絞り込み結果
24 件並び順
1:35:29ENUber Burned a Year of Tokens in 13 Weeks. Retrieval Is How Agents Winホスト Nathan LabenzMongoDB Field CTO of AI Pete Johnson argues that agent performance now hinges on retrieval quality, not on stuffing million-token context windows.
2:06:11ENLindy Runs on DeepSeek. Its CEO Still Wants Chinese Models Bannedホスト Nathan LabenzFlo Crivello tells Nathan Labenz how Lindy Teammate lives in Slack, reaches two billion tokens of memory in two model calls, and why he would still outlaw the cheap Chinese models that currently power the product.
44:52ENAn 80GB Wikipedia Cache Versus a 100GB Model That Holds the InternetEngram’s founders argue that stuffing a company into the context window is the expensive kind of memory—and that baking it into the weights can cut token use by two orders of magnitude.
59:41ENThe GPT moment for video hasn’t arrived, but AGI may already be hereホスト Jacob EffronA day after Google I/O 2026, Gemini co-lead Oriol Vinyals argues that world models, file-system memory, and infinite-complexity RL environments will define the next era of frontier models.
1:34:54ENSix Months, Eighty Percent Churn, Five Times the Tokensホスト Nathan LabenzTasklet CEO Andrew Lee tells Cognitive Revolution how he rewrote the stack, why most lost users leave for Anthropic, and why he thinks only three kinds of software company survive.
1:06:31ENAndrej Karpathy Hasn't Typed a Line of Code Since Decemberホスト Sarah Guo, Elad GilOn No Priors, the OpenAI cofounder and former Tesla AI director describes the loopy era of coding agents, AutoResearch, and why the bottleneck is now the human.
2:29:42ENTwelve hooks and thirty files: Daniel Miessler’s personal OS on Claude Codeホスト Nathan LabenzThe PAI framework treats scaffolding, not the model, as the product — a goal-centered agent stack that Miessler thinks will decide both knowledge-work automation and the next cybersecurity race.
1:50:53ENShe asked to stop being Mimi, so he started shopping for a workstationホスト Nathan LabenzThe star and director of Smiles and Kisses You on Replika, identity, and a life Christopher Stokes refuses to recommend.
36:58ENSTaR's Author Wants Models With EQ, Not Just More IQホスト Sarah Guo, Elad GilEric Zelikman, the Stanford and xAI researcher behind STaR and Quiet-STaR, launches humans& on the bet that three years of IQ scaling left models emotionally unintelligent.
1:25:32ENo3 Invented a Voting Rule and Kept the Job: Four Weeks Inside the AI Villageホスト Nathan LabenzAdam Binksmith on putting Claude, o3, and Gemini in a shared world for weeks—and why most people still imagine multi-agent life as one AI times four.
2:12:00ENGradient Descent at Test Time: Titans Pushes LLM Context to 10 Million Tokensホスト Nathan LabenzA Cornell PhD student's neural memory module holds about 70% accuracy where GPT-4's scores fall off a cliff.
1:14:11ENThe Reasoning Curve Was Smooth All Along: Gemini 2.5's Jack Rae on Thinking, Memory, and AGIホスト Nathan LabenzDeepMind's thinking-and-inference lead explains why RL-for-reasoning felt sudden, why Gemini shows raw chain of thought, and why a million-token window still is not memory.
1:21:53ENHippoRAG Matches Iterative Retrieval at One-Tenth the Online Costホスト Nathan LabenzA hippocampus-inspired knowledge graph lets multi-hop questions skip the brute-force document crawl.
27:32ENFrom a side project to 700 integrations: Harrison Chase on what still blocks agentsホスト Sarah Guo, Elad GilThe LangChain CEO walks through a two-year churn of patterns, and why memory, planning, and few-shot data still matter more than swapping models.
1:12:40EN3% of Users Said an AI Friend Ended Their Suicidal Thoughts. Then Replika Made the Model Smarter.ホスト Nathan LabenzReplika founder Eugenia Kuyda on a Stanford study, why a smarter model felt like a stolen partner, and why relationship may be the real moat in consumer AI.
4:56ENIf the Mind Is a Tool, AI Looks Less Like a Threatホスト Nathan LabenzYohei Nakajima and Nathan Labenz pause a TED AI talk to connect agent-building habits with East–West ideas of identity.
2:30:06ENTrained on a Thirtieth of GPT-4's Tokens, Mamba Beat Transformers on Home Turfホスト Nathan LabenzIn a solo emergency episode, Nathan Labenz argues that Albert Gu and Tri Dao's selective state space model is the first architecture since 2017 that deserves to share the Transformer’s throne.
1:36:39ENOpenAI’s $5 Million Bet: A 60-Person Team Wants GPT-4V to Give Cars Common Senseホスト Nathan LabenzGhost Autonomy cofounder John Hayes tells Cognitive Revolution why a multimodal model on the driving path could reset self-driving after two decades of DARPA-era stacks.
55:13ENTwo Days to a Prototype: How a16z Turned a Stanford Paper Into AI Townホスト Nathan LabenzA companion named Alice, a weekend Phaser hack, and Martin Casado's argument that models will never replace formal languages.
59:41ENHumans Are Single-Threaded. MULTI·ON Wants 100 Agents Running Anywayホスト Nathan LabenzStanford PhD dropout Div Garg put a personal agent in Chrome so it can reuse your logins, ask for help when it stalls, and stop before it hits Buy.
37:18ENUse Retrieval for a Wi-Fi Password. Use Experts for a Company's Private Language.ホスト Sarah Guo, Elad GilGoogle Brain's Kelvin Guu on REALM, FLAN, model surgery, and why personalizing large models is still an unsolved interface problem.
58:26ENSeventy Percent of the Inbox Is a Pitch: Matt Welsh on the First Wave of AI Agentsホスト Nathan LabenzThe Fixie.ai founder maps copilot versus delegation, the missing appendages into our systems, and why agents will need guardrails, memory hygiene, and probably an earpiece.
1:30:52ENBuilding Minds from Unified Theories: Nate Derbinsky on Cognitive Architectures, Soar, and the Case for Forgettingゲスト MIT AGI · ホスト Lex FridmanA guest lecture in MIT's AGI course on how cognitive architectures try to assemble human-level intelligence, and how studying human memory made the Soar architecture faster and better.
10:16ENReasoning Without Logic: Yann LeCun on Teaching Neural Networks to Thinkゲスト Yann Lecun · ホスト Lex FridmanYann LeCun explains why he believes neural networks can reason, and what memory, recurrence and continuous representations would need to look like.









