Загрузка видео...

Не удалось загрузить видео

На главную

Long Live Outputmaxxing: AI compute grids, Anthropic’s coding takeoff, data center backlash, & frontier systems AMP PBC founder Anjney Midha explains why 95% GPU utilization was considered an outage at Google, why the AI race is no longer just about buying more GPUs, how AMP is trying to make...

latentspacepod's profile picture

Latent.Space

31,303 subscribers

30,879 просмотров • 2 месяцев назад •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

AI is getting much better at hacking and fast. Data from the UK AI Security Institute, the Center for AI Standards and Innovations and others, shows how AI models are becoming dramatically more capable at cyberattacks. Sky's Rowland Manthorpe explains why this is no longer just a warning about one powerful model, but a much bigger industry-wide shift. 📺
2:58

AI is getting much better at hacking and fast. Data from the UK AI Security Institute, the Center for AI Standards and Innovations and others, shows how AI models are becoming dramatically more capable at cyberattacks. Sky's Rowland Manthorpe explains why this is no longer just a warning about one powerful model, but a much bigger industry-wide shift. 📺

Sky News

18,021 просмотров • 3 месяцев назад

Had a great talk with Anjney Midha about why he left a16z to start AMP PBC (raised $1.3B), his early bet on Anthropic, the importance of Japan in AI, his class at Stanford and a lot more. I didn't make him cry like Harry Stebbings, but I think it was a fun conversation about the future of AI Infra. Enjoy! 00:00 Introduction to AMP and AI Infrastructure 03:07 Transition from A16Z to AMP 06:01 Understanding AMP's Role in AI Compute 08:57 The Concept of a Compute Grid 11:59 The Need for Standardization in AI Infrastructure 14:59 Education and Collaboration in AI Development 17:59 Global Perspectives on AI Infrastructure 21:03 Japan's Role in AI and Robotics 23:14 Cultural Connections: Japan and Germany 25:08 The Journey to Teaching at Stanford 26:04 Security in Infrastructure: Lessons Learned 27:50 Cybersecurity and AI: A New Frontier 29:27 The Importance of Redundancy in Cybersecurity 30:39 Investing in AI: The Anthropic Experience 32:54 Wealth Creation and Public Participation 34:06 Infrastructure Bottlenecks in AI Development 36:57 Global Perspectives on Data Centers 39:11 The Need for Education on AI Benefits 42:07 Learning from China: A Systems Approach 45:47 Conclusion: Coordinating the Western Frontier
46:14

Had a great talk with Anjney Midha about why he left a16z to start AMP PBC (raised $1.3B), his early bet on Anthropic, the importance of Japan in AI, his class at Stanford and a lot more. I didn't make him cry like Harry Stebbings, but I think it was a fun conversation about the future of AI Infra. Enjoy! 00:00 Introduction to AMP and AI Infrastructure 03:07 Transition from A16Z to AMP 06:01 Understanding AMP's Role in AI Compute 08:57 The Concept of a Compute Grid 11:59 The Need for Standardization in AI Infrastructure 14:59 Education and Collaboration in AI Development 17:59 Global Perspectives on AI Infrastructure 21:03 Japan's Role in AI and Robotics 23:14 Cultural Connections: Japan and Germany 25:08 The Journey to Teaching at Stanford 26:04 Security in Infrastructure: Lessons Learned 27:50 Cybersecurity and AI: A New Frontier 29:27 The Importance of Redundancy in Cybersecurity 30:39 Investing in AI: The Anthropic Experience 32:54 Wealth Creation and Public Participation 34:06 Infrastructure Bottlenecks in AI Development 36:57 Global Perspectives on Data Centers 39:11 The Need for Education on AI Benefits 42:07 Learning from China: A Systems Approach 45:47 Conclusion: Coordinating the Western Frontier

Nathan Lands

28,872 просмотров • 4 месяцев назад

"Compute is the currency of the future." "Compute is the new OIL." Great, so how do I invest? You can play with the NVIDIA stock, or spin up an AI-Ready Data Center. Challenges of Spinning up AI-Ready Compute: Cost: Even a small cluster of H100 machines is upward to $10M GPU Scarcity: Unallocated or “underutilized” H100, A100, and RTX4090 GPUs are almost non-existent. Very Specific Physical & Technical Requirements: A data center with the right infrastructure is also very rare. There are only 10 in North America that deploy thousands of GPUs, in the open market. Domain Expertise: Teams of experts with the years of experience required to operate, deploy, and manage AI compute at scale are as rare as the GPUs they deploy. Exabits checks all of these boxes and we are the premier Compute Tokenization Platform. We are backed by REAL COMPUTE used by REAL CUSTOMERS.
0:30

"Compute is the currency of the future." "Compute is the new OIL." Great, so how do I invest? You can play with the NVIDIA stock, or spin up an AI-Ready Data Center. Challenges of Spinning up AI-Ready Compute: Cost: Even a small cluster of H100 machines is upward to $10M GPU Scarcity: Unallocated or “underutilized” H100, A100, and RTX4090 GPUs are almost non-existent. Very Specific Physical & Technical Requirements: A data center with the right infrastructure is also very rare. There are only 10 in North America that deploy thousands of GPUs, in the open market. Domain Expertise: Teams of experts with the years of experience required to operate, deploy, and manage AI compute at scale are as rare as the GPUs they deploy. Exabits checks all of these boxes and we are the premier Compute Tokenization Platform. We are backed by REAL COMPUTE used by REAL CUSTOMERS.

Exabits

13,896 просмотров • 2 лет назад

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. We sat down with Jeff to unpack what it really means to “own the Pareto frontier,” why distillation is the quiet force behind every generation of faster, cheaper models, how energy not FLOPs is becoming the true constraint on AI compute, what it takes to co-design hardware and models 2–6 years into the future, why unified multimodal systems will outperform specialized ones, what it was like leading the charge to unify all of Google’s AI teams, and his prediction that deeply personalized models with access to your full digital context will redefine what useful AI looks like. Jeff Dean Google DeepMind Google
0:44

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. We sat down with Jeff to unpack what it really means to “own the Pareto frontier,” why distillation is the quiet force behind every generation of faster, cheaper models, how energy not FLOPs is becoming the true constraint on AI compute, what it takes to co-design hardware and models 2–6 years into the future, why unified multimodal systems will outperform specialized ones, what it was like leading the charge to unify all of Google’s AI teams, and his prediction that deeply personalized models with access to your full digital context will redefine what useful AI looks like. Jeff Dean Google DeepMind Google

Latent.Space

529,014 просмотров • 6 месяцев назад

Why AI Can Now Make Discoveries - my conversation with Dan Roberts, Lead of the Foundations of Reinforcement Learning team at OpenAI 00:00 Intro: AI's wild week in mathematics 01:21 What OpenAI's Foundations of RL team does 03:08 Dan's journey: from black holes and quantum gravity to frontier AI 07:04 Are AI systems becoming useful for real science 08:21 The AI math moment: Erdős, OpenAI, DeepMind, and Anthropic 08:52 Why the OpenAI result was an act of exploration 10:25 OpenAI vs. DeepMind: informal reasoning vs. formal proof 12:13 RL 101: learning by doing, not just watching 15:10 Why reinforcement learning works 15:58 How RL breaks: sparse feedback and long-horizon tasks 17:03 RLHF: how human feedback shaped early language models 18:48 Move 37, self-play, and the search for novel strategies 22:16 Explore vs. exploit in scientific discovery 24:49 Why RL may now be "the cake," not the cherry on top 25:46 Why RL started working with large language models 27:29 Is RL "sucking supervision through a straw"? 28:47 Why language may be the grounding layer for intelligence 31:46 A contrarian take on the Bitter Lesson 32:41 What test-time compute actually is 34:50 How RL gives models the ability to think 35:40 Verifiable rewards, math, coding, and the messy real world 38:00 What physics can teach us about AI 42:08 Is there a thermodynamics of AI? 43:08 From Erdős problems to Einstein-level AI 45:16 Is AI already doing original science? 45:51 How far are we from AI automating AI research 47:41 Why Dan is excited about the future of science
49:06

Why AI Can Now Make Discoveries - my conversation with Dan Roberts, Lead of the Foundations of Reinforcement Learning team at OpenAI 00:00 Intro: AI's wild week in mathematics 01:21 What OpenAI's Foundations of RL team does 03:08 Dan's journey: from black holes and quantum gravity to frontier AI 07:04 Are AI systems becoming useful for real science 08:21 The AI math moment: Erdős, OpenAI, DeepMind, and Anthropic 08:52 Why the OpenAI result was an act of exploration 10:25 OpenAI vs. DeepMind: informal reasoning vs. formal proof 12:13 RL 101: learning by doing, not just watching 15:10 Why reinforcement learning works 15:58 How RL breaks: sparse feedback and long-horizon tasks 17:03 RLHF: how human feedback shaped early language models 18:48 Move 37, self-play, and the search for novel strategies 22:16 Explore vs. exploit in scientific discovery 24:49 Why RL may now be "the cake," not the cherry on top 25:46 Why RL started working with large language models 27:29 Is RL "sucking supervision through a straw"? 28:47 Why language may be the grounding layer for intelligence 31:46 A contrarian take on the Bitter Lesson 32:41 What test-time compute actually is 34:50 How RL gives models the ability to think 35:40 Verifiable rewards, math, coding, and the messy real world 38:00 What physics can teach us about AI 42:08 Is there a thermodynamics of AI? 43:08 From Erdős problems to Einstein-level AI 45:16 Is AI already doing original science? 45:51 How far are we from AI automating AI research 47:41 Why Dan is excited about the future of science

Matt Turck

66,936 просмотров • 2 месяцев назад

SoftBank is creating the biggest data center. On Fox Business, Arm CEO and SoftBank Group International CEO Rene Haas was asked what he thinks about Meta selling compute, which could mean a compute supply glut. Haas disagreed, pointing to SpaceX's compute deal with Anthropic as proof of real demand, not oversupply. He said Meta's situation is similar, and that only a handful of companies, SpaceX, Meta, and SoftBank, have the financial muscle to compete at gigawatt scale, which is why SoftBank is moving in.
0:44

SoftBank is creating the biggest data center. On Fox Business, Arm CEO and SoftBank Group International CEO Rene Haas was asked what he thinks about Meta selling compute, which could mean a compute supply glut. Haas disagreed, pointing to SpaceX's compute deal with Anthropic as proof of real demand, not oversupply. He said Meta's situation is similar, and that only a handful of companies, SpaceX, Meta, and SoftBank, have the financial muscle to compete at gigawatt scale, which is why SoftBank is moving in.

Lark Davis

34,017 просмотров • 1 месяц назад

DROPS E27: Ben from Talus Labs: Building the Future of Decentralized AI Agents In this episode Ben breaks down why AI isn’t a bubble, how Talus is redefining on-chain agents, and why optimism is something you earn and not something you just feel. We talk about: - Why AI is not a bubble and how mega-cap tech is funding the next AI wave through massive CapEx - How Talus enables decentralized AI agents that can act and transact trustlessly - The three big crypto × AI intersections: compute, data ownership, and intelligent on-chain workflows - Why AI needs verifiable, censorship-resistant execution layers and why blockchains provide that - Nuclear energy as a prerequisite for AGI-scale compute demand - Why SWE’s object model + parallelization is perfect for autonomous agent payments and priority fees And much more! Timestamps: 0:00 Introduction 1:31 Welcome To Drops 2:10 Getting Backers Like Polychain 3:16 Building While In College 7:51 VC’s vs Founders Differences Explained 10:01 Is There An AI Bubble 11:53 How Do We Quantify The AI Opportunity In Crypto 12:33 AI Is Difficult To Invest In, Why? 13:35 Risks To Look Out For To Ensure AI Investments Deliver 16:47 Energy Being A Big Risk For AI 20:50 How You Became Fascinated By Crypto & Blockchain 22:00 Why Does AI & Blockchain Intersection Make Sense Compute Side 24:39 Why Does AI & Blockchain Intersection Make Sense for Data Ownership 26:30 Complex Workflows With Intelligent Execution Meaning 28:19 Why Do We Need This On Blockchain 30:26 How Adding This Decentralized Layer Helps With The Black Box 32:01 Explain Talus To Your Mom 33:39 Talus Solved Blockchain Determinism Limitation Explained 35:43 Two Examples That Can Be Built On The Zapier Of Web 3 39:32 Why Get Investment From SUI 42:56 What’s Needed To Get To The Next Level 43:35 One Takeaway From Conversation 43:51 One Thing Pessimistic People Are Missing
45:43

DROPS E27: Ben from Talus Labs: Building the Future of Decentralized AI Agents In this episode Ben breaks down why AI isn’t a bubble, how Talus is redefining on-chain agents, and why optimism is something you earn and not something you just feel. We talk about: - Why AI is not a bubble and how mega-cap tech is funding the next AI wave through massive CapEx - How Talus enables decentralized AI agents that can act and transact trustlessly - The three big crypto × AI intersections: compute, data ownership, and intelligent on-chain workflows - Why AI needs verifiable, censorship-resistant execution layers and why blockchains provide that - Nuclear energy as a prerequisite for AGI-scale compute demand - Why SWE’s object model + parallelization is perfect for autonomous agent payments and priority fees And much more! Timestamps: 0:00 Introduction 1:31 Welcome To Drops 2:10 Getting Backers Like Polychain 3:16 Building While In College 7:51 VC’s vs Founders Differences Explained 10:01 Is There An AI Bubble 11:53 How Do We Quantify The AI Opportunity In Crypto 12:33 AI Is Difficult To Invest In, Why? 13:35 Risks To Look Out For To Ensure AI Investments Deliver 16:47 Energy Being A Big Risk For AI 20:50 How You Became Fascinated By Crypto & Blockchain 22:00 Why Does AI & Blockchain Intersection Make Sense Compute Side 24:39 Why Does AI & Blockchain Intersection Make Sense for Data Ownership 26:30 Complex Workflows With Intelligent Execution Meaning 28:19 Why Do We Need This On Blockchain 30:26 How Adding This Decentralized Layer Helps With The Black Box 32:01 Explain Talus To Your Mom 33:39 Talus Solved Blockchain Determinism Limitation Explained 35:43 Two Examples That Can Be Built On The Zapier Of Web 3 39:32 Why Get Investment From SUI 42:56 What’s Needed To Get To The Next Level 43:35 One Takeaway From Conversation 43:51 One Thing Pessimistic People Are Missing

MR SHIFT 🦁

41,057 просмотров • 8 месяцев назад

Joelle built the AI lab at Meta She worked and learnt from Zuck She is one of the leading minds in AI What everyone thinks they know about AI but get wrong 👇 - Why Scaling Laws Will Continue - Why You Can’t Buy Success in AI with Talent Acquisitions - Future of Synthetic Data & What It Means for Models - Why AI Coding is Akin to Image Generation in 2015 Spotify 👉 Youtube 👉 Apple Podcasts 👉 Timestamps: 00:00 Intro 01:16 How Meta Shaped How I Think About AI Research 02:22 Challenges in Reinforcement Learning 08:33 Is It Possible To Be Capital Efficient in AI 13:47 AI in Enterprise: Efficiency and Adoption 21:51 Security Concerns with AI Agents 28:06 Can Zuck Win By Buying The Superstars of AI 32:11 The Rising Cost of Data 36:38 Synthetic Data and Model Degradation 38:42 Why AI Coding is Akin to Image Generation in 2015 51:16 If Joelle Was a VC Where Would She Invest? 51:50 Quick-Fire Round: Lessons from Zuck, Biggest Mindset Shift My 5 takeaways with Joelle Pineau 👇
59:02

Joelle built the AI lab at Meta She worked and learnt from Zuck She is one of the leading minds in AI What everyone thinks they know about AI but get wrong 👇 - Why Scaling Laws Will Continue - Why You Can’t Buy Success in AI with Talent Acquisitions - Future of Synthetic Data & What It Means for Models - Why AI Coding is Akin to Image Generation in 2015 Spotify 👉 Youtube 👉 Apple Podcasts 👉 Timestamps: 00:00 Intro 01:16 How Meta Shaped How I Think About AI Research 02:22 Challenges in Reinforcement Learning 08:33 Is It Possible To Be Capital Efficient in AI 13:47 AI in Enterprise: Efficiency and Adoption 21:51 Security Concerns with AI Agents 28:06 Can Zuck Win By Buying The Superstars of AI 32:11 The Rising Cost of Data 36:38 Synthetic Data and Model Degradation 38:42 Why AI Coding is Akin to Image Generation in 2015 51:16 If Joelle Was a VC Where Would She Invest? 51:50 Quick-Fire Round: Lessons from Zuck, Biggest Mindset Shift My 5 takeaways with Joelle Pineau 👇

Harry Stebbings

19,398 просмотров • 10 месяцев назад

Baseten's Charlie O'Neill on why the gap to the frontier is just compute and data: "Despite OpenAI and Anthropic and the closed source labs having more compute and more data, the recipe is the same. We're all scaling the same recipe here, and we're at slightly different points of the scaling curve. It's not like OpenAI or Anthropic have any secret sauce that differentiates them scaling up their multi-trillion parameter models from the rest of the open source ecosystem." "It really is just a matter of taking the core underlying architecture and training recipe, and throwing more compute and data at it, which is really promising." "Of course, these closed source labs probably have a long tail of optimizations they've made, which we're gonna continue to try and figure out in the open source community. But the core mechanism is there." "K3 is a great example of, well, if you do scale this recipe up, we're gonna get a pretty good model that's very, very close to the frontier, if not frontier itself." Charlie O'Neill Baseten
1:51

Baseten's Charlie O'Neill on why the gap to the frontier is just compute and data: "Despite OpenAI and Anthropic and the closed source labs having more compute and more data, the recipe is the same. We're all scaling the same recipe here, and we're at slightly different points of the scaling curve. It's not like OpenAI or Anthropic have any secret sauce that differentiates them scaling up their multi-trillion parameter models from the rest of the open source ecosystem." "It really is just a matter of taking the core underlying architecture and training recipe, and throwing more compute and data at it, which is really promising." "Of course, these closed source labs probably have a long tail of optimizations they've made, which we're gonna continue to try and figure out in the open source community. But the core mechanism is there." "K3 is a great example of, well, if you do scale this recipe up, we're gonna get a pretty good model that's very, very close to the frontier, if not frontier itself." Charlie O'Neill Baseten

MTS

17,404 просмотров • 1 месяц назад

BREAKING: How a $15+ Trillion AUM Firm is Thinking About the Systemic $10T Rebuild Around AI Inside BlackRock's Fundamental Equities w/ Managing Director & Head of Global Tech, Tony Kim (Tony Kim) The irony of "the trillion dollars of CapEx this year & the $10T over the next 5 years that are coming.. is to move data centimeters & millimeters. That's AI." "Today we're all talking about compute, compute, compute. I think the primacy of memory will become even more important." Tony invests across public & private tech markets, covering semiconductors, memory, data centers, power, software, quantum, & robotics We Cover: › The shift from a software centric world to a compute centric world › The 10-20-30 trillion market cap breakdown, software, Mag 7, and hardware › RAMpocalypse, why memory becomes as important as compute › How BlackRock allocates, 90% in the 3-year AI vortex, 10% on frontier bets converging on 2030 › 140 Chinese robotics companies and the 30-40 IPOs coming this year › Following the "token flow" Special thank you to Brex, MongoDB, & AssemblyAI for helping make this RAISE AI Summit mini-series in Paris, France happen. 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 (00:00) Tony Kim, Head of BlackRock Fundamental Equities Global Technology (01:10) RAISE AI Summit in Paris (03:52) Why Compute now rules everything (05:17) Why old data centers can't survive AI (07:43) Why data centers are ditching Copper for Light (10:44) The shortage nobody saw coming: RAM (15:20) Only three companies control memory (16:03) Tony's Playbook for Investing in the AI Era (18:16) The 20-year lie: "Compute is just a Commodity" (22:38) How Hardware quietly became bigger than Software (27:49) The Investing Rule: Will you still be cool in 5 years? (32:41) 2030: the year every frontier technology bet converges (38:16) Chips were never a Commodity (41:30) Rack design, materials science, & the physical-world renaissance (43:30) Inside the architecture of a Robot's mind (45:03) Why China Is winning the Robotics race (47:37) Why the biggest Robotics market might be companionship (51:34) Speech models are growing faster than anyone expected (53:07) "Token Flow": Tony's Framework for the Future Enterprise (57:56) Are PE roll-ups the next big AI disruption play? (1:00:48) What Could Go Right (& Wrong) in AI This Year (1:03:51) The mentors who shaped Tony Kim's worldview
1:08:44

BREAKING: How a $15+ Trillion AUM Firm is Thinking About the Systemic $10T Rebuild Around AI Inside BlackRock's Fundamental Equities w/ Managing Director & Head of Global Tech, Tony Kim (Tony Kim) The irony of "the trillion dollars of CapEx this year & the $10T over the next 5 years that are coming.. is to move data centimeters & millimeters. That's AI." "Today we're all talking about compute, compute, compute. I think the primacy of memory will become even more important." Tony invests across public & private tech markets, covering semiconductors, memory, data centers, power, software, quantum, & robotics We Cover: › The shift from a software centric world to a compute centric world › The 10-20-30 trillion market cap breakdown, software, Mag 7, and hardware › RAMpocalypse, why memory becomes as important as compute › How BlackRock allocates, 90% in the 3-year AI vortex, 10% on frontier bets converging on 2030 › 140 Chinese robotics companies and the 30-40 IPOs coming this year › Following the "token flow" Special thank you to Brex, MongoDB, & AssemblyAI for helping make this RAISE AI Summit mini-series in Paris, France happen. 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 (00:00) Tony Kim, Head of BlackRock Fundamental Equities Global Technology (01:10) RAISE AI Summit in Paris (03:52) Why Compute now rules everything (05:17) Why old data centers can't survive AI (07:43) Why data centers are ditching Copper for Light (10:44) The shortage nobody saw coming: RAM (15:20) Only three companies control memory (16:03) Tony's Playbook for Investing in the AI Era (18:16) The 20-year lie: "Compute is just a Commodity" (22:38) How Hardware quietly became bigger than Software (27:49) The Investing Rule: Will you still be cool in 5 years? (32:41) 2030: the year every frontier technology bet converges (38:16) Chips were never a Commodity (41:30) Rack design, materials science, & the physical-world renaissance (43:30) Inside the architecture of a Robot's mind (45:03) Why China Is winning the Robotics race (47:37) Why the biggest Robotics market might be companionship (51:34) Speech models are growing faster than anyone expected (53:07) "Token Flow": Tony's Framework for the Future Enterprise (57:56) Are PE roll-ups the next big AI disruption play? (1:00:48) What Could Go Right (& Wrong) in AI This Year (1:03:51) The mentors who shaped Tony Kim's worldview

Molly O’Shea

356,369 просмотров • 1 месяц назад

“What if the most important AI infrastructure isn’t the biggest model — but the one you can actually trust?” Sachi Miyasaki sits down with Karan Sirdesai Karan, Founder of Mira, to explore how Mira is approaching decentralized AI from a more focused angle: making machine intelligence reliable enough for real-world use. They discuss why Mira is not trying to become a full-stack AI protocol, how its approach differs from networks like Bittensor, Ritual, and Sahara, and why focusing on trust and verification could make Mira a stronger fit inside the wider AI infrastructure stack. The episode also breaks down Mira’s reliability architecture, from its core verification layer to node-level consensus, and why the next evolution of AI APIs may not be about simply accessing more models, but about knowing whether the answers you receive can actually be trusted.
2:22

“What if the most important AI infrastructure isn’t the biggest model — but the one you can actually trust?” Sachi Miyasaki sits down with Karan Sirdesai Karan, Founder of Mira, to explore how Mira is approaching decentralized AI from a more focused angle: making machine intelligence reliable enough for real-world use. They discuss why Mira is not trying to become a full-stack AI protocol, how its approach differs from networks like Bittensor, Ritual, and Sahara, and why focusing on trust and verification could make Mira a stronger fit inside the wider AI infrastructure stack. The episode also breaks down Mira’s reliability architecture, from its core verification layer to node-level consensus, and why the next evolution of AI APIs may not be about simply accessing more models, but about knowing whether the answers you receive can actually be trusted.

BeaconLayer

79,111 просмотров • 1 месяц назад

My conversation with Tarun Chitra As a co-founder of Gauntlet and GP at Robot Ventures, Tarun has one of the sharpest frameworks for understanding market structure across both crypto and AI. In this episode we dig into why open source AI is unbundling faster than most people expect, and how the resulting stack looks surprisingly similar to DeFi. We spend time mapping the AI infrastructure layers directly onto crypto primitives and examining where value is actually going to accrue as models, harnesses, routers, and inference providers separate. At the center of the conversation is the belief that AI’s unbundling is creating a new competitive order-flow market (data centers competing like nodes, MEV-like dynamics for tokens/GPUs) while crypto itself has settled into a more mature “TradFi plus+” phase focused on trading, payments, and bringing real assets on-chain. We discuss: - The current state of crypto as TradFi+ and the decline of speculative narratives - Why AI is killing Bitcoin mining economics and weakening ETH value accrual - The architectural parallel between AI stacks and DeFi (Harnesses = Wallets, Routers = DEX aggregators, Models = Protocols, Inference Providers = LPs) - Why open source models are unbundling faster than traditional software - Agents as the next interface layer and the potential unbundling of ETFs -Cryptography, verifiable compute, and turning GPUs into digital assets - Onchain compute trading as the real crypto × AI opportunity - Sustainable business models and where value will ultimately capture Timestamps: 0:00 – Introduction & State of the Crypto Market 2:00 – Speculative Narratives Fade, Trading & Payments Remain 7:00 – AI’s Impact on Bitcoin Economics & Data Center Opportunity Cost 9:00 – ETH Value Accrual, Solana Positioning & DeFi Token Sustainability 15:00 – Trading Design Space & On-Chain Volume Upside 25:00 – AI Unbundling Thesis: Open-Source Models vs Data Centers 35:00 – The DeFi Mapping (Harnesses, Routers, Models, Inference) 48:00 – Agents, Preference Expression & Unbundling Traditional Products 1:00:00 – Real-Time Harness Generation & Active Learning 1:05:00 – Cryptography, Verifiable Compute & On-Chain GPU Markets 1:10:00 – Closing Thoughts: Where Value Accrues Next Enjoy!
1:13:04

My conversation with Tarun Chitra As a co-founder of Gauntlet and GP at Robot Ventures, Tarun has one of the sharpest frameworks for understanding market structure across both crypto and AI. In this episode we dig into why open source AI is unbundling faster than most people expect, and how the resulting stack looks surprisingly similar to DeFi. We spend time mapping the AI infrastructure layers directly onto crypto primitives and examining where value is actually going to accrue as models, harnesses, routers, and inference providers separate. At the center of the conversation is the belief that AI’s unbundling is creating a new competitive order-flow market (data centers competing like nodes, MEV-like dynamics for tokens/GPUs) while crypto itself has settled into a more mature “TradFi plus+” phase focused on trading, payments, and bringing real assets on-chain. We discuss: - The current state of crypto as TradFi+ and the decline of speculative narratives - Why AI is killing Bitcoin mining economics and weakening ETH value accrual - The architectural parallel between AI stacks and DeFi (Harnesses = Wallets, Routers = DEX aggregators, Models = Protocols, Inference Providers = LPs) - Why open source models are unbundling faster than traditional software - Agents as the next interface layer and the potential unbundling of ETFs -Cryptography, verifiable compute, and turning GPUs into digital assets - Onchain compute trading as the real crypto × AI opportunity - Sustainable business models and where value will ultimately capture Timestamps: 0:00 – Introduction & State of the Crypto Market 2:00 – Speculative Narratives Fade, Trading & Payments Remain 7:00 – AI’s Impact on Bitcoin Economics & Data Center Opportunity Cost 9:00 – ETH Value Accrual, Solana Positioning & DeFi Token Sustainability 15:00 – Trading Design Space & On-Chain Volume Upside 25:00 – AI Unbundling Thesis: Open-Source Models vs Data Centers 35:00 – The DeFi Mapping (Harnesses, Routers, Models, Inference) 48:00 – Agents, Preference Expression & Unbundling Traditional Products 1:00:00 – Real-Time Harness Generation & Active Learning 1:05:00 – Cryptography, Verifiable Compute & On-Chain GPU Markets 1:10:00 – Closing Thoughts: Where Value Accrues Next Enjoy!

Logan Jastremski

132,018 просмотров • 21 дней назад

From early engineer at Glean to Partner at Menlo Ventures co-leading the Anthology Fund with Anthropic, Deedy Das has gone from shipping “boring” enterprise search to backing frontier labs, infra, and research plays like Anthropic, OpenRouter, Goodfire, Prime Intellect, and Whisper. We sat down with Deedy to unpack how Glean quietly built a real AI moat before LLMs were cool, why enterprise search is significantly different than consumer search, where value actually accrues in the model vs app stack, Anthropic’s rise and the importance of Claude Code, the Anthology Fund thesis, OpenRouter’s wedge, mechanistic interpretability, the coming compute arms race, coding agents and “LLM psychosis,” and what all of this means for AI engineers deciding what to build. Deedy swyx 🔜 NeurIPS + #DevWritersRetreat Alessio Fanelli
1:15

From early engineer at Glean to Partner at Menlo Ventures co-leading the Anthology Fund with Anthropic, Deedy Das has gone from shipping “boring” enterprise search to backing frontier labs, infra, and research plays like Anthropic, OpenRouter, Goodfire, Prime Intellect, and Whisper. We sat down with Deedy to unpack how Glean quietly built a real AI moat before LLMs were cool, why enterprise search is significantly different than consumer search, where value actually accrues in the model vs app stack, Anthropic’s rise and the importance of Claude Code, the Anthology Fund thesis, OpenRouter’s wedge, mechanistic interpretability, the coming compute arms race, coding agents and “LLM psychosis,” and what all of this means for AI engineers deciding what to build. Deedy swyx 🔜 NeurIPS + #DevWritersRetreat Alessio Fanelli

Latent.Space

269,074 просмотров • 9 месяцев назад

The CPO who regrets that product management exists Tom Verrilli (Tom Verrilli) is Chief Product Officer at Whatnot and a long-time product leader at Twitch and Twitter. We discuss: 🔸 How AI is exposing rampant “product theater” 🔸 The PM skill trending up fastest (hint: systems thinking) 🔸 Why AI’s biggest unlock for PMs is now in data science, not prototyping 🔸 Why you need to stop promoting your A-players out of the work they’re great at 🔸 Why “hire great people and get out of their way” is wrong Full episode for your enjoyment 👇
1:24:57

The CPO who regrets that product management exists Tom Verrilli (Tom Verrilli) is Chief Product Officer at Whatnot and a long-time product leader at Twitch and Twitter. We discuss: 🔸 How AI is exposing rampant “product theater” 🔸 The PM skill trending up fastest (hint: systems thinking) 🔸 Why AI’s biggest unlock for PMs is now in data science, not prototyping 🔸 Why you need to stop promoting your A-players out of the work they’re great at 🔸 Why “hire great people and get out of their way” is wrong Full episode for your enjoyment 👇

Lenny Rachitsky

133,026 просмотров • 29 дней назад

I spoke with Jordi Visser to break down why the last two weeks may be the most important stretch in bitcoin's history, along with the US-Japan yen intervention and a weakening labor market. We also discuss the AI stock rebound, physical AI and Tesla's Terafab, Google's talent exodus, and how to value SpaceX. YouTube: Spotify: Apple: TIMESTAMPS: 0:00 - Intro 0:54 - Why the last two weeks may be the most important in Bitcoin's history 8:17 - What should the Fed do next? 13:39 - AI stocks bounce hard: was that the bottom? 21:40 - Physical AI: Hadrian, Tesla's Terafab & the compute race 25:57 - How to position your portfolio for the compute boom 33:26 - What frontier models get wrong (and why big companies move slow) 38:40 - Google's talent exodus: is the stock at risk? 41:23 - Does calendar seasonality still matter? 42:43 - Valuing SpaceX: the compute business vs. the space hype 44:42 - What's coming in Jordi's video this week
47:30

I spoke with Jordi Visser to break down why the last two weeks may be the most important stretch in bitcoin's history, along with the US-Japan yen intervention and a weakening labor market. We also discuss the AI stock rebound, physical AI and Tesla's Terafab, Google's talent exodus, and how to value SpaceX. YouTube: Spotify: Apple: TIMESTAMPS: 0:00 - Intro 0:54 - Why the last two weeks may be the most important in Bitcoin's history 8:17 - What should the Fed do next? 13:39 - AI stocks bounce hard: was that the bottom? 21:40 - Physical AI: Hadrian, Tesla's Terafab & the compute race 25:57 - How to position your portfolio for the compute boom 33:26 - What frontier models get wrong (and why big companies move slow) 38:40 - Google's talent exodus: is the stock at risk? 41:23 - Does calendar seasonality still matter? 42:43 - Valuing SpaceX: the compute business vs. the space hype 44:42 - What's coming in Jordi's video this week

Anthony Pompliano 🌪

79,205 просмотров • 24 дней назад

AMP PBC Founder Anjney Midha says the massive amount of AI compute being wasted in the U.S. is a national security crisis: “Every dollar of long-term lease that an AI lab takes today, you lose about 30–40% of FLOPs in just bad scheduling. The nodes are just straight-up not allocated.” “Then within the chip, you only have like 15% of the chip being utilized. That’s called MFU, model FLOP utilization, because the chip is waiting around for some other process to complete: storage, memory, networking, or some other chip to hand it off.” “If you compound those two things, of every dollar, only 15% of FLOPs are actually being utilized. That 85%? That’s a national security crisis, because that’s what’s keeping us from staying at the frontier.” “Am I building up more supply? But also, are we utilizing the existing capacity to get the maximum output possible? Both those things are what’s going to keep us at the frontier.”
1:56

AMP PBC Founder Anjney Midha says the massive amount of AI compute being wasted in the U.S. is a national security crisis: “Every dollar of long-term lease that an AI lab takes today, you lose about 30–40% of FLOPs in just bad scheduling. The nodes are just straight-up not allocated.” “Then within the chip, you only have like 15% of the chip being utilized. That’s called MFU, model FLOP utilization, because the chip is waiting around for some other process to complete: storage, memory, networking, or some other chip to hand it off.” “If you compound those two things, of every dollar, only 15% of FLOPs are actually being utilized. That 85%? That’s a national security crisis, because that’s what’s keeping us from staying at the frontier.” “Am I building up more supply? But also, are we utilizing the existing capacity to get the maximum output possible? Both those things are what’s going to keep us at the frontier.”

TBPN

117,549 просмотров • 1 месяц назад

Open-source AI is catching up to proprietary frontier models, but the hardware needed to run them are gatekept. B3 is launching B3IQ so anyone can own their own intelligence. daryl and Viktoriya (B3) and Yorke Ξ Rhodes III 🟧 🇺🇦 (Microsoft/NYU) join Stateful, hosted by Franklin Bi In this episode, they discuss why owning your compute is the next unlock for open source AI: - Frontier labs have locked up hyperscaler capacity through 2030. Everyone else is out of line. - NYU students researching human trafficking and war zone evacuations get account-banned by closed models - Two years of renting an H200 node costs the same as buying it. Rent-to-own drops the entry cost to 30%. - Kimi K3 matches frontier performance on some coding tasks. You just need two H200 nodes to run it. - Idle compute gets matched with offtake, so the machine pays for itself 05:28 Content Filters and Token Limits at NYU 11:02 The Hyperscaler Long Tail Problem 16:16 Why Frontier Labs Locked In Their Model Gains 18:19 How the B3IQ Financing Model Works 20:10 The iPhone Moment for Compute 29:27 Why Own the Box When Neo Clouds Exist? 40:24 The Five-Year Vision
46:03

Open-source AI is catching up to proprietary frontier models, but the hardware needed to run them are gatekept. B3 is launching B3IQ so anyone can own their own intelligence. daryl and Viktoriya (B3) and Yorke Ξ Rhodes III 🟧 🇺🇦 (Microsoft/NYU) join Stateful, hosted by Franklin Bi In this episode, they discuss why owning your compute is the next unlock for open source AI: - Frontier labs have locked up hyperscaler capacity through 2030. Everyone else is out of line. - NYU students researching human trafficking and war zone evacuations get account-banned by closed models - Two years of renting an H200 node costs the same as buying it. Rent-to-own drops the entry cost to 30%. - Kimi K3 matches frontier performance on some coding tasks. You just need two H200 nodes to run it. - Idle compute gets matched with offtake, so the machine pays for itself 05:28 Content Filters and Token Limits at NYU 11:02 The Hyperscaler Long Tail Problem 16:16 Why Frontier Labs Locked In Their Model Gains 18:19 How the B3IQ Financing Model Works 20:10 The iPhone Moment for Compute 29:27 Why Own the Box When Neo Clouds Exist? 40:24 The Five-Year Vision

Pantera Capital

22,485 просмотров • 21 дней назад

Adaption Labs CEO Sara Hooker says data centers in space are "pretty bonkers", and explains what the current consensus gets wrong: "Most co-located hardware is pretty much for training. I think that's why you care [about Space], otherwise you can distribute. Inference compute, which is where everything's moving towards, you can spread that compute more easily. You can have multiple data centers [on Earth]." "So if you care about space, you probably only care about training compute. I think people underestimate the amount of failures that happen, and you don't want to get your training job interrupted." "The real issue is that GPUs still have failure rates. The 2% of GPUs that are just considered done every year, that's really your cost. It's how quickly you can replace those [in space] and what that looks like."
2:29

Adaption Labs CEO Sara Hooker says data centers in space are "pretty bonkers", and explains what the current consensus gets wrong: "Most co-located hardware is pretty much for training. I think that's why you care [about Space], otherwise you can distribute. Inference compute, which is where everything's moving towards, you can spread that compute more easily. You can have multiple data centers [on Earth]." "So if you care about space, you probably only care about training compute. I think people underestimate the amount of failures that happen, and you don't want to get your training job interrupted." "The real issue is that GPUs still have failure rates. The 2% of GPUs that are just considered done every year, that's really your cost. It's how quickly you can replace those [in space] and what that looks like."

TBPN

37,463 просмотров • 6 месяцев назад

What if your personal data became an asset you actually owned? In this May 2026 AMA, Matrix AI Network explores one of the most important questions facing AI, Web3, and the future digital economy: Who owns the data that powers artificial intelligence? As AI models continue to improve, the next major bottleneck is no longer computing power—it is access to high-quality, verified human data. The discussion covers how personal biodata can become a valuable digital asset, how data quality is measured, and why future AI companies may depend directly on users as data providers. Topics include: ✅ Biodata as a new asset class ✅ AI-grade data quality scoring ✅ Wearable devices for raw data ✅ Data ownership and monetization models ✅ Preventing fake data and device spoofing ✅ Real-world applications in healthcare, sports science and mental wellness ✅ Hypnus and the future biodata ecosystem ✅ Project direction The future may not belong to companies that own data. It belongs to the people who create it. Watch now 👇 #AI #Web3 #DataOwnership
29:51

What if your personal data became an asset you actually owned? In this May 2026 AMA, Matrix AI Network explores one of the most important questions facing AI, Web3, and the future digital economy: Who owns the data that powers artificial intelligence? As AI models continue to improve, the next major bottleneck is no longer computing power—it is access to high-quality, verified human data. The discussion covers how personal biodata can become a valuable digital asset, how data quality is measured, and why future AI companies may depend directly on users as data providers. Topics include: ✅ Biodata as a new asset class ✅ AI-grade data quality scoring ✅ Wearable devices for raw data ✅ Data ownership and monetization models ✅ Preventing fake data and device spoofing ✅ Real-world applications in healthcare, sports science and mental wellness ✅ Hypnus and the future biodata ecosystem ✅ Project direction The future may not belong to companies that own data. It belongs to the people who create it. Watch now 👇 #AI #Web3 #DataOwnership

Matrix AI Network

60,855 просмотров • 3 месяцев назад

$META is already using AI to improve its ad engine which makes it a clear winner in the AI economy. The market still wants proof that its massive infrastructure spend can create revenue beyond better ads which is why the compute business matters. If Meta monetizes just 20% of its data center capacity then it could generate $25B in profits and that optionality still doesn't look fully priced into the stock.
1:37

$META is already using AI to improve its ad engine which makes it a clear winner in the AI economy. The market still wants proof that its massive infrastructure spend can create revenue beyond better ads which is why the compute business matters. If Meta monetizes just 20% of its data center capacity then it could generate $25B in profits and that optionality still doesn't look fully priced into the stock.

Shay Boloor

85,425 просмотров • 1 месяц назад