Podcasts about compute

Solving Poker and Diplomacy, Debating RL+Reasoning with Ilya, what's *wrong* with the System 1/2 analogy, and where Test-Time Compute hits a wall Timestamps 00:00 Intro – Diplomacy, Cicero & World Championship 02:00 Reverse Centaur: How AI Improved Noam's Human Play 05:00 Turing Test Failures in Chat: Hallucinations & Steerability 07:30 Reasoning Models & Fast vs. Slow Thinking Paradigm 11:00 System 1 vs. System 2 in Visual Tasks (GeoGuessr, Tic-Tac-Toe) 14:00 The Deep Research Existence Proof for Unverifiable Domains 17:30 Harnesses, Tool Use, and Fragility in AI Agents 21:00 The Case Against Over-Reliance on Scaffolds and Routers 24:00 Reinforcement Fine-Tuning and Long-Term Model Adaptability 28:00 Ilya's Bet on Reasoning and the O-Series Breakthrough 34:00 Noam's Dev Stack: Codex, Windsurf & AGI Moments 38:00 Building Better AI Developers: Memory, Reuse, and PR Reviews 41:00 Multi-Agent Intelligence and the “AI Civilization” Hypothesis 44:30 Implicit World Models and Theory of Mind Through Scaling 48:00 Why Self-Play Breaks Down Beyond Go and Chess 54:00 Designing Better Benchmarks for Fuzzy Tasks 57:30 The Real Limits of Test-Time Compute: Cost vs. Time 1:00:30 Data Efficiency Gaps Between Humans and LLMs 1:03:00 Training Pipeline: Pretraining, Midtraining, Posttraining 1:05:00 Games as Research Proving Grounds: Poker, MTG, Stratego 1:10:00 Closing Thoughts – Five-Year View and Open Research Directions Chapters 00:00:00 Intro & Guest Welcome 00:00:33 Diplomacy AI & Cicero Insights 00:03:49 AI Safety, Language Models, and Steerability 00:05:23 O Series Models: Progress and Benchmarks 00:08:53 Reasoning Paradigm: Thinking Fast and Slow in AI 00:14:02 Design Questions: Harnesses, Tools, and Test Time Compute 00:20:32 Reinforcement Fine-tuning & Model Specialization 00:21:52 The Rise of Reasoning Models at OpenAI 00:29:33 Data Efficiency in Machine Learning 00:33:21 Coding & AI: Codex, Workflows, and Developer Experience 00:41:38 Multi-Agent AI: Collaboration, Competition, and Civilization 00:45:14 Poker, Diplomacy & Exploitative vs. Optimal AI Strategy 00:52:11 World Models, Multi-Agent Learning, and Self-Play 00:58:50 Generative Media: Image & Video Models 01:00:44 Robotics: Humanoids, Iteration Speed, and Embodiment 01:04:25 Rapid Fire: Research Practices, Benchmarks, and AI Progress 01:14:19 Games, Imperfect Information, and AI Research Directions

What do you do with the world's largest supercomputer? This week, Technology Now looks further at the world of supercomputers and explores what the world's largest supercomputer, El Capitan, and it's sister machine, Tuolumne, are used for. Rob Rieben, a computational physicist at Lawrence Livermore National Laboratory, tells us more.This is Technology Now, a weekly show from Hewlett Packard Enterprise. Every week, hosts Michael Bird and Aubrey Lovell look at a story that's been making headlines, take a look at the technology behind it, and explain why it matters to organizations and what can be learnt from it.About Rob: https://www.linkedin.com/in/rieben1/SourcesWhat are supercomputers used for:https://www.anl.gov/science-101/supercomputingToday I learned: https://www.england.nhs.uk/2025/05/nhs-first-in-world-to-roll-out-revolutionary-blood-test-for-cancer-patients/https://www.theguardian.com/society/2025/may/29/revolutionary-dna-blood-test-to-offer-thousands-in-england-tailored-cancer-careThis week in history:https://www.esa.int/About_Us/50_years_of_ESA/50_years_of_humans_in_space/First_woman_in_space_Valentina

Explore FSx for Lustre's new intelligent storage tiering that delivers cost savings and unlimited scalability for file storage in the cloud. Plus, discover how the new Model Context Protocol (MCP) servers are revolutionizing AI-assisted development across ECS, EKS, and serverless platforms with real-time contextual responses and automated resource management. 00:00 - Intro, 00:52 - Introduction new storage class, 03:43 - MCP Servers, 07:18 - Analytics, 09:34 - Application Integration, 15:52 - Business Applications, 16:21 - Cloud Financial Management, 17:44 - Compute, 20:44 - Containers, 21:31 - Databases, 24:25 - Developer Tools, 25:42 - End User Computing, 25:58 - Gaming, 26:34 - Management and Governance, 28:35 - Marketplace, 28:51 - Media Services, 29:29 - Migration and Transfer, 30:01 - Networking and Content Delivery, 34:01 - Security Identity and Compliance, 34:43 - Serverless, 35:06 - Storage, 36:55 - Wrap up Show Notes: https://dqkop6u6q45rj.cloudfront.net/shownotes-20250613-185437.html

Chris Lattner of Modular (https://modular.com) joined us (again!) to talk about how they are breaking the CUDA monopoly, what it took to match NVIDIA performance with AMD, and how they are building a company of "elite nerds". X: https://x.com/latentspacepod Substack: https://latent.space 00:00:00 Introductions 00:00:12 Overview of Modular and the Shape of Compute 00:02:27 Modular's R&D Phase 00:06:55 From CPU Optimization to GPU Support 00:11:14 MAX: Modular's Inference Framework 00:12:52 Mojo Programming Language 00:18:25 MAX Architecture: From Mojo to Cluster-Scale Inference 00:29:16 Open Source Contributions and Community Involvement 00:32:25 Modular's Differentiation from VLLM and SGLang 00:41:37 Modular's Business Model and Monetization Strategy 00:53:17 DeepSeek's Impact and Low-Level GPU Programming 01:00:00 Inference Time Compute and Reasoning Models 01:02:31 Personal Reflections on Leading Modular 01:08:27 Daily Routine and Time Management as a Founder 01:13:24 Using AI Coding Tools and Staying Current with Research 01:14:47 Personal Projects and Work-Life Balance 01:17:05 Hiring, Open Source, and Community Engagement

Carmen Li spent decades in financial services across trading floors and data companies before spotting a massive inefficiency in the AI/compute economy. After managing global data partnerships at Bloomberg, she witnessed AI startups struggling with unpredictable compute costs that could swing their margins from healthy profits to devastating losses overnight. Drawing parallels to how airlines hedge oil prices through futures markets, Carmen realized that compute—despite being one of the fastest-growing commodities—lacked basic risk management tools. Within months of leaving Bloomberg, she built Silicon Data into the world's first GPU compute risk management platform, raising $5.7M without ever creating a pitch deck and publishing the industry's first GPU compute index on Bloomberg Terminal. Topics Discussed: The systemic problem of compute cost volatility destroying AI company margins Why compute lacks the risk management tools available in every other commodity market Building the world's first GPU compute index and benchmarking service Raising venture capital without pitch decks through product-first demonstrations Operating as a solo non-technical founder leading a team of engineers The unique buyer dynamics when selling to CTOs, portfolio managers, and AI researchers simultaneously GTM Lessons For B2B Founders: Price on value, not cost, and let customer conversations reshape your understanding: Carmen admits that every client conversation changes her valuation of the product's impact, typically making it bigger than initially thought. She prices based on the value delivered rather than cost structure. B2B founders should remain flexible in their value proposition and pricing as they learn more about customer impact through direct engagement. Product demonstrations beat pitch decks for technical buyers: Carmen raised $5.7M without ever creating a pitch deck, instead letting prospects interact directly with her product and writing a simple memo. For technical products solving complex problems, demonstrating actual capabilities often proves more effective than polished presentations. B2B founders should prioritize building working products over perfecting sales materials. Embrace being the "dumbest person in the room" for learning velocity: Carmen describes consistently being the least technical person in rooms full of CTOs, AI researchers, and GPU experts, but leverages this as a learning advantage. She asks hard questions and co-creates products on the fly based on these conversations. B2B founders should view knowledge gaps as opportunities for rapid learning rather than weaknesses to hide. Target systemic problems that span multiple sophisticated buyer types: Silicon Data serves everyone from chip designers to hedge funds to AI companies, requiring Carmen to handle technical GPU questions, financial modeling queries, and AI workflow concerns in single meetings. This breadth creates natural expansion opportunities and defensibility. B2B founders should look for problems that affect multiple stakeholder types within their target market. Leverage unique background intersections to spot obvious-but-overlooked opportunities: Carmen's combination of financial services expertise and data company experience let her quickly identify that compute needed the same risk management tools available in every other commodity market. The solution was "extremely intuitive" to her but invisible to others. B2B founders should examine how their unique background combinations reveal opportunities others might miss.   //   Sponsors: Front Lines — We help B2B tech companies launch, manage, and grow podcasts that drive demand, awareness, and thought leadership. www.FrontLines.io The Global Talent Co. — We help tech startups find, vet, hire, pay, and retain amazing marketing talent that costs 50-70% less than the US & Europe.  www.GlobalTalent.co   //   Don't Miss: New Podcast Series — How I Hire Senior GTM leaders share the tactical hiring frameworks they use to build winning revenue teams. Hosted by Andy Mowat, who scaled 4 unicorns from $10M to $100M+ ARR and launched Whispered to help executives find their next role. Subscribe here: https://open.spotify.com/show/53yCHlPfLSMFimtv0riPyM   

How do you make the world's fastest supercomputer? This week, Technology Now dives into the world of supercomputers, and how El Capitan, the world's largest supercomputer, was built. We will explore the software and hardware requirements as well as investigating the physical requirements needed to even be able to run a supercomputer on your premises. Bronis de Supinski, CTO of Livermore Computing at Lawrence Livermore National Laboratory, tells us more.This is Technology Now, a weekly show from Hewlett Packard Enterprise. Every week, hosts Michael Bird and Aubrey Lovell look at a story that's been making headlines, take a look at the technology behind it, and explain why it matters to organizations and what can be learnt from it.About Bronis: https://www.linkedin.com/in/bronis-de-supinski-607a441/SourcesEl Capitanhttps://www.hpe.com/us/en/newsroom/press-release/2024/11/hewlett-packard-enterprise-delivers-worlds-fastest-direct-liquid-cooled-exascale-supercomputer-el-capitan-for-lawrence-livermore-national-laboratory.htmlWhat are FLOPShttps://www.techtarget.com/whatis/definition/FLOPS-floating-point-operations-per-secondToday I LearnedMa. Y., et all, 2025, Near-infrared spatiotemporal colour vision in humans enabled by upconversion contact lenses, ISSN 0092-8674, 10.1016/j.cell.2025.04.019 https://www.cell.com/cell/fulltext/S0092-8674(25)00454-4This Week in Historyhttps://www.bbc.co.uk/future/article/20201028-history-of-the-ballpoint-penhttps://spinoff.nasa.gov/space-pens

Forget everything you think you know about AI timelines. The world's top AI labs now say AGI—artificial intelligence that surpasses humans at all tasks—is 5 years away, not 50. ChatGPT's 800 million users was just the beginning. When AI can teach better than any professor, diagnose better than any doctor, and code better than any developer, what happens to human work? To education? To you? This isn't futurism—it's a timeline shock that's sending universities into crisis mode. The race to AGI is on, and if you're not preparing now, you're already behind.Opening: From Basketball to AGI's Moon Shot (00:00:00)The Knicks' 25-year drought as metaphor for timeline compressionAGI as "today's moon shot" with labs racing at full throttleThe shift from "will AI get smart?" to "AI is already outthinking us"The Context: Why This Matters Today (00:04:21)ChatGPT's explosive growth: 800 million users in under 3 years—history's fastest tech adoptionCurrent AI still "narrow"—brilliant but limited like a writer who can't do mathSam Altman's bombshell: "We now know how to build AGI"Why education leaders need to understand this shift immediatelyWhat Is AGI and How Is It Different? (00:11:33)The specialist doctor vs. Leonardo da Vinci comparisonAGI capabilities: learns without training, transfers knowledge, solves new problemsOpenAI's definition: outperforms humans at most economically valuable workGoogle DeepMind's bar: matching Einstein's scientific breakthroughsMachines shifting from assistants to peers—or superiorsThe Timeline Shock: Why Now? (00:20:54)Sam Altman: AGI by 2029 (current presidential term)Dario Amodei: AI outsmarts humans by 20262,778 AI researchers: median prediction now 2047, down 13 years in one year10% chance of AGI by 2027—"Would you board that plane?"Compute growing 10x yearly, costs down 99.7%, $212 billion invested in 2024What This Means for the Knowledge Economy (00:27:23)80% of US workforce faces task disruption—all wage and education levelsNot just blue-collar: lawyers, doctors, engineers, executives all impactedMcKinsey already cutting 10% workforce, replacing with AIThe paradox: GDP could 10x while jobs disappearGitHub Copilot at 77,000 organizations, growing 180% yearlyThe Education Paradox (00:33:02)Universities' three pillars: knowledge transfer, certification, communityAGI breaks the first two completelyWhen AI teaches perfectly for free, "what's a lecture for?"When AI outperforms any graduate, "what's a diploma for?"Chegg's stock crash, coding bootcamps struggling, MBA programs questioning valueThree Futures for Universities (00:40:06)Human-Edge College: Small seminars, mentorship, emotional intelligence, ethicsResearch Steward: Ethical guardians of AGI, independent from profit motivesLifelong Learning Platform: Continuous upskilling partner using AI at scaleThe shift from information delivery to human developmentYear One Projection: The AI Teammate Era (00:48:29)AI agents as actual teammates handling meetings, reports, decisionsStudents using AI regardless of policies—adaptation isn't optionalSchools must decide how to integrate, not whetherThe institutions experimenting now vs. those falling behindYear Five Projection: The Hybrid Intelligence Era (00:50:01)2030: Seamless human-AI collaboration across all knowledge workNew jobs emerge: AI psychologists, algorithm auditors, team coordinatorsEvery student with 24/7 personalized AI tutorProfessors transform into coaches, degrees become competency-basedYear Ten Projection: Abundance or Inequality (00:52:51)Best case: Monthly scientific breakthroughs, disease prevention, climate solutionsWorst case: Mass unemployment, extreme inequality, human purpose crisisLarry Ellison's $600B Stargate project for personalized medicineEducation's new role: helping humans find meaning when machines do the workThe Call to Action: Lead or Be Led (00:59:37)Change is coming faster than anyone thinks—decades are now yearsStart experimenting immediately—not next semester, nowFocus on uniquely human qualities: empathy, creativity, ethical reasoningUniversities must become thought leaders, not bystanders"The future isn't fixed—we decide if AGI amplifies or replaces us" - - - -Connect With Our Co-Hosts:Ardis Kadiuhttps://www.linkedin.com/in/ardis/https://twitter.com/ardisDr. JC Bonillahttps://www.linkedin.com/in/jcbonilla/https://twitter.com/jbonillxAbout The Enrollify Podcast Network:Generation AI is a part of the Enrollify Podcast Network. If you like this podcast, chances are you'll like other Enrollify shows too! Enrollify is made possible by Element451 — the next-generation AI student engagement platform helping institutions create meaningful and personalized interactions with students. Learn more at element451.com. Attend the 2025 Engage Summit! The Engage Summit is the premier conference for forward-thinking leaders and practitioners dedicated to exploring the transformative power of AI in education. Explore the strategies and tools to step into the next generation of student engagement, supercharged by AI. You'll leave ready to deliver the most personalized digital engagement experience every step of the way.Register now to secure your spot in Charlotte, NC, on June 24-25, 2025! Early bird registration ends February 1st -- https://engage.element451.com/register

Audio note: this article contains 127 uses of latex notation, so the narration may be difficult to follow. There's a link to the original text in the episode description. Confidence: Medium, underlying data is patchy and relies on a good amount of guesswork, data work involved a fair amount of vibecoding. Intro:  Tom Davidson has an excellent post explaining the compute bottleneck objection to the software-only intelligence explosion.[1] The rough idea is that AI research requires two inputs: cognitive labor and research compute. If these two inputs are gross complements, then even if there is recursive self-improvement in the amount of cognitive labor directed towards AI research, this process will fizzle as you get bottlenecked by the amount of research compute. The compute bottleneck objection to the software-only intelligence explosion crucially relies on compute and cognitive labor being gross complements; however, this fact is not [...] ---Outline:(00:35) Intro:(02:16) Model(02:19) Baseline CES in Compute(04:07) Conditions for a Software-Only Intelligence Explosion(07:39) Deriving the Estimation Equation(09:31) Alternative CES Formulation in Frontier Experiments(10:59) Estimation(11:02) Data(15:02) Trends(15:58) Estimation Results(18:52) ResultsThe original text contained 13 footnotes which were omitted from this narration. --- First published: June 1st, 2025 Source: https://forum.effectivealtruism.org/posts/xoX936hEvpxToeuLw/estimating-the-substitutability-between-compute-and --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

In this episode of We Are DePIN, we sit down with Daniel Keller, Co-Founder of Flux, to explore how they're building a decentralised, scalable, and censorship-resistant alternative to Big Tech cloud providers.Daniel breaks down:The origin story and mission behind FluxWhy centralised cloud is a threat to Web3How Flux is enabling DePIN projects, dApps, and node infrastructureThe role of FluxNodes and $FLUX in powering the networkHis vision for a fully trustless, user-owned cloudIf you're interested in the future of decentralised infrastructure, edge computing, or how Flux is positioning itself as the backbone of Web3, this conversation is essential listening. Hosted on Acast. See acast.com/privacy for more information.

In Episode 52 of the Nextflow podcast, Phil Ewels is joined by Ziad Al Bkhetan and Steven Manos from the Australian BioCommons to explore how Nextflow and Seqera Platform are transforming bioinformatics research across Australia.Discover how the Australian BioCommons emerged as a national infrastructure investment to support molecular life sciences researchers, building digital platforms and tools that serve diverse research communities - from genome assembly to proteomics. Learn about their journey from identifying Nextflow as the leading workflow management system in Australia to successfully deploying Seqera Platform for over 90 users across 18+ research institutions.The conversation covers fascinating success stories including the collaborative development of the nf-core/proteinfold pipeline, supporting structural biology researchers with advanced protein folding tools, and the innovative Ozark community phylogenetic trees workflow. Steven and Ziad share insights on overcoming challenges like integrating web platforms with traditional HPC infrastructure at Australia's tier-one supercomputing facilities (NCI and Pawsey), and building hybrid compute environments that seamlessly span local HPC and cloud resources.Key highlights include Australia's growing contribution to the global nf-core community, culminating in their first official participation in the March 2025 nf-core hackathon with a dedicated Sydney hub. The team also discusses practical advice for other countries looking to establish national Nextflow services, emphasizing the importance of building local expertise and community engagement.For Australian researchers interested in accessing these services, visit biocommons.org.au to learn more about available platforms, training opportunities, and community resources.00:00 Podcast Ep 52: BioCommons00:06 Welcome00:41 Ziad - introduction01:36 Steven - introduction03:40 Introduction to Australian BioCommons08:13 Nextflow usage in Australia09:05 BioCommons collaboration with Seqera12:24 Compute infrastructures in Australia15:05 Nextflow support & training16:43 Highlights and lowlights19:11 Protein fold21:50 Ozard Community23:39 Automation with Seqera Platform25:52 Nextflow Communty in Australia29:27 2025 nf-core hackathon31:57 Future plans for Nextflow & Seqera Platform33:04 Tips for building a National Nextflow platform36:21 Next steps37:51 Wrap upResources:Australian BioCommons: https://biocommons.org.aunf-core office hours: https://nf-co.re/blog/2025/apac-helpdesk-2025

Claude 4 Opus and Sonnet, the most powerful models from Anthropic for coding and advanced reasoning, are now available on Amazon Bedrock. Plus, AWS Transform can now accelerate your Mainframe, .NET and VMWare workload modernization. Learn about these updates and more with your hosts Shruthi and Jillian. 00:00 - Intro, 00:21 - Amazon Bedrock, 03:40 - AWS Transform, 07:51 - EC2 P6/B200 Instances, 11:35 - Analytics, 16:55 - Business Applications, 17:27 - Cloud Financial Management, 17:46 - Compute, 20:14 - Containers, 21:27 - ECS, 21:54 - Databases, 27:08 - Developer Tools, 29:15 - End User Computing, 29:53 - Management & Governance, 32:36 - Migration & Transfer, 35:12 - Networking and Content Delivery, 35:35 - Security Entity and Compliance, 36:07 - Service Changes, 37:06 - Services end of support, 38:26 - Wrap up Show Notes: https://dqkop6u6q45rj.cloudfront.net/shownotes-20250530-194643.html

Interview with James Nagle, Reboot of Compute's Gazette Magazine Patreon: https://www.patreon.com/FloppyDays Sponsors: 8-Bit Classics  Arcade Shopper  FutureVision Research   Hello, and welcome to episode 151 of the Floppy Days Podcast for May, 2025.  My name is Randy Kindig and I'm the host for this ode to computers that only survive in our memories and our collections. This month, I'm continuing to step aside from the ongoing series of episodes about the HP 97/67 programmable calculators to bring you a timely interview that basically constitutes current news.  I don't often do this, but this news was so exciting to me that I wanted to bring this to all my listeners as soon as possible.  The interviewee is James Nagle and the topic is the sudden and welcome news that James is planning to revive the iconic Compute! magazine under the equally-iconic name Compute's Gazette.  I hope you'll stick around to hear about James' plans and are as excited as I am to find out where this goes. For upcoming shows, we do have one more episode in the series on the HP97 with HP calculator historian Wlodek Mier-Jedrzejowicz (“Vwahdek Meer-Yeng SHAY of itch”).  I will air that episode very soon. New Acquisitions and What I've Been Up To C64OS - https://c64os.com/  PiStorm: https://www.hackster.io/news/hands-on-with-the-pistorm-the-ultimate-raspberry-pi-powered-accelerator-for-your-commodore-amiga-449ef0634f3e  https://www.amigastore.com/pistorm-edition-amiga-p-91328.html  Quick Reference Book - https://floppydaysqr.my.canva.site/  Upcoming Vintage Computer Shows Retrofest 2025 - May 31-June1 - Steam Museum of the Great Western Railway, Swindon, UK - https://retrofest.uk/  Vancouver Retro Gaming Expo - June 14 - New Westminster, BC, Canada - https://www.vancouvergamingexpo.com/index.html  VCF Southwest - June 20-22, 2025 - Davidson-Gundy Alumni Center at UT Dallas - https://www.vcfsw.org/  Southern Fried Gaming Expo and VCF Southeast - June 20-22, 2025 - Atlanta, GA - https://gameatl.com/  Pacific Commodore Expo NW v4 - June 21-22 - Old Rainier Brewery Intraspace, Seattle, WA - https://www.portcommodore.com/dokuwiki/doku.php?id=pacommex:start  KansasFest - July 18-20 - Virtual only - https://www.kansasfest.org/  INIT HELLO Apple II Conference - July 26-27 - System Source Computer Museum in Hunt Valley, MD - https://init-hello.org/  Silly Venture SE (Summer Edition) - July 31-Aug. 3 - Gdansk, Poland - https://www.demoparty.net/silly-venture/silly-venture-2025-se  VCF West - August 1-2 - Computer History Museum in Mountain View, CA - https://vcfed.org/2025/03/05/vcf-west-2025-save-the-date/  Fujiama - August 11-17 - Lengenfeld, Germany - http://atarixle.ddns.net/fuji/2025/  VCF Midwest - September 13-14, 2025 - Renaissance Schaumburg Convention Center in Schaumburg, IL - http://vcfmw.org/  Tandy Assembly - September 26-28 - Courtyard by Marriott Springfield - Springfield, OH - http://www.tandyassembly.com/  Portland Retro Gaming Expo - October 17-19 - Oregon Convention Center, Portland, OR - https://retrogamingexpo.com/  Chicago TI International World Faire - October 25 - Evanston Public Library, Evanston, IL - https://www.chicagotiug.org/home  Schedule Published on Floppy Days Website - https://docs.google.com/document/d/e/2PACX-1vSeLsg4hf5KZKtpxwUQgacCIsqeIdQeZniq3yE881wOCCYskpLVs5OO1PZLqRRF2t5fUUiaKByqQrgA/pub  Feedback HP-97S: https://www.hpmuseum.org/hp97s.htm https://www.johnwolff.id.au/calculators/laboratory/laboratory.htm Interview Links New Compute's Gazette Website - https://www.computesgazette.com/ Compute's Gazette collection at archive.org - https://archive.org/details/computes.gazette

Leo Fan is Co-Founder of Cysic, Assistant Professor of Computer Science at Rutgers University, and a former core researcher at Algorand. He leads the development of ComputeFi, Cysic's hardware tokenisation model that turns real-world compute into on-chain assets — part of Cysic's broader mission to build low-latency, cost-efficient infrastructure for real-time Web3, AI, and decentralised applications.

Leo Fan is Co-Founder of Cysic, Assistant Professor of Computer Science at Rutgers University, and a former core researcher at Algorand. He leads the development of ComputeFi, Cysic's hardware tokenisation model that turns real-world compute into on-chain assets — part of Cysic's broader mission to build low-latency, cost-efficient infrastructure for real-time Web3, AI, and decentralised applications.

The Vault is a morning show hosted on Twitter Spaces and YouTube Live on Tuesdays, Wednesdays, and Thursdays at 11:30 am EST. The show focuses on multi-chain communities, emerging protocols, NFTFi, DeFi, Gaming, and, most importantly, collecting digital assets.Adam McBride: https://twitter.com/adamamcbrideJake Gallen: https://twitter.com/jakegallen_Chris Devitte: https://twitter.com/chris_devvEmblem Vault: https://twitter.com/EmblemVault

What if your rooftop solar could do more than just power your fridge?Karl Andersen believes it can—and should—power the future of AI.Karl unpacks the grid's biggest vulnerabilities, why data centers lack critical power infrastructure, and how we can turn solar-powered homes into the building blocks of a decentralized compute network. Lektra's tech fuses distributed energy with cloud computing—think “Raspberry Pi meets Tesla Powerwall meets AI.” The result? A game-changing business model where solar homeowners become micro data centers—and start earning like one.From national security concerns to GPU monetization, Karl walks us through why our energy and data systems are broken—and how his patented solution bridges both.Expect to learn:

Nikolay Filichkin is the Co-Founder and Chief Business Officer of Compute Labs, a cutting-edge infrastructure company transforming access to compute by financializing GPU assets. With over a decade of experience in technology, M&A, and strategic partnerships, Nikolay plays a key role in creating the financial ecosystem for compute as an asset class.Prior to Compute Labs, Nikolay spent 6+ years at Xsolla, where he led strategic partnerships and helped scale the company from 200 to over 1,200 employees. He drove five acquisitions, launched three new product lines, and helped expand Xsolla's global presence across gaming and digital commerce.Nikolay is a builder and operator who understands how to take ideas from 0 to 1 and then scale them. At Compute Labs, he leads business development and investor relations, helping institutional allocators and partners gain access to the financial upside of AI infrastructure during this key period of growth.

Nikolay Filichkin is the Co-Founder and Chief Business Officer of Compute Labs, a cutting-edge infrastructure company transforming access to compute by financializing GPU assets. With over a decade of experience in technology, M&A, and strategic partnerships, Nikolay plays a key role in creating the financial ecosystem for compute as an asset class.Prior to Compute Labs, Nikolay spent 6+ years at Xsolla, where he led strategic partnerships and helped scale the company from 200 to over 1,200 employees. He drove five acquisitions, launched three new product lines, and helped expand Xsolla's global presence across gaming and digital commerce.Nikolay is a builder and operator who understands how to take ideas from 0 to 1 and then scale them. At Compute Labs, he leads business development and investor relations, helping institutional allocators and partners gain access to the financial upside of AI infrastructure during this key period of growth.

Nova Premier is our most advanced AI model yet, featuring a million-token context window and enhanced capabilities at nearly half the cost of competitors. Dive into this update and more with hosts Simon and Jillian. 00:00 - Intro 00:31 - Amazon Nova Premier 02:56 - Analytics 04:46 - Artificial Intelligence 11:02 - Business Applications 11:38 - Cloud Financial Management 11:57 - Compute 12:10 - Contact Center 14:50 - Containers 15:13 - Database 17:52 -Developer Tools 18:08 - Management and Governance 20:25 - Networking 22:48 - Marketplace 24:04 - Security Identity End Compliance 26:09 - Storage 27:56 - Outro Show Notes: https://dqkop6u6q45rj.cloudfront.net/shownotes-20250516-191312.html

A large contingent of Silicon Valley CEOs followed President Donald Trump to Saudi Arabia this week, where a number of them announced billions of dollars in AI-related investments and business partnerships. Mohammed Soliman, a senior fellow at the Middle East Institute, says this is the new Middle East — where the relationship with the U.S. is driven by tech and innovation, not just oil and security. On POLITICO Tech, Soliman tells host Steven Overly how this new arrangement benefits tech companies and Gulf nations — and why it's necessary if the U.S. hopes to stay ahead of China.  Steven Overly is the host of POLITICO Tech and covers the intersection of trade and technology. Nirmal Mulaikal is the co-host and producer of POLITICO Energy and producer of POLITICO Tech. Learn more about your ad choices. Visit megaphone.fm/adchoices

Piotr Tomasik, Co-Founder & President of TensorWave, who’s powering the next wave of AI compute with AMD-optimized super-cloud infrastructure and building Las Vegas into a … Read more The post Powering the AI Revolution: How TensorWave’s AMD Supercloud Is Solving Compute Bottlenecks appeared first on Top Entrepreneurs Podcast | Enterprise Podcast Network.

Lennart Heim, a researcher and information scientist at RAND Corporation, joins Azeem Azhar to unpack a provocative claim: China is catching up with US AI capabilities, but it doesn't matter. Timestamps: (00:00) Episode trailer (01:19) Lennart's core thesis (03:26)   Why compute matters so much (07:31)  The investment split between model R&D and model execution (11:18)  How test-time compute impacts costs (16:14) The geopolitics of compute (21:32) Why does the U.S have more compute capacity than China? (25:01)  The trade-off between economic needs and national-security needs (31:54)  How technology change might shift the battlegrounds (35:33)  Dealing with compute and power concentration (48:19)  Concluding quick-fire question  Lennart's links: Twitter/X: https://twitter.com/ohlennartPersonal blog: https://heim.xyz/Azeem's links:Substack: https://www.exponentialview.co/Website: https://www.azeemazhar.com/LinkedIn: https://www.linkedin.com/in/azharTwitter/X: https://x.com/azeemThis was originally recorded for "Friday with Azeem Azhar", a new show that takes place every Friday at 9am PT and 12pm ET. You can tune in through Exponential View on Substack. Produced by supermix.io and EPIIPLUS1 Ltd

Hey there, listeners! Welcome back to the show! Today, we've got an incredible episode lined up for you. Host Jake Aaron Villarreal sits down with Gennady Pihimenko, the co-founder and CEO of CentML, for a deep dive into the wild world of AI and machine learning. Gennady takes us on his fascinating journey—from a math and programming whiz in Russia to a trailblazer in the AI industry. He breaks down the evolution of AI, spills the tea on why optimizing machine learning workloads is a game-changer, and gets real about the balancing act of being a professor and a startup CEO.Gennady doesn't hold back, sharing sharp insights on the hidden costs of AI—think training and inference—and who CentML's solutions are built for. He's all about solving real-world problems and unlocking transformative impact through smarter, more efficient compute solutions. Plus, he dishes on the challenges and opportunities in the AI industry, from the skyrocketing demand for computational power to the art of building a killer team and nailing hiring strategies in a startup. Gennady also gives us a sneak peek into CentML's future, teasing their growth plans and how customer feedback is shaping their roadmap.Buckle up for a conversation packed with big ideas, practical wisdom, and a front-row seat to the future of AI. Let's dive in! Host: Jake Aaron Villarreal, leads the top AI Recruitment Firm in Silicon Valley www.matchrelevant.com, uncovering stories of funded startups and goes behinds to scenes to tell their founders journey.  If you are growing AI Startup or have a great storytelling, email us at: jake.villarreal@matchrelevant.com  

Carl Peterson, CEO of Thunder Compute uncovers how Thunder Computer is redefining GPU utilization by enabling network-attached virtual GPUs—dramatically slashing costs and democratizing access. Carl shares the startup's Y Combinator origin story, the impact of DeepSeek, and how virtualization is transforming AI development for individuals and enterprises alike. We also unpack GPU security, job disruption from AI, and the accelerating arms race in model development. A must-listen for anyone navigating AI, compute efficiency, and data protection.

Description: Learn how you can use the all new Amazon Q Developer integration with GitLab Duo to automate code generation and review, plus even more updates from AWS. 00:00:00 - Intro, 00:00:28 - SWE Holly Bench, 00:04:31 - Analytics, 00:06:49 - Application Integration, 00:07:14 - Artificial Intelligence, 00:08:53 - Amazon Bedrock Data Automation, 00:14:11 - AWS Health Omex, 00:14:21 - Compute, 00:16:37 - Contact Centers, 00:17:25 - Containers, 00:17:46 - Databases, 00:18:18 - Front end Web and Mobile, 00:18:59 - Management and Governance, 00:20:07 - Migration and Transfer, 00:20:17 - Networking and Content Delivery, 00:20:44 - Security Identity End Compliance, 00:23:24 - Serverless, 00:24:01 - Storage, 00:24:41 - Wrap up Shownotes: https://d29iemol7wxagg.cloudfront.net/719ExtendedShownotes.html

What if your LLM could think ahead—preparing answers before questions are even asked?In this week's paper read, we dive into a groundbreaking new paper from researchers at Letta, introducing sleep-time compute: a novel technique that lets models do their heavy lifting offline, well before the user query arrives. By predicting likely questions and precomputing key reasoning steps, sleep-time compute dramatically reduces test-time latency and cost—without sacrificing performance.​We explore new benchmarks—Stateful GSM-Symbolic, Stateful AIME, and the multi-query extension of GSM—that show up to 5x lower compute at inference, 2.5x lower cost per query, and up to 18% higher accuracy when scaled.​You'll also see how this method applies to realistic agent use cases and what makes it most effective.If you care about LLM efficiency, scalability, or cutting-edge research.Explore more AI research, or sign up to hear the next session live: arize.com/ai-research-papersLearn more about AI observability and evaluation, join the Arize AI Slack community or get the latest on LinkedIn and X.

Interview with David Greelish, Apple Lisa Documentary Patreon: https://www.patreon.com/FloppyDays Sponsors: 8-Bit Classics  Arcade Shopper  FutureVision Research   Hello, and welcome to episode 150 of the Floppy Days Podcast for April, 2025.  My name is Randy Kindig and I'm the host for this journey through the annals of home computer history. This month, I'm going to step aside from the ongoing series of episodes about the HP 97/67 programmable calculators to bring you a timely interview with a good friend about an interesting topic.  That friend is David Greelish, a computer historian, and the topic is his recent publication of a film documentary about the Apple Lisa, called "Before Macintosh: The Apple Lisa".  David tells us all about the film, why he produced it, why the Apple Lisa was an important part of home computer history, who he interviewed for the film (he had some amazing guests) and much more.  It's a great film and should interest a lot of the listeners, so please consider going out and purchasing the film in order to support David's efforts. For upcoming shows, we do have one more episode in the series on the HP97 with HP calculator historian Wlodek Mier-Jedrzejowicz.  I will air that episode very soon. New Acquisitions/What I've Been Up To Indy Classic Expo - https://www.indyclassic.org  Vintage Computer Center - https://www.vintagecomputercenter.com  OmniView 80 card for Atari 800 - https://archive.org/details/Atari_OMNIVIEW_manual  Commodore 16 - https://en.wikipedia.org/wiki/Commodore_16  6502 Plus 4 upgrade for C16 from Lotharek - (https://lotharek.pl/productdetail.php?id=257  News Reboot of Compute's Gazette Magazine - https://www.computesgazette.com/iconic-computes-gazette-magazine-returns-after-35-years-expanding-focus-to-entire-retro-computing-community/  Upcoming Shows The 32nd Annual “Last” Chicago CoCoFEST! - May 2-3, 2025 - Holiday Inn & Suites Chicago-Carol Stream (Wheaton), Carol Stream, Illinois - https://www.glensideccc.com/cocofest/  VCF Europe - May 3-4 - Munich, Germany - https://vcfe.org/E/  Retrofest 2025 - May 31-June1 - Steam Museum of the Great Western Railway, Swindon, UK - https://retrofest.uk/  Vancouver Retro Gaming Expo - June 14 - New Westminster, BC, Canada - https://www.vancouvergamingexpo.com/index.html  VCF Southwest - June 20-22, 2025 - Davidson-Gundy Alumni Center at UT Dallas - https://www.vcfsw.org/  Southern Fried Gaming Expo and VCF Southeast - June 20-22, 2025 - Atlanta, GA - https://gameatl.com/  Pacific Commodore Expo NW v4 - June 21-22 - Old Rainier Brewery Intraspace, Seattle, WA - https://www.portcommodore.com/dokuwiki/doku.php?id=pacommex:start  KansasFest - July 18-20 - Virtual only - https://www.kansasfest.org/  VCF West - August 1-2 - Computer History Museum in Mountain View, CA - https://vcfed.org/2025/03/05/vcf-west-2025-save-the-date/  VCF Midwest - September 13-14, 2025 - Renaissance Schaumburg Convention Center in Schaumburg, IL - http://vcfmw.org/  Tandy Assembly - September 26-28 - Courtyard by Marriott Springfield - Springfield, OH - http://www.tandyassembly.com/  Portland Retro Gaming Expo - October 17-19 - Oregon Convention Center, Portland, OR - https://retrogamingexpo.com/  Chicago TI International World Faire - October 25 - Evanston Public Library, Evanston, IL - https://www.chicagotiug.org/home  Schedule Published on Floppy Days Website - https://docs.google.com/document/d/e/2PACX-1vSeLsg4hf5KZKtpxwUQgacCIsqeIdQeZniq3yE881wOCCYskpLVs5OO1PZLqRRF2t5fUUiaKByqQrgA/pub  Documentary and Classic Computing Links Classic Computing Website - https://www.classiccomputing.com/Classic_Computing/Blog/Blog.html  https://www.youtube.com/watch?v=psAeTDYezdo - "Before Macintosh: The Apple Lisa" Full Documentary Film  Exidy Sorcerer at VCFSE 2 - https://floppydays.libsyn.com/floppy-days-episode-17-the-exidy-sorcerer-live-from-vcfse-20  Stan Veit podcast - https://www.classiccomputing.com/CCPodcasts/Stan_Veit/Stan_Veit.html  Classic Computing - the book! - https://www.classiccomputing.com/Classic_Computing/My_Book.html  Documentary link at IMDB - https://www.imdb.com/title/tt31122934/   

Daniel ‌Marin is the Founder and Chief Executive Officer of Nexus. Daniel founded Nexus in 2022 while he was at Stanford with the mission to enable the Verifiable Internet, which will redefine digital trust and create a more transparent, secure, and efficient world. To achieve this mission, Nexus is building a globally distributed Layer-1 blockchain powered by a zkVM engine.Daniel earned a Bachelor of Science in Computer Science from Stanford University. He was named to Forbes' '30 Under 30' list in 2025, and earned Bronze medals at the International Physics Olympiad in 2018 and 2019.In this conversation, we discuss:- Are we back?- Enabling the Verifiable Internet- Parallels between AI and ZK- Aggregating unused compute power- Verifiable AI- Solving critical issues around privacy, trust, and security- 2.1 million users and 3.6 million nodes already connected to the network- With Nexus, more nodes = faster blockchain- Verifiable computation will impact many markets, blockchain is just one example- The power of zkEVM- The future of AI & BlockchainNexusWebsite: nexus.xyzX: @NexusLabsDiscord: discord.gg/nexus-xyzDaniel MarinX: @danielmarinqLinkedIn: Daniel Marin---------------------------------------------------------------------------------  This episode is brought to you by PrimeXBT.  PrimeXBT offers a robust trading system for both beginners and professional traders that demand highly reliable market data and performance. Traders of all experience levels can easily design and customize layouts and widgets to best fit their trading style. PrimeXBT is always offering innovative products and professional trading conditions to all customers.   PrimeXBT is running an exclusive promotion for listeners of the podcast. After making your first deposit, 50% of that first deposit will be credited to your account as a bonus that can be used as additional collateral to open positions.  Code: CRYPTONEWS50  This promotion is available for a month after activation. Click the link below:  PrimeXBT x CRYPTONEWS50

Last year during my trip to Silicon Valley, I was invited to visit a company called PsiQuantum. When you think about quantum computing, your mind might conjure up those chandeliers. Qubits plunged to super cold temperatures. PsiQuantum is working on something a little different. Quantum computing using photons. In this video, a form of quantum compute with intriguing possibilities. Does it “work” like silicon does today? Is quantum compute really here? I can't really answer those questions in this video. But we can explore the ideas and the ideas are certainly mind-bending.

In this episode, David Aronchick, CEO and Co-founder of Expanso discusses his experiences and insights from working with Kubernetes since its early days at Google. David shares his journey from working on Kubernetes to co-founding Kubeflow and his latest project, Bacalhau, which focuses on combining compute and data management in distributed systems. Highlighting the challenges of data processing and privacy, particularly in edge computing and regulated environments, David emphasizes cost-saving benefits and the importance of local data processing. Throughout, privacy and regulatory concerns are underscored along with solutions for efficient and secure data handling. 00:00 Introduction and Welcome 00:23 Early Days of Kubernetes 01:05 Kubernetes Community and Evolution 02:23 AI, ML, and KubeFlow 03:40 Current Work and Data Challenges 08:20 Privacy and Security Concerns 14:21 Real-World Applications and Benefits 20:42 Conclusion Guest: David Aronchick, Founder and CEO at Expanso, formerly led open source machine learning strategy at Azure, managed Kubernetes product development at Google, and co-founded Kubeflow. Previous roles at Microsoft, Amazon, and Chef.  

Last year during my trip to Silicon Valley, I was invited to visit a company called PsiQuantum. When you think about quantum computing, your mind might conjure up those chandeliers. Qubits plunged to super cold temperatures. PsiQuantum is working on something a little different. Quantum computing using photons. In this video, a form of quantum compute with intriguing possibilities. Does it “work” like silicon does today? Is quantum compute really here? I can't really answer those questions in this video. But we can explore the ideas and the ideas are certainly mind-bending.

Learn about the latest new FM in the Nova family that simplifies conversational AI with low latency, and build safely with new capabilities for Amazon Bedrock Guardrails. 00:00 - Intro, 00:27 - Amazon Nova Sonic, 03:13 - Amazon Bedrock Guardrails, 05:23 - Analytics, 08:18 - Application Integration, 08:37 - Artificial Intelligence, 12:06 - Business Applications, 13:01 - Cloud Financial Management, 13:44 - Compute, 15:04 - Contact Center, 16:29 - Containers, 16:49 - Databases, 19:57 - Developer Tools, 20:59 - Frontend Web and Mobile, 21:20 - Management and Governance, 23:39 - Media Services, 25:37 - Migration and Transfer, 26:46 - Networking and Content Delivery, 28:45 - Artificial Intelligence, 29:58 - Security, Identity, and Compliance, 32:51 - Serverless, 33:57 - Storage, 37:29 - Wrap up Show Notes: https://dqkop6u6q45rj.cloudfront.net/run-sheet-20250418-173723.html

This episode is sponsored by Oracle. OCI is the next-generation cloud designed for every workload – where you can run any application, including any AI projects, faster and more securely for less.  On average, OCI costs 50% less for compute, 70% less for storage, and 80% less for networking. Join Modal, Skydance Animation, and today's innovative AI tech companies who upgraded to OCI…and saved.    Offer only for new US customers with a minimum financial commitment. See if you qualify for half off at http://oracle.com/eyeonai     In this episode of Eye on AI, Craig Smith sits down with Brice Challamel, Head of AI Products and Innovation at Moderna, to explore how one of the world's leading biotech companies is embedding artificial intelligence across every layer of its business—from drug discovery to regulatory approval.   Brice breaks down how Moderna treats AI not just as a tool, but as a utility—much like electricity or the internet—designed to empower every employee and drive innovation at scale. With over 1,800 GPTs in production and thousands of AI solutions running on internal platforms like Compute and MChat, Moderna is redefining what it means to be an AI-native company.   Key topics covered in this episode: How Moderna operationalizes AI at scale GenAI as the new interface for machine learning AI's role in speeding up drug approvals and clinical trials The future of personalized cancer treatment (INT) Moderna's platform mindset: AI + mRNA = next-gen medicine Collaborating with the FDA using AI-powered systems   Don't forget to like, comment, and subscribe for more interviews at the intersection of AI and innovation.     Stay Updated: Craig Smith on X:https://x.com/craigss Eye on A.I. on X: https://x.com/EyeOn_AI     (00:00) Preview  (02:49) Brice Challamel's Background and Role at Moderna (05:51) Why AI Is Treated as a Utility at Moderna (09:01) Moderna's AI Infrastructure (11:53) GenAI vs Traditional ML (14:59) Combining mRNA and AI as Dual Platforms (18:15) AI's Impact on Regulatory & Clinical Acceleration (23:46) The Five Core Applications of AI at Moderna (26:33) How Teams Identify AI Use Cases Across the Business (29:01) Collaborating with the FDA Using AI Tools (33:55) How Moderna Is Personalizing Cancer Treatments (36:59) The Role of GenAI in Medical Care (40:10) Producing Personalized mRNA Medicines (42:33) Why Moderna Doesn't Sell AI Tools (45:30) The Future: AI and Democratized Biotech

In this episode, Patrick McKenzie (@patio11) is joined by Tim Fist, Director of Emerging Technologies at the Institute for Progress, to discuss how energy constraints could bottleneck AI development. They explore how AI training clusters will soon require gigawatts of power—equivalent to multiple nuclear plants—with projections showing a single cluster needing 5 gigawatts by 2030. Tim explains why behind-the-meter generation and geothermal energy offer promising solutions while regulatory hurdles like NEPA and transmission permitting create "litigation doom loops" that threaten America's competitiveness. The conversation covers the global race for compute infrastructure, with China and the UAE making aggressive investments while the US struggles with permitting delays, highlighting how energy policy will determine which nations lead the AI revolution. –Full transcript available here: www.complexsystemspodcast.com/the-ai-energy-bottleneck-with-tim-fist/–Sponsor:  VantaVanta automates security compliance and builds trust, helping companies streamline ISO, SOC 2, and AI framework certifications. Learn more at https://vanta.com/complex–Recommended in this episode:Compute in America https://ifp.org/compute-in-america/Tim Fist on Twitter https://x.com/fiiiiiist The Enchippening by Sarah Constantin https://sarahconstantin.substack.com/p/the-enchippening Solar economics with Casey Handmer https://open.spotify.com/episode/0GHegWgLSubYxvATmbWhQu?si=VKJYaSwaRJq_YcK8kJIdvQ AI & Power economics with Azeem Azhar https://open.spotify.com/episode/3KkvPiYpGvXCRukWxHP7Ch?si=RPEjrs67S9CFA0lLak6OVAFracking with Austin Vernon  https://open.spotify.com/episode/0YDV1XyjUCM2RtuTcBGYH9?si=hSniC3N0QkqhF74ra-XAcA Economics of the grid with Travis Dauwalter https://open.spotify.com/episode/5JY8e84sEXmHFlc8IR2kRb?si=BsqMZGu6Qr-2F7-RSyyEhw–Timestamps:(00:00) Intro(00:40) Energy bottlenecks in AI development(02:56) Technical and policy solutions for energy needs(05:18) Challenges in transmission infrastructure(12:14) Behind the meter generation explained(17:50) Solar and storage: The future of energy(18:47) Sponsor: Vanta(20:05) Solar and storage: The future of energy (part 2)(29:07) Power purchase agreements and financing(33:17) Financing geothermal wells(33:53) The promise of geothermal energy(35:25) Challenges in geothermal adoption(36:59) Industrial applications of geothermal heat(45:01) Geothermal energy and national security(49:27) Global investments in AI and energy infrastructure(56:29) Policy and technical expertise in AI(01:00:54) The role of government in technological advancements(01:05:07) Wrap

Dive deep into the fascinating world of modern data centers with Sr. Principal Engineer at AWS, Stephen Callahan. Discover how AI is revolutionizing data center design, why nothing is uninteresting at scale, and the innovative ways AWS is tackling sustainability while powering the future of cloud computing. Learn more: AWS Global Infrastructure: https://aws.amazon.com/about-aws/global-infrastructure/ More about Data Center Innovations: https://press.aboutamazon.com/2024/12/aws-announces-new-data-center-components-to-support-ai-innovation-and-further-improve-energy-efficiency

Evan Conrad, co-founder of SF Compute, joined us to talk about how they started as an AI lab that avoided bankruptcy by selling GPU clusters, why CoreWeave financials look like a real estate business, and how GPUs are turning into a commodities market. Chapters: 00:00:05 - Introductions 00:00:12 - Introduction of guest Evan Conrad from SF Compute 00:00:12 - CoreWeave Business Model Discussion 00:05:37 - CoreWeave as a Real Estate Business 00:08:59 - Interest Rate Risk and GPU Market Strategy Framework 00:16:33 - Why Together and DigitalOcean will lose money on their clusters 00:20:37 - SF Compute's AI Lab Origins 00:25:49 - Utilization Rates and Benefits of SF Compute Market Model 00:30:00 - H100 GPU Glut, Supply Chain Issues, and Future Demand Forecast 00:34:00 - P2P GPU networks 00:36:50 - Customer stories 00:38:23 - VC-Provided GPU Clusters and Credit Risk Arbitrage 00:41:58 - Market Pricing Dynamics and Preemptible GPU Pricing Model 00:48:00 - Future Plans for Financialization? 00:52:59 - Cluster auditing and quality control 00:58:00 - Futures Contracts for GPUs 01:01:20 - Branding and Aesthetic Choices Behind SF Compute 01:06:30 - Lessons from Previous Startups 01:09:07 - Hiring at SF Compute Chapters 00:00:00 Introduction and Background 00:00:58 Analysis of GPU Business Models 00:01:53 Challenges with GPU Pricing 00:02:48 Revenue and Scaling with GPUs 00:03:46 Customer Sensitivity to GPU Pricing 00:04:44 Core Weave's Business Strategy 00:05:41 Core Weave's Market Perception 00:06:40 Hyperscalers and GPU Market Dynamics 00:07:37 Financial Strategies for GPU Sales 00:08:35 Interest Rates and GPU Market Risks 00:09:30 Optimal GPU Contract Strategies 00:10:27 Risks in GPU Market Contracts 00:11:25 Price Sensitivity and Market Competition 00:12:21 Market Dynamics and GPU Contracts 00:13:18 Hyperscalers and GPU Market Strategies 00:14:15 Nvidia and Market Competition 00:15:12 Microsoft's Role in GPU Market 00:16:10 Challenges in GPU Market Dynamics 00:17:07 Economic Realities of the GPU Market 00:18:03 Real Estate Model for GPU Clouds 00:18:59 Price Sensitivity and Chip Design 00:19:55 SF Compute's Beginnings and Challenges 00:20:54 Navigating the GPU Market 00:21:54 Pivoting to a GPU Cloud Provider 00:22:53 Building a GPU Market 00:23:52 SF Compute as a GPU Marketplace 00:24:49 Market Liquidity and GPU Pricing 00:25:47 Utilization Rates in GPU Markets 00:26:44 Brokerage and Market Flexibility 00:27:42 H100 Glut and Market Cycles 00:28:40 Supply Chain Challenges and GPU Glut 00:29:35 Future Predictions for the GPU Market 00:30:33 Speculations on Test Time Inference 00:31:29 Market Demand and Test Time Inference 00:32:26 Open Source vs. Closed AI Demand 00:33:24 Future of Inference Demand 00:34:24 Peer-to-Peer GPU Markets 00:35:17 Decentralized GPU Market Skepticism 00:36:15 Redesigning Architectures for New Markets 00:37:14 Supporting Grad Students and Startups 00:38:11 Successful Startups Using SF Compute 00:39:11 VCs and GPU Infrastructure 00:40:09 VCs as GPU Credit Transformators 00:41:06 Market Timing and GPU Infrastructure 00:42:02 Understanding GPU Pricing Dynamics 00:43:01 Market Pricing and Preemptible Compute 00:43:55 Price Volatility and Market Optimization 00:44:52 Customizing Compute Contracts 00:45:50 Creating Flexible Compute Guarantees 00:46:45 Financialization of GPU Markets 00:47:44 Building a Spot Market for GPUs 00:48:40 Auditing and Standardizing Clusters 00:49:40 Ensuring Cluster Reliability 00:50:36 Active Monitoring and Refunds 00:51:33 Automating Customer Refunds 00:52:33 Challenges in Cluster Maintenance 00:53:29 Remote Cluster Management 00:54:29 Standardizing Compute Contracts 00:55:28 Unified Infrastructure for Clusters 00:56:24 Creating a Commodity Market for GPUs 00:57:22 Futures Market and Risk Management 00:58:18 Reducing Risk with GPU Futures 00:59:14 Stabilizing the GPU Market 01:00:10 SF Compute's Anti-Hype Approach 01:01:07 Calm Branding and Expectations 01:02:07 Promoting San Francisco's Beauty 01:03:03 Design Philosophy at SF Compute 01:04:02 Artistic Influence on Branding 01:05:00 Past Projects and Burnout 01:05:59 Challenges in Building an Email Client 01:06:57 Persistence and Iteration in Startups 01:07:57 Email Market Challenges 01:08:53 SF Compute Job Opportunities 01:09:53 Hiring for Systems Engineering 01:10:50 Financial Systems Engineering Role 01:11:50 Conclusion and Farewell

Send us a textSubscribe to AG Dillon Pre-IPO Stock Research at agdillon.com/subscribe;- Wednesday = secondary market valuations, revenue multiples, performance, index fact sheets- Saturdays = pre-IPO news and insights00:00 - Intro00:08 - Thinking Machines Targets $10B Valuation with $2B Seed Round  01:12 - ByteDance Revenue Hits $155B; Valuation Diverges  02:15 - Anysphere Revenue Quadruples; Eyes $10B Valuation  03:01 - Nuro Raises $106M at $6B Valuation  03:51 - Base Power Raises $200M to Scale Affordable Home Batteries  05:06 - Anthropic Launches Claude Max, Valued at $61.5B  06:15 - Ripple Acquires Hidden Road for $1.25B  07:13 - Canva Adds GenAI Tools; Valued at $37.9B  08:19 - Electricity Demand for AI Surges Globally  10:31 - OpenAI Rolls Out ChatGPT Memory Feature  11:30 - Google Joins Anthropic's Model Context Protocol  12:43 - Safe Superintelligence Taps Google Cloud for Compute  

In this episode of Web3 with Sam Kamani, Sam is joined by co-host Amanda Whitcroft to interview Hoansoo Lee, co-founder of Exabits.ai. With a PhD from Harvard and deep expertise in edge computing, Hoansoo shares how Exabits is decentralizing the GPU cloud for AI by combining high-performance chips like the H100 and Blackwell with tokenized infrastructure on Web3 rails.They explore why AI compute is the "new energy," how Exabits differentiates from competitors like CoreWeave, and the opportunities for DeFi and structured finance in this emerging landscape. Hoansoo also discusses the limitations of decentralized compute, the challenges around AI experimentation, and how data, compute, and causality intersect in building next-gen AI.Whether you're a founder building in AI, a researcher, or a curious investor, this episode is packed with deep insights into the future of decentralized compute and what's next in the AI x Web3 convergence.Key Timestamps[00:00:00] Introduction: Sam introduces co-host Amanda and guest Hoansoo Lee from Exabits.ai.[00:01:00] What is Exabits?: Hoansoo explains Exabits in one sentence—high-quality GPU compute for AI.[00:02:00] Who Uses It: Discussing their customer base across Web2 and Web3.[00:03:00] Hardware Stack: Exabits runs 60,000+ GPUs including H100s and Blackwells.[00:04:00] Competitive Landscape: Why Exabits is different from other Web3 dePIN projects.[00:05:00] Founding Story: How a background in edge computing led to building Exabits.[00:06:00] Go-to-Market: Customer acquisition through partnerships, referrals, and conferences.[00:07:00] Growth Opportunity: Why structured finance and GPU financialization is the next big thing.[00:08:00] AI Efficiency vs. Demand: DeepSeek, scaling laws, and the compute boom.[00:10:00] Energy + Compute: AI's demand for energy and its parallels to historical tech trends.[00:11:00] Decentralized Compute: Limitations of latency-sensitive decentralized AI infrastructure.[00:13:00] AI = Bitcoin Mining 2.0: The evolution from minting Bitcoin to minting intelligence.[00:14:00] Pillars of AI: From compute/data/models to experimentation and causal inference.[00:17:00] AI Limits: Why synthetic data can't replace real-world experimentation.[00:18:00] Scarcity & Innovation: How chip scarcity could spark further innovation.[00:20:00] In-House Servers: Why building H200 racks in-house is a differentiator.[00:21:00] How It Works: A user's experience on Exabits from login to compute access.[00:23:00] Founder Advice: Hoansoo's take on building something with real customers and solid fundamentals[00:24:00] Roadmap: Data center expansion, orchestration features, and governance via staking.[00:25:00] TGE Ahead: Exabits' upcoming token generation event and next steps.Connecthttps://www.exabits.ai/https://www.linkedin.com/company/exabitsai/https://x.com/exa_bitshttps://www.linkedin.com/in/hoansoo-lee-21586b9/https://www.linkedin.com/in/amanda-whitcroft-324879164/DisclaimerNothing mentioned in this podcast is investment advice and please do your own research. Finally, it would mean a lot if you can leave a review of this podcast on Apple Podcasts or Spotify and share this podcast with a friend.Be a guest on the podcast or contact us - https://www.web3pod.xyz/

Hosts Simon and Jillian discuss how you can uncover hidden trends and make data-driven decisions - all through natural conversation, with Amazon Q in Quicksight, plus, more of the latest updates from AWS. 00:00 - Intro, 00:22 - Top Stories, 02:50 - Analytics, 03:35 - Application Integrations, 04:48 - Amazon Sagemaker, 05:29 - Amazon Bedrock Knowledge Bases, 05:48- Amazon Polly, 06:46 - Amazon Bedrock, 07:31 - Amazon Bedrock Model Evolution LLM, 08:29 - Business Application, 08:58 - Compute, 09:51 - Contact Centers, 10:54 - Containers, 11:12 - Database, 14:21 - Developer Tools, 15:20 - Front End Web and Mobile, 15:45 - Games, 16:04 - Management and Governance, 16:35 - Media Services, 16:47 - Network and Content Delivery, 19:39 - Security Identity and Compliance, 20:24 - Serverless, 21:48 - Storage, 22:43 - Wrap up Show Notes: https://dqkop6u6q45rj.cloudfront.net/shownotes-20250404-184823.html

In this episode of the ABCDs Roundup, we break down OpenAI's massive $40 billion funding round, led by SoftBank, and its impact on AI infrastructure. We explore AMD's $4.9 billion acquisition of ZT Systems as it challenges Nvidia in the AI data center wars and take a broader look at tariffs affecting the AI and blockchain industries. We also cover the latest developments in TikTok's U.S. ownership battle and Fidelity's Bitcoin market update, which predicts a potential acceleration phase in this week's chart. Remember to Stay Current! To learn more, visit us on the web at https://www.morgancreekcap.com/morgan-creek-digital/. To speak to a team member or sign up for additional content, please email mcdigital@morgancreekcap.com   Legal Disclaimer This podcast is for informational purposes only and should not be construed as investment advice or a solicitation for the sale of any security, advisory, or other service. Investments related to the themes and ideas discussed may be owned by funds managed by the host and podcast guests. Any conflicts mentioned by the host are subject to change. Listeners should consult their personal financial advisors before making any investment decisions.

Kevin Scott is the CTO of Microsoft, where he leads the company's AI and technology strategy at global scale and played a pivotal role in Microsoft's partnership with OpenAI. Prior to Microsoft, Kevin spent six years at Linkedin as SVP of Engineering. Kevin has also enjoyed advisory positions with Pinterest, Box, Code.org and more.  In Today's Episode We Discuss: 04:10 Where is Enduring Value in a World of AI 10:53 Why Scaling Laws are BS 12:26 What is the Bottleneck Today: Data, Compute or Algorithms 15:38: In 10 Years Time: What % of Data Usage will be Synthetic 20:04 How Will AI Agents Evolve Over the Next Five Years 23:34: Deepseek Evalution: Do We Underestimate China 28:34 The Future of Software Development 31:53 The Thing That Most Excites Me in AI is Tech Debt 35:01 Leadership Lessons from Satya Nadella 41:13 Quickfire Round  

Today, we're joined by Jonas Geiping, research group leader at Ellis Institute and the Max Planck Institute for Intelligent Systems to discuss his recent paper, “Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach.” This paper proposes a novel language model architecture which uses recurrent depth to enable “thinking in latent space.” We dig into “internal reasoning” versus “verbalized reasoning”—analogous to non-verbalized and verbalized thinking in humans, and discuss how the model searches in latent space to predict the next token and dynamically allocates more compute based on token difficulty. We also explore how the recurrent depth architecture simplifies LLMs, the parallels to diffusion models, the model's performance on reasoning tasks, the challenges of comparing models with varying compute budgets, and architectural advantages such as zero-shot adaptive exits and natural speculative decoding. The complete show notes for this episode can be found at https://twimlai.com/go/723.

Tim Fist, Director of Emerging Technology Policy at the Institute for Future Progress, and Arnab Datta, Director of Infrastructure Policy at IFP and Managing Director of Policy Implementation at Employ America, join Kevin Frazier, a Contributing Editor at Lawfare and adjunct professor at Delaware Law, to dive into the weeds of their thorough report on building America's AI infrastructure. The duo extensively studied the gulf between the stated goals of America's AI leaders and the practical hurdles to realizing those ambitious aims.Check out the entire report series here: Compute in AmericaTo receive ad-free podcasts, become a Lawfare Material Supporter at www.patreon.com/lawfare. You can also support Lawfare by making a one-time donation at https://givebutter.com/lawfare-institute.Support this show http://supporter.acast.com/lawfare. Hosted on Acast. See acast.com/privacy for more information.