The podcast & newsletter where 170,000+ AI Engineers gather to talk models, tools and ideas. Breaking news today you will use at work tomorrow! Full show notes and newsletter at https://latent.space
Moonlake founders Fan-yun Sun and Chris Manning argue that true world models require action-conditioned symbolic reasoning about physics and consequences, not just pixel prediction, enabling spatial intelligence with orders of magnitude less data than pure scaling approaches.
Sam D'Amico, former Meta and Apple hardware engineer, demonstrates the Impulse Cooktop, a high-performance induction stove featuring a built-in 3kWh lithium iron phosphate battery that delivers 10,000 watts per burner and boils water in 40 seconds, while functioning as distributed grid storage.
Mistral releases Voxtral TTS, a 3B parameter open-weights speech generation model using a novel auto-regressive flow matching architecture that delivers state-of-the-art performance at a fraction of competitors' costs while enabling enterprises to leverage proprietary domain data.
MIT professor Heather Kulik explains how AI discovered quantum phenomena to create 4x tougher polymers and why materials science lacks an 'AlphaFold' equivalent due to missing experimental datasets, emphasizing that domain expertise remains essential to validate AI predictions in chemistry.
David Singleton introduces Dreamer as an 'Agent OS' that combines a personal AI Sidekick with a marketplace of tools and agents, enabling both non-technical users and engineers to build, customize, and deploy AI applications through natural language while maintaining privacy through centralized, OS-level architecture.
Anthropic's Felix Rieseberg explains why AI agents need their own virtual computers to be effective, arguing that confining Claude to chat interfaces severely limits capability. He details how this philosophy shaped Claude Cowork and why product development is shifting from lengthy planning to rapidly building multiple prototypes simultaneously.
Samuel Colvin from Pydantic introduces Monty, a Rust-based Python interpreter designed specifically for AI agents that achieves sub-microsecond execution latency by running in-process, bridging the gap between rigid tool calling and heavy containerized sandboxes.
NVIDIA engineers discuss securing AI agents through the 'two of three' capability rule, the evolution of Brev from startup to NVIDIA's developer experience layer, and how DGX Spark bridges local and cloud GPU workflows for a broader developer audience.
Cursor launched Cloud Agents that provide AI models with full virtual machine access to autonomously write, test, and demonstrate code through video recordings, shifting from simple code generation to end-to-end software engineering workflows.
Box CEO Aaron Levie argues that enterprises face a fundamental infrastructure shift as AI agents proliferate 10-100x beyond human employees, requiring secure 'boxes'—governed data containers with distinct identities and permissions—to prevent data leaks and manage liability in autonomous workflows.
Pulsia, an AI platform founded by Ben, reached $1M ARR in one month by acting as an autonomous 'AI CEO' that builds and operates companies overnight—handling coding, marketing, sales, and strategy while sending users daily morning briefings. The platform currently manages over 1,000 concurrent businesses, demonstrating a shift from AI as a tool to AI as a full-stack operator.
METR (Model Evaluation and Threat Research) tracks AI capabilities using a 'time horizon' framework, revealing remarkably consistent exponential growth in task difficulty that models can solve, while noting recent jumps like Claude Opus 4.5 and the critical distinction between human-equivalent task difficulty and actual autonomous runtime.