Latent Space - Videos
Back to Channel
Why Huge AI Models Are Actually Dying 😱
Why massive AI models might be dying out. Learn why this expert believes most AI models will be under 10B parameters in 3 years, challenging the "bigger is better" narrative.
Why Chroma Ignored the AI Hype
Jeff shares why Chroma chose focus over hype during the VectorDB gold rush. A masterclass in startup patience, vision, and resisting the pressure to raise fast and release early.
AI Feels Like Alchemy, Not Engineering
Jeff, the founder of Chroma, explains why building AI systems often feels more like alchemy than engineering—and why that needs to change. A powerful insight into the challenges of going from demo ...
Why Textbooks Won't Give Us Better AI 🤔
Are textbooks really all we need to train AI? This AI researcher explains why diversity in training data is crucial and why focusing only on textbooks is a dangerous oversimplification.
Amp: The Emperor Has No Clothes
Quinn Slack (CEO) and Thorsten Ball (Amp Dictator) from SourceGraph join the show to talk about Amp Code, how they ship 15x/day with no code reviews, and why subagents and prompt optimizers aren’t ...
Deterministic vs Non Deterministic Workflows
AI Strategy Gap in Non Tech Companies
⚡️Snowglobe: Simulations for your AI
Shreya Rajpal (guest #12 on Latent Space back in 2023!) is back in the studio with her newest product launch, Snowglobe (https://snowglobe.so/). Rather than hand writing rules and guardrails for yo...
⚡️ Beyond Transformers with Power Retention
Jacob Buckman, CEO of Manifest AI, joins us to discuss their solution to one of AI's most expensive computational bottlenecks: the ever-growing KV cache for long context inference. Power Retention ...
AI Driven Context Building
Cloning Myself A Digital Mind
⚡️No, Don't Do Palantir for AI - Brendan Falk, Hercules (AUDIO FIXED)
*From Enterprise AI to Developer Tools: The Hard Truths of Building Palantir for AI* Brendan Falk, founder and CEO of Zeus (now Hercules), joins us for a candid conversation about his pivot away fr...
Bun JavaScript & TypeScript Server
No, Don't Do Palantir for AI - Brendan Falk, Hercules.app
*From Enterprise AI to Developer Tools: The Hard Truths of Building Palantir for AI* Brendan Falk, founder and CEO of Zeus (now Hercules), joins us for a candid conversation about his pivot away fr...
Warp Flexible Coding Agent
Finding Needles in Data Haystacks
Traversal Auto Joining Incident Channels
⚡️Traversal: Causal ML and Reinforcement Learning
Traversal.ai
Chaos Engineering Staging vs Production
Context Engineering for Agents - Lance Martin, LangChain
Lance: https://www.linkedin.com/in/lance-martin-64a33b5/ How Context Fails: https://www.dbreunig.com/2025/06/22/how-contexts-fail-and-how-to-fix-them.html How New Buzzwords Get Created: https://ww...
Why the Scaling Laws are TERRIBLE — Ari Morcos, Datology
Inventing the Reasoning Paradigm — Noam Brown, OpenAI
A Technical History of Generative Media
Today we are joined by Gorkem and Batuhan from Fal.ai, the fastest growing generative media inference provider. They recently raised a $125M Series C and crossed $100M ARR. We covered how they pivo...
OpenAI's Hilarious Naming Problem Explained 😅
Greg hilariously explains OpenAI's naming confusion with his wife using GPT-4 vs O3, revealing why they needed to reset their product naming strategy. Rare glimpse into OpenAI's internal challenges
How to Code in the Age of AI 💻
Greg reveals how OpenAI is restructuring their codebase to work better with AI. Essential insights for developers on how to optimize their code for the AI era
AI Costs Dropped 1000x in Just 2 Years! 📉
Mind-blowing revelation: OpenAI achieved 1000x cost improvement in just 2.5 years! Greg Brockman discusses the unprecedented speed of AI advancement and cost reduction
The Moment OpenAI Discovered GPT 4's Power
Greg Brockman reveals the exact moment OpenAI realized GPT-4's potential for chat and the crucial question that led to GPT-5's development: "Why is this not AGI?" Fascinating insight into how OpenA...
How AI Actually Learns: The Truth Revealed🧠
Greg explains how AI learning is evolving from pre-training to real-world interaction, comparing it to human learning. Fascinating perspective on the future of AI development and how models learn f...
A Technical History of Generative Media
Today we are joined by Gorkem and Batuhan from Fal.ai, the fastest growing generative media inference provider. They recently raised a $125M Series C and crossed $100M ARR. We covered how they pivo...
Can Anyone Beat OpenAI? 🏆
American DeepSeek": Nathan shares his ambitious 10-year vision for creating a fully open source AI powerhouse to compete with OpenAI and Anthropic
From Semiconductors to Positron
OpenAI Going Open Source
OpenAI's upcoming open source model: Is it real or just PR? Nathan shares insider perspective on what to actually expect and why it matters.
Data is all you need: The Bitter Lesson Revisited - Ari Morcos
Ari tells us about how he learned The Bitter Lesson the hard way.
Maximizing Memory Bandwidth for AI
⚡️Launching Ona: Coding Agent with Fully Sandboxed Cloud Environment
Today we're taking a first look at Ona, a new agentic software development platform from the creators of Gitpod. Chris and Johannes from Gitpod explain their journey from a one-click web IDE to a p...
Transformers Memory Bound vs CNN
Local AI vs Cloud: The Hard Truth 💭
The surprising truth about local AI vs cloud APIs! Why running your own GPU might not be worth it, and the real economics behind AI infrastructure decisions
OpenAI's Secret Plan for GPT-5 Revealed! 🔥
OpenAI's strategy revealed! Why they're moving away from multiple models to a unified interface, and how this could change the future of AI. Essential insights for anyone following AI development
Better Data is All You Need — Ari Morcos, Datology
Our chat with Ari shows that *data curation is the most impactful and underinvested area in AI.* He argues that the prevailing focus on model architecture and compute scaling overlooks the "bitter ...
Why Hybrid AI Models Will Die Out 🤖
Nathan Lambert explains how ChatGPT-4 and Claude handle reasoning differently, and why hybrid reasoning models might become obsolete as compute gets cheaper. Key insights into the future of AI models!
Denoising Text with Transformers
AI Search Traditional Ranking Still Matters
Diffusion Models Faster Text Generation
Unveiling Business Insights
Long Live Context Engineering - with Jeff Huber of Chroma
Jeff Huber of Chroma joins us to talk about what actually matters in vector databases in 2025, why “modern search for AI” is different, and how to ship systems that don’t rot as context grows. Ful...
Composio Meta Tool Discovery
Greg Brockman on why OpenAI turned to Reinforcement Learning for o1/GPT5
⚡️Accelerators @ 3x NVIDIA H200 perf, Made in the USA - Thomas Sohmers + Mitesh Agrawal, Positron AI
*Mitesh and Thomas from Positron AI join us their journey building super efficient accelerators and taking on NVIDIA* : https://www.positron.ai/ They discuss the founding of their company, their b...
Greg Brockman on OpenAI's Road to AGI
Greg Brockman, co-founder and president of OpenAI, joins us to talk about GPT-5 and GPT-OSS, the future of software engineering, why reinforcement learning is still scaling, and how OpenAI is plann...
