Latent Space: The AI Engineer Podcast
Latent.Space
The podcast by and for AI Engineers! In 2025, over 10 million readers and listeners came to Latent Space to hear about news, papers and interviews in Software 3.0.
We cover Foundation Models changing every domain in Code Generation, Multimodality, AI Agents, GPU Infra and more, directly from the founders, builders, and thinkers involved in pushing the cutting edge. Striving to give you both the definitive take on the Current Thing down to the first introduction to the tech you'll be using in the next 3 months! We break news and exclusive interviews from OpenAI, Anthropic, Gemini, Meta (Soumith Chintala), Sierra (Bret Taylor), tiny (George Hotz), Databricks/MosaicML (Jon Frankle), Modular (Chris Lattner), Answer.ai (Jeremy Howard), et al.
Full show notes always on https://latent.space www.latent.space
We cover Foundation Models changing every domain in Code Generation, Multimodality, AI Agents, GPU Infra and more, directly from the founders, builders, and thinkers involved in pushing the cutting edge. Striving to give you both the definitive take on the Current Thing down to the first introduction to the tech you'll be using in the next 3 months! We break news and exclusive interviews from OpenAI, Anthropic, Gemini, Meta (Soumith Chintala), Sierra (Bret Taylor), tiny (George Hotz), Databricks/MosaicML (Jon Frankle), Modular (Chris Lattner), Answer.ai (Jeremy Howard), et al.
Full show notes always on https://latent.space www.latent.space
Episodes
Mentioned books
269 snips
Mar 24, 2026 • 35min
🔬Why There Is No "AlphaFold for Materials" — AI for Materials Discovery with Heather Kulik
Heather Kulik, MIT chemical engineering professor in computational chemistry, explores why materials discovery has no AlphaFold moment yet. She digs into AI-designed polymers that turned out four times tougher. She talks about active learning for CO2 capture, why quantum models need ML speedups, where LLMs still fail simple chemistry tasks, and how noisy data and lab bottlenecks slow progress.
955 snips
Mar 20, 2026 • 1h 4min
Dreamer: the Personal Agent OS — David Singleton
David Singleton, former Stripe CTO and early Android leader, talks about Dreamer and its Sidekick, a personal AI that builds apps and agents from plain language. He explores agent app stores, forkable tools, temporary apps for trips and events, security and permissions, paid builder ecosystems, and why taste still matters in AI software.
1,946 snips
Mar 17, 2026 • 1h 27min
Why Anthropic Thinks AI Should Have Its Own Computer — Felix Rieseberg of Claude Cowork & Claude Code Desktop
Felix Rieseberg, Anthropic engineer behind Claude Cowork and Claude Code Desktop, joins with a background in Slack, Electron, and the Windows 95-in-JavaScript project. They dig into how Cowork emerged from messy knowledge work, why AI may need its own computer, local-first agents, fast prototype culture, reusable skills, browser vision, and how automation could reshape finance, junior careers, and vertical AI.
802 snips
Mar 12, 2026 • 1h 1min
Retrieval After RAG: Hybrid Search, Agents, and Database Design — Simon Hørup Eskildsen of Turbopuffer
Simon Eskildsen, founder and CEO of Turbopuffer and former Shopify infra engineer, digs into why search for unstructured data broke old cost models. He gets into hybrid retrieval for code, agents making parallel tool calls, bold database design around object storage and NVMe, and the wild stories behind Readwise, Notion, and Cursor.
359 snips
Mar 10, 2026 • 1h 24min
NVIDIA's AI Engineers: Agent Inference at Planetary Scale and "Speed of Light" — Nader Khalil (Brev), Kyle Kranen (Dynamo)
Kyle Kranen, an engineering leader behind NVIDIA Dynamo who builds datacenter-scale inference systems. Nader Khalil, a DevRel leader focused on GPU developer UX and Brev’s developer onboarding. They discuss Dynamo’s scale-out inference approach, prefill vs decode disaggregation, Kubernetes-based scaling, SOL (Speed of Light) urgency culture, model‑hardware co-design, long‑context limits, and agent security and tooling.
1,039 snips
Mar 6, 2026 • 1h 7min
Cursor's Third Era: Cloud Agents
They dig into cloud agents that control full VMs, pixels and remote input to reproduce bugs, run tests and record demo videos. They demo slash commands, subagents and parallel agents for context management and auto-routing. They debate UI choices, onboarding, memory and scaling tradeoffs as agents shift workflows from tab autocomplete to high-level delegation.
662 snips
Mar 5, 2026 • 1h 17min
Every Agent Needs a Box — Aaron Levie, Box
Aaron Levie, CEO and co-founder of Box, a leader in enterprise content platforms. He talks about why filesystems and sandboxed 'boxes' are central for safe, governed agents. Short takes cover agent identity and permissions, reworking workflows for agent-first work, challenges in search and context limits, and how teams and evals measure agent performance.
589 snips
Feb 27, 2026 • 56min
METR’s Joel Becker on exponential Time Horizon Evals, Threat Models, and the Limits of AI Productivity
Joel Becker, researcher leading METR (model evaluation and threat research), explains the origin and limits of the Time Horizon chart. He walks through how tasks were chosen and biases in benchmarks. The conversation covers developer productivity studies, milestone-like model jumps, threat modeling for discontinuous takeoff, and trade-offs in building evaluation scaffolds versus real-world deployment.
428 snips
Feb 26, 2026 • 52min
[LIVE] Anthropic Distillation & How Models Cheat (SWE-Bench Dead) | Nathan Lambert & Sebastian Raschka
Swyx, AI writer and commentator, and Sebastian Raschka, ML professor specializing in interpretability, join to dissect distillation, benchmarks, and model memorization. They debate detection limits for API-based distillation, teacher-student dynamics, and logits vs open-weight approaches. Sweebench and its vulnerabilities to leakage and curation issues are also explored.
273 snips
Feb 25, 2026 • 34min
🔬Searching the Space of All Possible Materials — Prof. Max Welling, CuspAI
Max Welling, renowned theoretical physicist and ML researcher building CuspAI to speed materials discovery for climate solutions. He sketches nature as a kind of computer and the idea of a “physics processing unit.” Topics include symmetry and equivariant networks, diffusion models tied to thermodynamics, why materials may limit the energy transition, and how AI and lab experiments can form an automated discovery loop.
