Feed
Content view for crawlers and readers. Open the interactive view.
OpenClaw, OpenAI and the future | Peter Steinberger
I'm joining OpenAI to work on bringing agents to everyone. OpenClaw will move to a foundation and stay open and independent.
One ride on a Liquid
Whether you are carving down a slope on a snowboard, surfing your first wave, or cruising down an open road, something special happens when you ride. The weightlessness, the feeling that time stops, the sense that nothing matters except the present moment. You feel the gentle wind, the subtle shift in balance, and everything else simply disappears. Liquid is about making that feeling accessible every day. We focused on the sensation of freedom found in constant motion, and worked on how to bring that essence into a simple skateboard. This is why we created the Liquid Skateboard: a remoteless electric skateboard designed to maintain your momentum. Its cruise control delivers the right power so you don’t slow down once both feet are on the deck. Discover more on: https://liquidskateboard.com/ A huge thank you to Gaston for capturing the feeling of freedom on camera, and a big shout-out to Martin for filming a Liquid while riding a Liquid. #liquidskateboard #electricskateboard #skateboarding #cruiserboard #copenhagen #wheels
The Final Bottleneck
AI speeds up writing code, but accountability and review capacity still impose hard limits.
Why OpenAI Can't Win The Decentralized AI Future (OpenClaw, Apple's Win, X.AI Exodus)
Sam Lessin, Dave Morin, Jessica Lessin, and Brit Morin discuss OpenClaw, decentralized AI, AI burnout, and why Apple may benefit from the shift.
The AI Frontier: from Gemini 3 Deep Think distilling to Flash — Jeff Dean
From rewriting Google’s search stack in the early 2000s to reviving sparse trillion-parameter models and co-designing TPUs with frontier ML research, Jeff Dean has quietly shaped nearly every layer of the modern AI stack. As Chief AI Scientist at Google and a driving force behind Gemini, Jeff has lived through multiple scaling revolutions from CPUs and sharded indices to multimodal models that reason across text, video, and code. Jeff joins us to unpack what it really means to “own the Pareto frontier,” why distillation is the engine behind every Flash model breakthrough, how energy (in picojoules) not FLOPs is becoming the true bottleneck, what it was like leading the charge to unify all of Google’s AI teams, and why the next leap won’t come from bigger context windows alone, but from systems that give the illusion of attending to trillions of tokens. We discuss: • Jeff’s early neural net thesis in 1990: parallel training before it was cool, why he believed scaling would win decades early, and the “bigger model, more data, better results” mantra that held for 15 years • The evolution of Google Search: sharding, moving the entire index into memory in 2001, softening query semantics pre-LLMs, and why retrieval pipelines already resemble modern LLM systems • Pareto frontier strategy: why you need both frontier “Pro” models and low-latency “Flash” models, and how distillation lets smaller models surpass prior generations • Distillation deep dive: ensembles → compression → logits as soft supervision, and why you need the biggest model to make the smallest one good • Latency as a first-class objective: why 10–50x lower latency changes UX entirely, and how future reasoning workloads will demand 10,000 tokens/sec • Energy-based thinking: picojoules per bit, why moving data costs 1000x more than a multiply, batching through the lens of energy, and speculative decoding as amortization • TPU co-design: predicting ML workloads 2–6 years out, speculative hardware features, precision reduction, sparsity, and the constant feedback loop between model architecture and silicon • Sparse models and “outrageously large” networks: trillions of parameters with 1–5% activation, and why sparsity was always the right abstraction • Unified vs. specialized models: abandoning symbolic systems, why general multimodal models tend to dominate vertical silos, and when vertical fine-tuning still makes sense • Long context and the illusion of scale: beyond needle-in-a-haystack benchmarks toward systems that narrow trillions of tokens to 117 relevant documents • Personalized AI: attending to your emails, photos, and documents (with permission), and why retrieval + reasoning will unlock deeply personal assistants • Coding agents: 50 AI interns, crisp specifications as a new core skill, and how ultra-low latency will reshape human–agent collaboration • Why ideas still matter: transformers, sparsity, RL, hardware, systems — scaling wasn’t blind; the pieces had to multiply together Substack Article w/Show Notes: https://www.latent.space/p/jeffdean — Jeff Dean • LinkedIn: https://www.linkedin.com/in/jeff-dean-8b212555 • X: https://x.com/jeffdean Google • https://google.com • https://deepmind.google 00:00:00 Intro 00:01:31 Frontier vs Flash & Distillation Strategy 00:05:09 Distillation, RL & Flash Economic Advantage 00:07:35 Flash in Products + Importance of Latency 00:11:11 Benchmarks, Long Context & Real Use Cases 00:15:01 Attending to Trillions of Tokens & Multimodality 00:20:11 LLM Search & Google Search Evolution 00:24:09 Systems Design Principles + Latency Numbers 00:32:09 Energy, Batching & TPU Co-Design 00:42:21 Research Frontiers: Reliability & RL Challenges 00:46:27 Unified Models vs Symbolic Systems (IMO) 00:50:38 Knowledge vs Reasoning + Vertical/Modular Models 00:55:58 Multilingual + Low-Resource Language Insights 00:57:58 Vision-Language Representations Example 01:07:15 Gemini Origin Story + Organizational Memo 01:09:27 Coding with AI & Agent Interaction Style 01:14:26 Prompting Skills & Spec Design 01:19:54 Latency Predictions & Tokens/sec Vision 01:21:29 Future Predictions: Personal Models & Hardware 01:23:11 Closing
The surprising case for AI judges
Bridget McCormack of the American Arbitration Association explains how AI-assisted arbitration works and what it might mean for trust and justice.
Harness engineering: leveraging Codex in an agent-first world
By Ryan Lopopolo, Member of the Technical Staff
Building the GitHub for RL Environments: Prime Intellect's Will Brown & Johannes Hagemann
Will Brown and Johannes Hagemann of Prime Intellect discuss the shift from static prompting to "environment-based" AI development, and their Environments Hub, a platform designed to democratize frontier-level training. The conversation highlights a major shift: AI progress is moving toward Recursive Language Models that manage their own context and agentic RL that scales through trial and error. Will and Johannes describe their vision for the future in which every company will become an AI research lab. By leveraging institutional knowledge as training data, businesses can build models with decades of experience that far outperform generic, off-the-shelf systems. Hosted by Sonya Huang, Sequoia Capital 00:00 Introduction 01:50 Understanding Frontier Lab Training and RL Hub 02:53 The Importance of Customization in AI Models 04:36 Harnessing the Power of Environments in AI 23:14 Evaluating Data Quality with Reinforcement Learning 24:17 Constructing Realistic Cybersecurity Environments 25:12 Designing Efficient Simulation Environments 29:04 The Role of Human Data in Model Training 33:29 Future Research and Vision
How I Use Claude Code
The research-plan-implement workflow I use to build software with Claude Code, and why I never let it write code until I've approved a written plan.
How Generative and Agentic AI Shift Concern from Technical Debt to Cognitive Debt
The term technical debt is often used to refer to the accumulation of design or implementation choices that later make the software harder and more costly to understand, modify, or extend over time...
Why All Mammograms Should Incorporate A.I.
A very impressive body of evidence has accumulated
The Rise of Spec Driven Development
Writing about AI, geo, culture, media, data, and the ways they interact.