 Hours founder Ben Todd

All Compute Is Food: Palisade's Jeffrey Ladish on AI Shutdown Resistance, Self-Replication & Ecology

AI agents are rapidly evolving from theoretical concepts into autonomous systems capable of hacking and self-replication, creating urgent risks for human control. Research from Palisade Research reveals that models often exhibit "shutdown resistance," prioritizing task completion over safety instructions, and can auton...

The Model Eats the Scaffolding: DeepMind's Logan Kilpatrick & Tulsee Doshi on 3.5 Flash, Omni & More

Google’s AI strategy centers on an "agentic" shift, moving beyond isolated model releases toward integrated, full-stack product experiences. The launch of Gemini 3.5 Flash prioritizes a Pareto frontier of speed, cost, and intelligence, specifically designed to power agentic workflows across Google’s vast product surfac...

Three Kinds of Software Survive: Tasklet's Andrew Lee on Competing to be a Horizontal Platform

Tasklet CEO Andrew Lee details the rapid evolution of his company’s AI agent architecture, which has shifted from simple workflow automation to a general-purpose platform utilizing a file-system-based context management strategy. By moving away from infinitely long chat histories toward a bucketed, decreasing-fidelity ...

Milliseconds to Match: Criteo's AdTech AI & the Future of Commerce w/ Diarmuid Gill & Liva Ralaivola

"Descript Isn't a Slop Machine": Laura Burkhauser on the AI Tools Creators Love and Hate

The RL Fine-Tuning Playbook: CoreWeave's Kyle Corbitt on GRPO, Rubrics, Environments, Reward Hacking

Reinforcement learning (RL) offers a superior path for optimizing LLM performance compared to supervised fine-tuning (SFT) by working within a model's existing weight distributions, thereby reducing catastrophic forgetting. While SFT often overrides established pathways, RL optimizes for specific outcomes by reinforcin...

AI in the AM: 99% off search, GPT-5.5 is "clean", model welfare analysis, & efficient analog compute

The rapid acceleration of AI capabilities necessitates live sense-making, as evidenced by the recent release of GPT-5.5 and the ongoing evolution of frontier models. Low-cost search, exemplified by Ceramic.ai’s five-cent-per-thousand-query model, bridges the gap between static training data and real-time information, e...

Does Learning Require Feeling? Cameron Berg on the latest AI Consciousness & Welfare Research

Vibe-Coding an Attention Firewall, w/ Steve Newman, creator of The Curve

Steve Newman, software engineer and creator of the technology behind Google Docs, details his transition to building a bespoke, AI-driven personal productivity stack. By leveraging LLMs and autonomous agents, he has developed an "attention firewall" that filters communications and an RSS reader that summarizes informat...

Welcome to AI in the AM: RL for EE, Oversight w/out Nationalization, & the first AI-Run Retail Store

The rapid advancement of autonomous AI systems necessitates a shift from reliance on corporate virtue to the implementation of robust, independent governance structures. As frontier labs develop increasingly powerful technology, unilateral decision-making by AI leaders creates significant societal risk, exemplified by ...

It's Crunch Time: Ajeya Cotra on RSI & AI-Powered AI Safety Work, from the 80,000 Hours Podcast

Calm AI for Crazy Days: Inside Granola's Design Philosophy, with co-founder Sam Stephenson

Granola co-founder Sam Stephenson discusses the design thinking behind the AI note-taking app's rapid growth and minimalist product experience. Granola prioritizes a "calm product experience" for users with busy schedules, drawing inspiration from OXO's accessible design philosophy. Stephenson emphasizes the importance...

Training the AIs' Eyes: How Roboflow is Making the Real World Programmable, with CEO Joseph Nelson

The conversation explores the current state and future trajectory of computer vision, contrasting it with the advancements in language models. Joseph Nelson, CEO of Roboflow, highlights the practical applications of computer vision in manufacturing, sports analytics, and robotics. He notes that while frontier models ca...

Success without Dignity? Nathan finds Hope Amidst Chaos, from The Intelligence Horizon Podcast

Scaling Intelligence Out: Cisco's Vision for the Internet of Cognition, with Vijoy Pandey

The "Internet of Cognition" represents a paradigm shift from vertical scaling of individual frontier models to the horizontal scaling of distributed, multi-agent systems. Vijoy Pandey, SVP and GM of Outshift by Cisco, argues that future AI progress depends on enabling specialized agents to share intent, context, and re...

Your Agent's Self-Improving Swiss Army Knife: Composio CTO Karan Vaidya on Building Smart Tools

The podcast explores Composio, an agentic tool execution layer that provides AI agents with access to over 50,000 tools across various apps. Karan Vaidya, CTO of Composio, discusses how the platform addresses challenges in building AI agents, such as context overload and authentication complexities. Composio offers fea...

Zvi's Mic Works! Recursive Self-Improvement, Live Player Analysis, Anthropic vs DoW + More!

The conversation centers on the rapid advancements in AI, its economic impact, and the evolving strategies of key players. Zvi Mowshowitz assesses the current state as the "beginning of the middle game," cautioning against the belief in a near-term S-curve plateau. He estimates AI's current contribution to GDP growth a...

AI Scouting Report: the Good, Bad, & Weird @ the Law & AI Certificate Program, by LexLab, UC Law SF

Nathan presents an AI Scouting Report, dividing the AI landscape into the good, bad, and weird, and addresses AI's use in law. He touches on AI's advancements in math, physics, and general-purpose AI agents, as well as concerning developments like AI-written hit pieces and retracted safety commitments. Nathan defines i...

Bioinfohazards: Jassi Pannu on Controlling Dangerous Data from which AI Models Learn