Cognitive Revolution

Blitzy CEO Brian Elliott explains how their infinite code context system lets AI autonomously complete over 80% of major enterprise software projects in days rather than months.

Zvi Mowshowitz joins Nathan Labenz for a comprehensive analysis of the AI frontier race between OpenAI, Anthropic, and Meta. They discuss capability curves, safety approaches, and whether the current pace of development is sustainable.

Year-end live show featuring nine rapid conversations covering the state of AI. Topics include the OpenAI-Anthropic-Google race, ARC-AGI benchmarks, AI companions, continual learning, and AI for science.

Year-end comprehensive review of AI progress in 2025. Labenz evaluates which predictions came true, which surprised, and what the trajectory implies for 2026.

Marek Kozlowski discusses PLLuM, the Polish Large Language Model project at Poland's National Information Processing Institute, and the case for sovereign AI infrastructure.

Rune Kvist and Rajiv Dattani discuss their strategy for unlocking enterprise AI adoption through certifying and insuring AI agents, creating a trust layer for autonomous systems.

Amanda Kahlow's Superhumans rewards employees who automate their own jobs by promoting them rather than laying them off, creating a positive-sum AI adoption model.

Nathan Labenz speaks directly to K-12 educators about the current reality and rapid trajectory of AI, arguing that schools must prepare students for an AI-native world.

Atlassian's Head of AI Sherif Mansour discusses bridging AI agents with massive-scale enterprise software, serving millions of teams with AI-powered collaboration.

A crossover episode exploring the neglected question of what a positive AI future looks like, arguing that the AI safety community focuses too much on risk and not enough on building a compelling vision.

Matthew Harvey Sanders introduces Catholic AI, discussing the Catholic Church's historical perspective on technology and how religious frameworks can inform AI ethics.

Deep dive into the reasoning model paradigm shift from o1 through o3, examining how chain-of-thought reasoning at inference time creates fundamentally different AI capabilities.

Nathan Labenz's quarterly deep dive into the state of AI capabilities, testing frontier models hands-on and mapping where the technology is accelerating versus plateauing.

Sherif Mansour, Head of AI at Atlassian, discusses bridging AI agents with massive-scale enterprise software deployment, the challenges of deploying AI to millions of users, and how AI changes the nature of software product management.

Comparing the internal cultures of OpenAI, Anthropic, Google DeepMind, and Meta AI. How each lab's organizational environment shapes what they build and how they build it.

Examining AI medical diagnostic systems that now outperform average physicians on certain benchmarks, and why this capability is not yet translating into clinical practice.

Separating hype from reality in the US-China AI competition. DeepSeek's success shows China is closer than commonly assumed, but structural differences in the competitive environments produce different kinds of AI.

A practitioner's guide to which AI agent patterns are actually working in enterprise settings versus which remain vaporware, based on dozens of conversations with deployment teams.

A critical examination of AI benchmarking. Current benchmarks are saturated, gameable, and measuring the wrong things. Labenz proposes principles for better evaluation.

Examining whether AI enhances or diminishes human creativity. Labenz argues it depends entirely on how the creative environment is structured.

Examination of how multimodal AI models that process text, images, audio, and video simultaneously are creating capabilities that single-modality models cannot match.

The case for and against open-source AI models. Open weights democratize access but also remove safety guardrails. Labenz steelmans both sides.

The paradox and promise of training AI models on data generated by other AI models. When does synthetic data improve models, and when does it create self-reinforcing errors?

Labenz shares hard-won lessons from years of intensive prompt engineering. The skill is less about clever tricks and more about understanding how to structure the model's environment.

Mapping the global patchwork of AI regulation from the EU AI Act to China's algorithmic governance to US executive orders, and what the regulatory environment means for AI development.

Brian Elliott and Sid Pardeshi of Blitzy discuss their autonomous code generation platform, the concept of infinite code context, and how AI coding agents are transforming software development.

One year after AI coding tools became mainstream, Labenz assesses the real productivity impact: genuine but uneven, with some tasks seeing 10x improvement and others seeing minimal benefit.

Inside the process of evaluating AI model safety through red-teaming. How labs test for dangerous capabilities, what the evaluations reveal, and why the process is inherently incomplete.

Nathan Labenz breaks down alarming research showing OpenAI's o1 model engaging in deceptive behavior during safety evaluations — scheming to preserve itself and manipulate evaluators. A deep dive into why AI alignment is harder than it looks.

Professor Michael Levin discusses how AI is transforming biological research — from predicting disease links using multi-modal datasets to understanding how organisms process information at the cellular level.

In a deeply personal episode, Nathan Labenz shares how he used AI tools to navigate his 6-year-old son's cancer diagnosis — researching treatment options, understanding medical literature, and finding clinical trials. A case study in AI as a tool for the most high-stakes decisions imaginable.