Podcasts about Unified

Fish Discovering water is a great metaphor for Self Realisation. In this episode we explore that deeply along with some important questions about what is real and not real, what is the nature of reality! If you would like to see the video version of this Episode visit this link. For a free 14 day trial of Philip's Self Realisation App: The Living Soul use this Smartlink here. Timestamps: 00:00:00 Overview Invitation to Webinars and Community Calls 00:01:22 Music Intro & Book Overview of Webinars/Book Launch 00:04:33 Todays theme: A Fish Discovering Water   00:05:13 1st Slide (Pre-register for the book on Website.   00:06:09 The book is not about knowledge…   00:07:33 (Fish) video   00:12:54 We all go off into a voyage of discovery   00:14:35 Fibre Optic Lamp metaphor   00:15:51 2nd Slide (Pillars Of Creation)   00:18:30 (Meditation)   00:27:47 (End of meditation)   00:29:43 Unified whole   00:32:50 (Question) Julia - What Is It That We Perceive To Be Real?   00:43:10 (Question) - What Is Reality, Really?   00:51:22 Can't be acquired from effort   00:54:02 Slide, 2nd showing   00:56:49 Fish metaphor   00:57:34 Making the connection   00:58:37 The Flat-Landers…   1:02:00 Outro Monthly Episodes, Next: 3 April 2026, released at 4pm latest but usually takes until about 5pm to find its way to the various outlets.  

Hosts Grace Pratt, Monica Harrison, Bridget, Neftali Serrano, and late-arriving co-host Jen Thomas introduce themselves on the Integrated Care Podcast and share CFHA updates, including the virtual spring conference (May 6–7) and a call for proposals for the October fall conference in St. Louis, plus cohort-based trainings on cfha.net. The group workshops a planned fall mini-summit prompted by Naftali's concern that government momentum for behavioral health integration is waning while cost pressures, health tech fragmentation, and changes to primary care accelerate. They discuss the need for a unified message centered on team-based care, the operational barriers within health systems, and gaps between clinical progress, organizational change, and policy influence, emphasizing relationship-building with administrators and policymakers so integrated care is included in future payment models and legislation.

Luma introduced Luma Agents, powered by its new “Unified Intelligence” models, designed to coordinate multiple AI systems and generate end-to-end creative work across text, images, video and audio. Learn more about your ad choices. Visit podcastchoices.com/adchoices

In this Denatured episode, Jennifer C. Smith-Parker speaks to Ram May-Ron, managing partner at FreeMind Group, and Ravi Kiron, managing director at Biopharma Strategy Advisors. We'll discuss how best to tailor an investment approach of both nondilutive funding and family offices to overcome the drug development valley of death. Host⁠Jennifer Smith-Parker⁠, Director of Insights, BioSpaceGuestsRam May-Ron, Managing Partner, FreeMind GroupRavi Kiron, Managing Director, Biopharma Strategy AdvisorsDisclaimer: The views expressed in this discussion by guests are their own and do not represent those of their organizations.

Jonathan Schanzer reports that Iran's attacks on neutral Gulf nations backfire, pushing previously hesitant allies like Qatar and Oman toward a unified front with Israel and the United States. 12.XERXES OF PERSIA

Judy Dempsey reports that recent polls show US voters oppose intervention in Iran, while rumors of internal administration friction suggest a lack of unified strategy for the expanding war. 4.1890 PERSIA

We got a humdinger for you this week on The Loftus Party podcast with Michael Loftus. Yes. News? Music? Laughs and some spicy content? Oh, yeah. On this show, we'll reveal our unified conspiracy theory. It's the one theory that ties a lot of this BS together. Epstein, Iran and more. Speaking of Epstein it feels like the moment the Clintons got done testifying, we were jumping into a conflict with Iran. Hmmm...We'll get into the Iran stuff. Or as I'm calling it: The Persia stuff! There's news on the war on fraud and we've also got updates on Friday Night Bangers, Fallout season 2 and more! Let's go! Thanks for being here ya sexy heathens!Want to show your support and get all the cool extras? We're on Locals and Patreon! Join up. Join in. Let us begin! See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

This episode features Cliff Fisher, Senior Solutions Architect at Semperis and former Senior Technical Program Manager on Microsoft's Active Directory product group.With over a decade spent inside Microsoft supporting enterprise customers and helping guide Active Directory's security and roadmap, Cliff brings a rare insider perspective on what's actually happening behind the scenes of one of the world's most widely deployed identity platforms.In this episode, Cliff tackles the question many organizations are still asking: Is Active Directory really going away? He explains why the shift to cloud identity has moved far slower than expected, shares polling data that confirms hybrid environments are here for the long term, and breaks down how Microsoft is still investing in AD through security hardening, supportability improvements, and features like Windows LAPS.This episode offers a clearer look at why Active Directory remains central to enterprise identity and what defenders need to prepare for as hybrid becomes the default reality.Guest Bio With nearly 20 years of Active Directory experience across varied roles in system administration, support, debugging, and program management, Cliff spent over a decade at Microsoft supporting Premier and Unified customers and, most recently, managing the releases of Windows LAPS, new features for Server 2025, and monthly security and quality updates. In January of 2026, he joined Semperis, bringing his unique blend of skills, perspectives, and passion to their stacked roster of established identity experts.Guest Quote  “The easiest way to get everyone secure is to get people all to the cloud. What [Microsoft] didn't realize... is that customers just aren't going to be able to absorb change at that rate, and especially at that cost. Shifting to the cloud is not cheap.”Time stamps 01:45 Meet Cliff Fisher: Identity security expert 04:24 Microsoft's Vision for Active Directory 07:58 Challenges and Future of Active Directory 23:12 The Complexity of AD Code and Security Vulnerabilities 24:39 Understanding Fuzzing and Its Importance 27:28 Domain Join Hardening and Its Challenges 36:28 Windows LAPS and Future Security Measures 41:39 Why is RC4 Going Away? 45:14 Conclusion and Final ThoughtsSponsor The HIP Podcast is brought to you by Semperis, the leader in identity-driven cyber resilience for the hybrid enterprise. Trusted by the world's leading businesses, Semperis protects critical Active Directory and Entra ID environments from cyberattacks, ensuring rapid recovery and business continuity when every second counts. Visit semperis.com to learn more.LinksConnect with Cliff on LinkedInConnect with Sean on LinkedInDon't miss future episodesLearn more about SemperisSubmit your proposal to speak at HIP Conf 26: HIP Conf 26 Call for Papers Submission

Josiah and Micah Kennealy talk with Aaron Wadsworth from Grace Church in Arizona about Unified College. More about us: Learn more about youngadultstoday: www.youngadults.today Give to propel the ministry forward: https://tithe.ly/give?c=5350133 Resources: -Free eBook "10 Steps to Starting a Successful Young Adult Ministry": https://www.youngadults.today/book -Join our FaceBook Group Community with 2500+ leaders: https://www.facebook.com/groups/796270437396021 -Follow us on Instagram: https://www.instagram.com/youngadults.today/ -See you in Minneapolis this March 13-14th for the youngadultstoday leader conference: www.youngadults.today/conference -Limited Spots are available for our Coaching Communities launching February 16th: www.youngadults.today/coaching-communities 

Is American society more polarized than ever? How did the Founding Fathers handle division in their day? On this week's “Leaders and Legends” podcast, we interview renowned scholar Yuval Levin about his brilliant and compelling new book “American Covenant: How the Constitution Unified Our Nation—And Could Again”, and we do it courtesy of Marian University and its Richard G. Lugar Speaker Series.See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

Join Paul Gigliotti, CEO of California MBA, on Connect for Episode 254 with Tom Morelli from Wilco! Discover how Wilco is transforming mortgage origination with Charlie – a groundbreaking Production Optimization Platform (POP) that unites POS, LOS, AI, and business intelligence into one seamless system. Key topics: - Why traditional LOS/POS silos are broken and costing lenders $11-13K per loan - Making the borrower a true partner with real-time transparency and no more repeated document requests - Breaking workflows into discrete activities for parallel processing, higher quality, and lower fallout Wilco's founding vision from industry legends (DocuSign, Optimal Blue, Ellie Mae) - The Brimma acquisition: AI-powered content intelligence (DocFlow OCR/ICR) for quick wins – even if you're not ready for full Charlie - Business intelligence that self-diagnoses bottlenecks and sets data-driven expectations Whether you're a lender battling high origination costs, prepping for the next refi wave, or exploring AI in mortgage tech, this episode is packed with insights! Timestamps: 0:00 - Intro & Sponsor Shoutouts (AmeriHome, Western Alliance Bank, etc.) 2:30 - Meet Tom Morelli & Wilco's Founding Team 5:00 - The Broken State of Origination: Costs, Silos & Duplication 9:45 - Core Philosophy: Unified POS/LOS for Borrower-Centric Experience 15:20 - Discrete Activities Workflow & Parallel Processing 20:10 - Real Borrower Impact: Speed, Accuracy & Fewer Frustrations 25:40 - Reducing Late Fallout & Cost to Originate 28:00 - Brimma Acquisition: AI DocFlow & Modular Content Intelligence 32:15 - Business Intelligence: Real-Time Bottlenecks & Stats-Driven Insights Subscribe to California MBA for more industry-leading conversations on advocacy, tech, and mortgage innovation. Like & comment: What's your biggest origination pain point? Learn more about Wilco: https://www.wilco.io/ (or official site) California MBA: https://www.camortgage.org/ #MortgageTech #LoanOrigination #AITech #MortgageLending #ProductionOptimization #CharliePlatform #Wilco #CaliforniaMBA #Podcast

Listen to Full Audio at https://podcasts.apple.com/us/podcast/scientist-vs-storyteller-benchmarking-gpt-5-2-claude/id1684415169?i=1000752001078For years, Latent Diffusion Models—the tech behind Stable Diffusion and DALL-E—have relied on a bit of an 'art form' called KL-regularization. Basically, researchers had to manually guess how much to compress an image before the AI started to lose the details. If you compressed too much, the image got blurry. Too little, and the model became too expensive to train.Enter Unified Latents, or UL.In a new paper out of DeepMind Amsterdam, researchers have introduced a framework that replaces that guesswork with a single, cohesive mathematical objective. Instead of training the compressor and the generator separately, UL trains the Encoder, the Prior, and the Decoder all at once.The 'Secret Sauce' here is something called Fixed Gaussian Noise Encoding. By injecting a constant, specific amount of noise during the encoding process, DeepMind has created a 'Maximum Precision Link.' This forces the encoder to be incredibly efficient, focusing only on the most important structures of an image.The results are staggering: UL achieved a state-of-the-art Video Distance score on the Kinetics-600 dataset and hit a competitive 1.4 FID on ImageNet—all while using significantly less computational power than traditional methods.This episode is made possible by our sponsors:

In this Foojay Podcast, we're celebrating a major milestone in Java development history: 25 years of IntelliJ IDEA.Think about it: IntelliJ IDEA launched in 2000, and since then, it has become the go-to IDE for millions of Java developers worldwide. From its revolutionary code completion and refactoring tools to AI-powered features and the recent unified Community and Ultimate release, IntelliJ has shaped how we write Java, and keeps reinventing itself to stay ahead.For this episode, I'm joined by three people from the JetBrains team who know this story inside and out. Marit van Dijk, developer advocate and contributor to the Foojay community. Anton Arhipov, also a developer advocate at JetBrains. And Dmitry Jemerov, who has been part of the IntelliJ IDEA story for a very long time.GuestsMarit van Dijkhttps://foojay.io/today/author/marit-van-dijk/https://www.linkedin.com/in/maritvandijk/https://mastodon.social/@maritvandijkAnton Arhipovhttps://www.linkedin.com/in/antonarhipov/Dmitry Jemerovhttps://www.linkedin.com/in/dmitry-jemerov-3a59b43a5/LinksWebsiteDocumentationBlogYouTubeLinkedInBlueskyTwitterFoojay Podcast #81: Maven 4 – The Future of Java Build AutomationVideo: IntelliJ IDEA: The Documentary | [OFFICIAL TRAILER] | Coming March 5thIntroducing Mellum: JetBrains' New LLM Built for Developers Mellum: Explore code-intelligent large language models for IDEs, AI assistants, research, and educationBirthday game websiteGame plugin in IntelliJ IDEAYou're Invited to IntelliJ IDEA Conf 2025!The Unified IntelliJ IDEA: More Free Features, a Better Experience, Smoother FlowVideo: Troubleshooting Spring Boot Applications with the Spring DebuggerSpring Debugger pluginPlugin for IntelliJ IDEA (and other IDEs) created by Frank: Recent Projects OrganizedContent00:00 Introduction of topic and guests01:36 Now JetBrains started02:31 Licensed software in an open-source world06:37 Other JetBrains IDEs07:46 Why Kotlin was created08:50 The challenge of maintaining all the tools10:36 How the guests joined JetBrains14:03 IntelliJ versus IntelliJ IDEA, history of the name15:10 Most important ongoing changes in IDEs17:55 Unified distribution of IntelliJ IDEA and the history of the open-source version21:28 The number of people at JetBrains23:31 the "business model" behind Kotlin24:39 The impact of AI, LLM, Chat interfaces,...35:49 Upcoming evolutions in IntelliJ IDEA38:07 About shortcuts and the many features and plugins in IntelliJ IDEA46:36 Announcements: IntelliJ IDEA Conf 2026 and Documentary Trailer48:35 The IntelliJ IDEA Birthday Game49:24 Conclusions

A firefighter testifies that his warnings about the Lachman Fire were ignored. Netflix drops out of the Warner Brothers bidding war. The financial aid that's available to 3 out of every 5 California students. Plus, more from Morning Edition. Support The L.A. Report by donating at LAist.com/join and by visiting https://laist.comSupport the show: https://laist.com

Alan and Sam dive into configuration drift with M365 and assessment against security baselines. Here are a few things we covered:What is configuration drift, and why a security baseline is important.Microsoft's new Unified Tenant Configuration Management APIs for managing drifts.How to get it set up and advice for authentication.What's covered, the limitations and how it works.What did you think of this episode? Give us some feedback via our contact form, Or leave us a voice message in the bottom right corner of our site.

Adam Engst returns to the show to talk, in detail, about certain of the UI changes in iOS 26 and Apple's version 26 OSes overall. In particular, the new Unified view in the Phone app, and the Filter pop-up menu in both the Phone and Messages apps. Also: a shoutout to Balloon Help.

Digital transformation in procurement has been "imminent" for over a decade, however, Legacy Thinking Is the Real Bottleneck! Boards talk about automation. CFOs talk about control. Procurement leaders talk about value creation. And yet, across industries, source-to-pay (S2P) remains one of the most stubbornly legacy bound functions in the enterprise. The irony? Procurement should be one of the easiest functions to modernize. It is structured, process driven, data rich, and measurable. But in practice, S2P transformation efforts stall, underdeliver, or quietly die after expensive, lengthy and limited implementation cycles. Why? The bottleneck isn't technology. It's legacy gravity. The Hidden Cost of "Good Enough" Procurement Many organizations still operate on a patchwork of: ERP systems and bolt-ons built for another era Email based approvals Manual vendor onboarding Disconnected sourcing tools Excel driven reporting and even pen and paper These systems "work"… in the same way that a fax machine technically still works. The problem is that legacy procurement systems were designed for control and record keeping, not agility, collaboration, or strategic insight. They reflect a time when procurement was administrative. Today, it's expected to be strategic. That shift breaks the old model. Where Source-to-Pay Innovation Gets Stuck 1. ERP-Centric Thinking For years, procurement innovation meant adding modules to an ERP. But ERPs are transactional systems of record, not innovation platforms. They are excellent at posting journal entries. They are poor at enabling dynamic sourcing, supplier collaboration, or real time spend intelligence. Trying to build modern procurement on top of ERP architecture is like building a streaming service on top of a DVD player. 1. Change Fatigue and Organisational Inertia Procurement teams are often overworked and understaffed. Digital transformation becomes "another project" layered on top of operational pressure. Without clear ROI and intuitive user experience, adoption fails. Stakeholders revert to email. Maverick spend returns. The transformation narrative and urgency fades. 1. Fragmented Tool Stacks Organisations frequently assemble S2P capabilities from multiple vendors: One for sourcing One for contract management One for P2P Another for analytics Integration becomes the project. Data reconciliation becomes a full-time job. Innovation slows under its own complexity. 1. Supplier Experience Is an Afterthought Most legacy procurement systems optimize for internal compliance, not supplier usability. Clunky onboarding. Repetitive data entry. Limited transparency. In an era where supplier relationships are strategic assets, this friction is more than inconvenient — it's counterproductive. 1. Procurement Still Seen as Cost Control Perhaps the deepest legacy issue is philosophical. Many executive teams still view procurement primarily as a cost-cutting function. But modern S2P innovation unlocks: Risk visibility ESG traceability Working capital optimization Data driven negotiation leverage Cross functional alignment Actionable game changing business intelligence insights When procurement is framed as a back-office function, investment remains incremental. When it's framed as a strategic value driver, transformation becomes inevitable. What Modern Source-to-Pay Should Actually Look Like True S2P innovation isn't about digitising paperwork. It's about re-architecting the procurement experience. That includes: Consumer grade UX that drives adoption Unified workflows from sourcing through payment Real-time spend visibility Embedded analytics Supplier-first design Automation of approvals and compliance Configurability without heavy IT dependency In short, S2P should feel like modern SaaS, not a compliance portal from 2009, with the UX of teletext from the 1990's. The New Model: Agile, Unified, Intuitive Forward-thinking organizations are abandoning monolithic, ERP bound procurement stacks in favor of flexi...

In this sermon we look at John 5 and talk about the person of Jesus, at the center of the Biblical story.For more teaching, visit citizenscharlotte.com/teaching

Today's guest is Dan Keto, President and Co-founder at Easy Metrics, where he focuses on helping warehouse and distribution teams turn fragmented transactional data into a unified "single pane of glass" that supports faster diagnosis of variance and more defensible decision-making. Dan joins Emerj's Matthew DeMello to explore what a solid data foundation looks like in warehouse networks — and why it matters before teams attempt to layer AI on top. He also shares practical takeaways on how enterprises can align stakeholders around a common data language, avoid costly "AI-first" missteps, and use repeatable investigations and alerts to surface real cost drivers. This episode is sponsored by Easy Metrics. If you've enjoyed or benefited from some of the insights of this episode, consider leaving us a five-star review on Apple Podcasts, and let us know what you learned, found helpful, or liked most about this show!

The International Fresh Produce Association's vice president of sustainability shares how this new strategic path will standardize metrics and streamline efforts across the global supply chain.See omnystudio.com/listener for privacy information.

Can one person make a difference? Absolutely! But nothing can beat a group of people with a common bond and a common goal. Dr. Tony Evans explores ways to become that kind of committed community as we look at the power of a united church.

Can one person make a difference? Absolutely! But nothing can beat a group of people with a common bond and a common goal. Dr. Tony Evans explores ways to become that kind of committed community as we look at the power of a united church.

The latest North State and California news on our airwaves for Wednesday, Feb. 18, 2026.

Send a textIn the season premiere of NATA-Cast, hosts Mollie Pillman, MS, MBA, CAE, and Katie Scott, MS, ATC, CAE, share a special live recording from the recent AT Alliance meeting, bringing together leaders from across the athletic training profession. NATA President A.J. Duffy III, MS, ATC, PT  joins Brian Conway, LAT, ATC, of the Board of Certification, MaryBeth Horodyski, EdD, ATC, LAT of the NATA Research and Education Foundation, and Toni Torres-McGehee, PhD, SCAT, ATC of the Commission on Accreditation of Athletic Training Education to discuss how the Alliance has evolved into a unified effort to advance athletic training across workplace settings. The conversation highlights strategic planning and governance priorities, credentialing and accreditation updates, research and workforce initiatives and coordinated advocacy to strengthen recognition and value of athletic trainers. The episode also explores key challenges facing the profession, including recruitment and retention, transition to practice, compensation and work-life balance, concluding with audience dialogue on supporting mid-career professionals and sustaining the workforce.NATA-Cast is produced by Association Briefings.Follow The National Athletic Trainers' Association on social media!FacebookXInstagramLinkedInHave an idea for an episode or series? Send us an email! thenatacast@nata.org

PREVIEW FOR LATER TODAY Guest: Michael Vorenberg. Vorenberg discusses how President Johnson'sobstructionism inadvertently unified Republicans, enabling the passage of the Reconstruction Acts and the 14th and 15th Amendments.1865 INAUGURATION OF ANDREW JOHNSON FOLLOWING LINCOL'S DEATH.

As we get to the end of Ephesians 4, Paul teaches the church how to display the oneness and holiness that we have in Christ.

John talks about a Senate vote failing as the clock continues to tick down on DHS funding. A partial shutdown looms unless Republicans can meet the Dem demands. He also discusses Thom Homan who says the immigration crackdown on Minnesota is over, for now, and he and his goons believe they have left the place whiter than they found it. Then, he interviews Dan Flores who is the A. B. Hammond Professor Emeritus of Western History at the University of Montana and the author of eleven books on aspects of American history. They discuss his new book Coyote America which traces both the five-million-year-long biological story of coyotes, as well as their cultural evolution from preeminence in Native American religions to haplessness before the Road Runner. A deeply American tale, the story of the coyote in the American West and then across the entire country is a sort of Manifest Destiny in reverse, with a pioneering hero whose career holds up an uncanny mirror to the successes and failures of American expansionism. Then, John welcomes Stuart Delony. He is a writer and podcaster whose work examines faith, power, and the cultural consequences of certainty. A former pastor, he is the host of the Snarky Faith podcast and a columnist focused on American Christianity, politics, and end-times theology. John discusses his new book The Tribulation Survival Guide is for exvangelicals, spiritual misfits, and connoisseurs of dark humor. This isn't your typical devotional—it's a survivalist satire for anyone who's ever questioned faith, feared the Beast, or accidentally attended a prophecy conference. Delivered with the solemnity of a Cold War safety pamphlet and the wit of a burned-out prophet, this deadpan, government-grade field manual offers step-by-step guidance for navigating the world's most awkward apocalypse. Whether you've been left behind by the Rapture—or just by organized religion—you'll find something disturbingly familiar in its pages. From decoding Antichrist branding strategies to surviving plagues, televangelists, and HOA-controlled hellscapes, this guide blends biting satire with faux-instructional sincerity. Inside you'll find checklists, diagnostic quizzes, heavenly bureaucracy hacks, and DIY hell décor tips (lava optional)—all designed to help you stay alive, or at least mildly amused, through the end of all things.See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

From rewriting Google's search stack in the early 2000s to reviving sparse trillion-parameter models and co-designing TPUs with frontier ML research, Jeff Dean has quietly shaped nearly every layer of the modern AI stack. As Chief AI Scientist at Google and a driving force behind Gemini, Jeff has lived through multiple scaling revolutions from CPUs and sharded indices to multimodal models that reason across text, video, and code.Jeff joins us to unpack what it really means to “own the Pareto frontier,” why distillation is the engine behind every Flash model breakthrough, how energy (in picojoules) not FLOPs is becoming the true bottleneck, what it was like leading the charge to unify all of Google's AI teams, and why the next leap won't come from bigger context windows alone, but from systems that give the illusion of attending to trillions of tokens.We discuss:* Jeff's early neural net thesis in 1990: parallel training before it was cool, why he believed scaling would win decades early, and the “bigger model, more data, better results” mantra that held for 15 years* The evolution of Google Search: sharding, moving the entire index into memory in 2001, softening query semantics pre-LLMs, and why retrieval pipelines already resemble modern LLM systems* Pareto frontier strategy: why you need both frontier “Pro” models and low-latency “Flash” models, and how distillation lets smaller models surpass prior generations* Distillation deep dive: ensembles → compression → logits as soft supervision, and why you need the biggest model to make the smallest one good* Latency as a first-class objective: why 10–50x lower latency changes UX entirely, and how future reasoning workloads will demand 10,000 tokens/sec* Energy-based thinking: picojoules per bit, why moving data costs 1000x more than a multiply, batching through the lens of energy, and speculative decoding as amortization* TPU co-design: predicting ML workloads 2–6 years out, speculative hardware features, precision reduction, sparsity, and the constant feedback loop between model architecture and silicon* Sparse models and “outrageously large” networks: trillions of parameters with 1–5% activation, and why sparsity was always the right abstraction* Unified vs. specialized models: abandoning symbolic systems, why general multimodal models tend to dominate vertical silos, and when vertical fine-tuning still makes sense* Long context and the illusion of scale: beyond needle-in-a-haystack benchmarks toward systems that narrow trillions of tokens to 117 relevant documents* Personalized AI: attending to your emails, photos, and documents (with permission), and why retrieval + reasoning will unlock deeply personal assistants* Coding agents: 50 AI interns, crisp specifications as a new core skill, and how ultra-low latency will reshape human–agent collaboration* Why ideas still matter: transformers, sparsity, RL, hardware, systems — scaling wasn't blind; the pieces had to multiply togetherShow Notes:* Gemma 3 Paper* Gemma 3* Gemini 2.5 Report* Jeff Dean's “Software Engineering Advice fromBuilding Large-Scale Distributed Systems” Presentation (with Back of the Envelope Calculations)* Latency Numbers Every Programmer Should Know by Jeff Dean* The Jeff Dean Facts* Jeff Dean Google Bio* Jeff Dean on “Important AI Trends” @Stanford AI Club* Jeff Dean & Noam Shazeer — 25 years at Google (Dwarkesh)—Jeff Dean* LinkedIn: https://www.linkedin.com/in/jeff-dean-8b212555* X: https://x.com/jeffdeanGoogle* https://google.com* https://deepmind.googleFull Video EpisodeTimestamps00:00:04 — Introduction: Alessio & Swyx welcome Jeff Dean, chief AI scientist at Google, to the Latent Space podcast00:00:30 — Owning the Pareto Frontier & balancing frontier vs low-latency models00:01:31 — Frontier models vs Flash models + role of distillation00:03:52 — History of distillation and its original motivation00:05:09 — Distillation's role in modern model scaling00:07:02 — Model hierarchy (Flash, Pro, Ultra) and distillation sources00:07:46 — Flash model economics & wide deployment00:08:10 — Latency importance for complex tasks00:09:19 — Saturation of some tasks and future frontier tasks00:11:26 — On benchmarks, public vs internal00:12:53 — Example long-context benchmarks & limitations00:15:01 — Long-context goals: attending to trillions of tokens00:16:26 — Realistic use cases beyond pure language00:18:04 — Multimodal reasoning and non-text modalities00:19:05 — Importance of vision & motion modalities00:20:11 — Video understanding example (extracting structured info)00:20:47 — Search ranking analogy for LLM retrieval00:23:08 — LLM representations vs keyword search00:24:06 — Early Google search evolution & in-memory index00:26:47 — Design principles for scalable systems00:28:55 — Real-time index updates & recrawl strategies00:30:06 — Classic “Latency numbers every programmer should know”00:32:09 — Cost of memory vs compute and energy emphasis00:34:33 — TPUs & hardware trade-offs for serving models00:35:57 — TPU design decisions & co-design with ML00:38:06 — Adapting model architecture to hardware00:39:50 — Alternatives: energy-based models, speculative decoding00:42:21 — Open research directions: complex workflows, RL00:44:56 — Non-verifiable RL domains & model evaluation00:46:13 — Transition away from symbolic systems toward unified LLMs00:47:59 — Unified models vs specialized ones00:50:38 — Knowledge vs reasoning & retrieval + reasoning00:52:24 — Vertical model specialization & modules00:55:21 — Token count considerations for vertical domains00:56:09 — Low resource languages & contextual learning00:59:22 — Origins: Dean's early neural network work01:10:07 — AI for coding & human–model interaction styles01:15:52 — Importance of crisp specification for coding agents01:19:23 — Prediction: personalized models & state retrieval01:22:36 — Token-per-second targets (10k+) and reasoning throughput01:23:20 — Episode conclusion and thanksTranscriptAlessio Fanelli [00:00:04]: Hey everyone, welcome to the Latent Space podcast. This is Alessio, founder of Kernel Labs, and I'm joined by Swyx, editor of Latent Space. Shawn Wang [00:00:11]: Hello, hello. We're here in the studio with Jeff Dean, chief AI scientist at Google. Welcome. Thanks for having me. It's a bit surreal to have you in the studio. I've watched so many of your talks, and obviously your career has been super legendary. So, I mean, congrats. I think the first thing must be said, congrats on owning the Pareto Frontier.Jeff Dean [00:00:30]: Thank you, thank you. Pareto Frontiers are good. It's good to be out there.Shawn Wang [00:00:34]: Yeah, I mean, I think it's a combination of both. You have to own the Pareto Frontier. You have to have like frontier capability, but also efficiency, and then offer that range of models that people like to use. And, you know, some part of this was started because of your hardware work. Some part of that is your model work, and I'm sure there's lots of secret sauce that you guys have worked on cumulatively. But, like, it's really impressive to see it all come together in, like, this slittily advanced.Jeff Dean [00:01:04]: Yeah, yeah. I mean, I think, as you say, it's not just one thing. It's like a whole bunch of things up and down the stack. And, you know, all of those really combine to help make UNOS able to make highly capable large models, as well as, you know, software techniques to get those large model capabilities into much smaller, lighter weight models that are, you know, much more cost effective and lower latency, but still, you know, quite capable for their size. Yeah.Alessio Fanelli [00:01:31]: How much pressure do you have on, like, having the lower bound of the Pareto Frontier, too? I think, like, the new labs are always trying to push the top performance frontier because they need to raise more money and all of that. And you guys have billions of users. And I think initially when you worked on the CPU, you were thinking about, you know, if everybody that used Google, we use the voice model for, like, three minutes a day, they were like, you need to double your CPU number. Like, what's that discussion today at Google? Like, how do you prioritize frontier versus, like, we have to do this? How do we actually need to deploy it if we build it?Jeff Dean [00:02:03]: Yeah, I mean, I think we always want to have models that are at the frontier or pushing the frontier because I think that's where you see what capabilities now exist that didn't exist at the sort of slightly less capable last year's version or last six months ago version. At the same time, you know, we know those are going to be really useful for a bunch of use cases, but they're going to be a bit slower and a bit more expensive than people might like for a bunch of other broader models. So I think what we want to do is always have kind of a highly capable sort of affordable model that enables a whole bunch of, you know, lower latency use cases. People can use them for agentic coding much more readily and then have the high-end, you know, frontier model that is really useful for, you know, deep reasoning, you know, solving really complicated math problems, those kinds of things. And it's not that. One or the other is useful. They're both useful. So I think we'd like to do both. And also, you know, through distillation, which is a key technique for making the smaller models more capable, you know, you have to have the frontier model in order to then distill it into your smaller model. So it's not like an either or choice. You sort of need that in order to actually get a highly capable, more modest size model. Yeah.Alessio Fanelli [00:03:24]: I mean, you and Jeffrey came up with the solution in 2014.Jeff Dean [00:03:28]: Don't forget, L'Oreal Vinyls as well. Yeah, yeah.Alessio Fanelli [00:03:30]: A long time ago. But like, I'm curious how you think about the cycle of these ideas, even like, you know, sparse models and, you know, how do you reevaluate them? How do you think about in the next generation of model, what is worth revisiting? Like, yeah, they're just kind of like, you know, you worked on so many ideas that end up being influential, but like in the moment, they might not feel that way necessarily. Yeah.Jeff Dean [00:03:52]: I mean, I think distillation was originally motivated because we were seeing that we had a very large image data set at the time, you know, 300 million images that we could train on. And we were seeing that if you create specialists for different subsets of those image categories, you know, this one's going to be really good at sort of mammals, and this one's going to be really good at sort of indoor room scenes or whatever, and you can cluster those categories and train on an enriched stream of data after you do pre-training on a much broader set of images. You get much better performance. If you then treat that whole set of maybe 50 models you've trained as a large ensemble, but that's not a very practical thing to serve, right? So distillation really came about from the idea of, okay, what if we want to actually serve that and train all these independent sort of expert models and then squish it into something that actually fits in a form factor that you can actually serve? And that's, you know, not that different from what we're doing today. You know, often today we're instead of having an ensemble of 50 models. We're having a much larger scale model that we then distill into a much smaller scale model.Shawn Wang [00:05:09]: Yeah. A part of me also wonders if distillation also has a story with the RL revolution. So let me maybe try to articulate what I mean by that, which is you can, RL basically spikes models in a certain part of the distribution. And then you have to sort of, well, you can spike models, but usually sometimes... It might be lossy in other areas and it's kind of like an uneven technique, but you can probably distill it back and you can, I think that the sort of general dream is to be able to advance capabilities without regressing on anything else. And I think like that, that whole capability merging without loss, I feel like it's like, you know, some part of that should be a distillation process, but I can't quite articulate it. I haven't seen much papers about it.Jeff Dean [00:06:01]: Yeah, I mean, I tend to think of one of the key advantages of distillation is that you can have a much smaller model and you can have a very large, you know, training data set and you can get utility out of making many passes over that data set because you're now getting the logits from the much larger model in order to sort of coax the right behavior out of the smaller model that you wouldn't otherwise get with just the hard labels. And so, you know, I think that's what we've observed. Is you can get, you know, very close to your largest model performance with distillation approaches. And that seems to be, you know, a nice sweet spot for a lot of people because it enables us to kind of, for multiple Gemini generations now, we've been able to make the sort of flash version of the next generation as good or even substantially better than the previous generations pro. And I think we're going to keep trying to do that because that seems like a good trend to follow.Shawn Wang [00:07:02]: So, Dara asked, so it was the original map was Flash Pro and Ultra. Are you just sitting on Ultra and distilling from that? Is that like the mother load?Jeff Dean [00:07:12]: I mean, we have a lot of different kinds of models. Some are internal ones that are not necessarily meant to be released or served. Some are, you know, our pro scale model and we can distill from that as well into our Flash scale model. So I think, you know, it's an important set of capabilities to have and also inference time scaling. It can also be a useful thing to improve the capabilities of the model.Shawn Wang [00:07:35]: And yeah, yeah, cool. Yeah. And obviously, I think the economy of Flash is what led to the total dominance. I think the latest number is like 50 trillion tokens. I don't know. I mean, obviously, it's changing every day.Jeff Dean [00:07:46]: Yeah, yeah. But, you know, by market share, hopefully up.Shawn Wang [00:07:50]: No, I mean, there's no I mean, there's just the economics wise, like because Flash is so economical, like you can use it for everything. Like it's in Gmail now. It's in YouTube. Like it's yeah. It's in everything.Jeff Dean [00:08:02]: We're using it more in our search products of various AI mode reviews.Shawn Wang [00:08:05]: Oh, my God. Flash past the AI mode. Oh, my God. Yeah, that's yeah, I didn't even think about that.Jeff Dean [00:08:10]: I mean, I think one of the things that is quite nice about the Flash model is not only is it more affordable, it's also a lower latency. And I think latency is actually a pretty important characteristic for these models because we're going to want models to do much more complicated things that are going to involve, you know, generating many more tokens from when you ask the model to do so. So, you know, if you're going to ask the model to do something until it actually finishes what you ask it to do, because you're going to ask now, not just write me a for loop, but like write me a whole software package to do X or Y or Z. And so having low latency systems that can do that seems really important. And Flash is one direction, one way of doing that. You know, obviously our hardware platforms enable a bunch of interesting aspects of our, you know, serving stack as well, like TPUs, the interconnect between. Chips on the TPUs is actually quite, quite high performance and quite amenable to, for example, long context kind of attention operations, you know, having sparse models with lots of experts. These kinds of things really, really matter a lot in terms of how do you make them servable at scale.Alessio Fanelli [00:09:19]: Yeah. Does it feel like there's some breaking point for like the proto Flash distillation, kind of like one generation delayed? I almost think about almost like the capability as a. In certain tasks, like the pro model today is a saturated, some sort of task. So next generation, that same task will be saturated at the Flash price point. And I think for most of the things that people use models for at some point, the Flash model in two generation will be able to do basically everything. And how do you make it economical to like keep pushing the pro frontier when a lot of the population will be okay with the Flash model? I'm curious how you think about that.Jeff Dean [00:09:59]: I mean, I think that's true. If your distribution of what people are asking people, the models to do is stationary, right? But I think what often happens is as the models become more capable, people ask them to do more, right? So, I mean, I think this happens in my own usage. Like I used to try our models a year ago for some sort of coding task, and it was okay at some simpler things, but wouldn't do work very well for more complicated things. And since then, we've improved dramatically on the more complicated coding tasks. And now I'll ask it to do much more complicated things. And I think that's true, not just of coding, but of, you know, now, you know, can you analyze all the, you know, renewable energy deployments in the world and give me a report on solar panel deployment or whatever. That's a very complicated, you know, more complicated task than people would have asked a year ago. And so you are going to want more capable models to push the frontier in the absence of what people ask the models to do. And that also then gives us. Insight into, okay, where does the, where do things break down? How can we improve the model in these, these particular areas, uh, in order to sort of, um, make the next generation even better.Alessio Fanelli [00:11:11]: Yeah. Are there any benchmarks or like test sets they use internally? Because it's almost like the same benchmarks get reported every time. And it's like, all right, it's like 99 instead of 97. Like, how do you have to keep pushing the team internally to it? Or like, this is what we're building towards. Yeah.Jeff Dean [00:11:26]: I mean, I think. Benchmarks, particularly external ones that are publicly available. Have their utility, but they often kind of have a lifespan of utility where they're introduced and maybe they're quite hard for current models. You know, I, I like to think of the best kinds of benchmarks are ones where the initial scores are like 10 to 20 or 30%, maybe, but not higher. And then you can sort of work on improving that capability for, uh, whatever it is, the benchmark is trying to assess and get it up to like 80, 90%, whatever. I, I think once it hits kind of 95% or something, you get very diminishing returns from really focusing on that benchmark, cuz it's sort of, it's either the case that you've now achieved that capability, or there's also the issue of leakage in public data or very related kind of data being, being in your training data. Um, so we have a bunch of held out internal benchmarks that we really look at where we know that wasn't represented in the training data at all. There are capabilities that we want the model to have. Um, yeah. Yeah. Um, that it doesn't have now, and then we can work on, you know, assessing, you know, how do we make the model better at these kinds of things? Is it, we need different kind of data to train on that's more specialized for this particular kind of task. Do we need, um, you know, a bunch of, uh, you know, architectural improvements or some sort of, uh, model capability improvements, you know, what would help make that better?Shawn Wang [00:12:53]: Is there, is there such an example that you, uh, a benchmark inspired in architectural improvement? Like, uh, I'm just kind of. Jumping on that because you just.Jeff Dean [00:13:02]: Uh, I mean, I think some of the long context capability of the, of the Gemini models that came, I guess, first in 1.5 really were about looking at, okay, we want to have, um, you know,Shawn Wang [00:13:15]: immediately everyone jumped to like completely green charts of like, everyone had, I was like, how did everyone crack this at the same time? Right. Yeah. Yeah.Jeff Dean [00:13:23]: I mean, I think, um, and once you're set, I mean, as you say that needed single needle and a half. Hey, stack benchmark is really saturated for at least context links up to 1, 2 and K or something. Don't actually have, you know, much larger than 1, 2 and 8 K these days or two or something. We're trying to push the frontier of 1 million or 2 million context, which is good because I think there are a lot of use cases where. Yeah. You know, putting a thousand pages of text or putting, you know, multiple hour long videos and the context and then actually being able to make use of that as useful. Try to, to explore the über graduation are fairly large. But the single needle in a haystack benchmark is sort of saturated. So you really want more complicated, sort of multi-needle or more realistic, take all this content and produce this kind of answer from a long context that sort of better assesses what it is people really want to do with long context. Which is not just, you know, can you tell me the product number for this particular thing?Shawn Wang [00:14:31]: Yeah, it's retrieval. It's retrieval within machine learning. It's interesting because I think the more meta level I'm trying to operate at here is you have a benchmark. You're like, okay, I see the architectural thing I need to do in order to go fix that. But should you do it? Because sometimes that's an inductive bias, basically. It's what Jason Wei, who used to work at Google, would say. Exactly the kind of thing. Yeah, you're going to win. Short term. Longer term, I don't know if that's going to scale. You might have to undo that.Jeff Dean [00:15:01]: I mean, I like to sort of not focus on exactly what solution we're going to derive, but what capability would you want? And I think we're very convinced that, you know, long context is useful, but it's way too short today. Right? Like, I think what you would really want is, can I attend to the internet while I answer my question? Right? But that's not going to happen. I think that's going to be solved by purely scaling the existing solutions, which are quadratic. So a million tokens kind of pushes what you can do. You're not going to do that to a trillion tokens, let alone, you know, a billion tokens, let alone a trillion. But I think if you could give the illusion that you can attend to trillions of tokens, that would be amazing. You'd find all kinds of uses for that. You would have attend to the internet. You could attend to the pixels of YouTube and the sort of deeper representations that we can find. You could attend to the form for a single video, but across many videos, you know, on a personal Gemini level, you could attend to all of your personal state with your permission. So like your emails, your photos, your docs, your plane tickets you have. I think that would be really, really useful. And the question is, how do you get algorithmic improvements and system level improvements that get you to something where you actually can attend to trillions of tokens? Right. In a meaningful way. Yeah.Shawn Wang [00:16:26]: But by the way, I think I did some math and it's like, if you spoke all day, every day for eight hours a day, you only generate a maximum of like a hundred K tokens, which like very comfortably fits.Jeff Dean [00:16:38]: Right. But if you then say, okay, I want to be able to understand everything people are putting on videos.Shawn Wang [00:16:46]: Well, also, I think that the classic example is you start going beyond language into like proteins and whatever else is extremely information dense. Yeah. Yeah.Jeff Dean [00:16:55]: I mean, I think one of the things about Gemini's multimodal aspects is we've always wanted it to be multimodal from the start. And so, you know, that sometimes to people means text and images and video sort of human-like and audio, audio, human-like modalities. But I think it's also really useful to have Gemini know about non-human modalities. Yeah. Like LIDAR sensor data from. Yes. Say, Waymo vehicles or. Like robots or, you know, various kinds of health modalities, x-rays and MRIs and imaging and genomics information. And I think there's probably hundreds of modalities of data where you'd like the model to be able to at least be exposed to the fact that this is an interesting modality and has certain meaning in the world. Where even if you haven't trained on all the LIDAR data or MRI data, you could have, because maybe that's not, you know, it doesn't make sense in terms of trade-offs of. You know, what you include in your main pre-training data mix, at least including a little bit of it is actually quite useful. Yeah. Because it sort of tempts the model that this is a thing.Shawn Wang [00:18:04]: Yeah. Do you believe, I mean, since we're on this topic and something I just get to ask you all the questions I always wanted to ask, which is fantastic. Like, are there some king modalities, like modalities that supersede all the other modalities? So a simple example was Vision can, on a pixel level, encode text. And DeepSeq had this DeepSeq CR paper that did that. Vision. And Vision has also been shown to maybe incorporate audio because you can do audio spectrograms and that's, that's also like a Vision capable thing. Like, so, so maybe Vision is just the king modality and like. Yeah.Jeff Dean [00:18:36]: I mean, Vision and Motion are quite important things, right? Motion. Well, like video as opposed to static images, because I mean, there's a reason evolution has evolved eyes like 23 independent ways, because it's such a useful capability for sensing the world around you, which is really what we want these models to be. So I think the only thing that we can be able to do is interpret the things we're seeing or the things we're paying attention to and then help us in using that information to do things. Yeah.Shawn Wang [00:19:05]: I think motion, you know, I still want to shout out, I think Gemini, still the only native video understanding model that's out there. So I use it for YouTube all the time. Nice.Jeff Dean [00:19:15]: Yeah. Yeah. I mean, it's actually, I think people kind of are not necessarily aware of what the Gemini models can actually do. Yeah. Like I have an example I've used in one of my talks. It had like, it was like a YouTube highlight video of 18 memorable sports moments across the last 20 years or something. So it has like Michael Jordan hitting some jump shot at the end of the finals and, you know, some soccer goals and things like that. And you can literally just give it the video and say, can you please make me a table of what all these different events are? What when the date is when they happened? And a short description. And so you get like now an 18 row table of that information extracted from the video, which is, you know, not something most people think of as like a turn video into sequel like table.Alessio Fanelli [00:20:11]: Has there been any discussion inside of Google of like, you mentioned tending to the whole internet, right? Google, it's almost built because a human cannot tend to the whole internet and you need some sort of ranking to find what you need. Yep. That ranking is like much different for an LLM because you can expect a person to look at maybe the first five, six links in a Google search versus for an LLM. Should you expect to have 20 links that are highly relevant? Like how do you internally figure out, you know, how do we build the AI mode that is like maybe like much broader search and span versus like the more human one? Yeah.Jeff Dean [00:20:47]: I mean, I think even pre-language model based work, you know, our ranking systems would be built to start. I mean, I think even pre-language model based work, you know, our ranking systems would be built to start. With a giant number of web pages in our index, many of them are not relevant. So you identify a subset of them that are relevant with very lightweight kinds of methods. You know, you're down to like 30,000 documents or something. And then you gradually refine that to apply more and more sophisticated algorithms and more and more sophisticated sort of signals of various kinds in order to get down to ultimately what you show, which is, you know, the final 10 results or, you know, 10 results plus. Other kinds of information. And I think an LLM based system is not going to be that dissimilar, right? You're going to attend to trillions of tokens, but you're going to want to identify, you know, what are the 30,000 ish documents that are with the, you know, maybe 30 million interesting tokens. And then how do you go from that into what are the 117 documents I really should be paying attention to in order to carry out the tasks that the user has asked? And I think, you know, you can imagine systems where you have, you know, a lot of highly parallel processing to identify those initial 30,000 candidates, maybe with very lightweight kinds of models. Then you have some system that sort of helps you narrow down from 30,000 to the 117 with maybe a little bit more sophisticated model or set of models. And then maybe the final model is the thing that looks. So the 117 things that might be your most capable model. So I think it has to, it's going to be some system like that, that is really enables you to give the illusion of attending to trillions of tokens. Sort of the way Google search gives you, you know, not the illusion, but you are searching the internet, but you're finding, you know, a very small subset of things that are, that are relevant.Shawn Wang [00:22:47]: Yeah. I often tell a lot of people that are not steeped in like Google search history that, well, you know, like Bert was. Like he was like basically immediately inside of Google search and that improves results a lot, right? Like I don't, I don't have any numbers off the top of my head, but like, I'm sure you guys, that's obviously the most important numbers to Google. Yeah.Jeff Dean [00:23:08]: I mean, I think going to an LLM based representation of text and words and so on enables you to get out of the explicit hard notion of, of particular words having to be on the page, but really getting at the notion of this topic of this page or this page. Paragraph is highly relevant to this query. Yeah.Shawn Wang [00:23:28]: I don't think people understand how much LLMs have taken over all these very high traffic system, very high traffic. Yeah. Like it's Google, it's YouTube. YouTube has this like semantics ID thing where it's just like every token or every item in the vocab is a YouTube video or something that predicts the video using a code book, which is absurd to me for YouTube size.Jeff Dean [00:23:50]: And then most recently GROK also for, for XAI, which is like, yeah. I mean, I'll call out even before LLMs were used extensively in search, we put a lot of emphasis on softening the notion of what the user actually entered into the query.Shawn Wang [00:24:06]: So do you have like a history of like, what's the progression? Oh yeah.Jeff Dean [00:24:09]: I mean, I actually gave a talk in, uh, I guess, uh, web search and data mining conference in 2009, uh, where we never actually published any papers about the origins of Google search, uh, sort of, but we went through sort of four or five or six. generations, four or five or six generations of, uh, redesigning of the search and retrieval system, uh, from about 1999 through 2004 or five. And that talk is really about that evolution. And one of the things that really happened in 2001 was we were sort of working to scale the system in multiple dimensions. So one is we wanted to make our index bigger, so we could retrieve from a larger index, which always helps your quality in general. Uh, because if you don't have the page in your index, you're going to not do well. Um, and then we also needed to scale our capacity because we were, our traffic was growing quite extensively. Um, and so we had, you know, a sharded system where you have more and more shards as the index grows, you have like 30 shards. And then if you want to double the index size, you make 60 shards so that you can bound the latency by which you respond for any particular user query. Um, and then as traffic grows, you add, you add more and more replicas of each of those. And so we eventually did the math that realized that in a data center where we had say 60 shards and, um, you know, 20 copies of each shard, we now had 1200 machines, uh, with disks. And we did the math and we're like, Hey, one copy of that index would actually fit in memory across 1200 machines. So in 2001, we introduced, uh, we put our entire index in memory and what that enabled from a quality perspective was amazing. Um, and so we had more and more replicas of each of those. Before you had to be really careful about, you know, how many different terms you looked at for a query, because every one of them would involve a disk seek on every one of the 60 shards. And so you, as you make your index bigger, that becomes even more inefficient. But once you have the whole index in memory, it's totally fine to have 50 terms you throw into the query from the user's original three or four word query, because now you can add synonyms like restaurant and restaurants and cafe and, uh, you know, things like that. Uh, bistro and all these things. And you can suddenly start, uh, sort of really, uh, getting at the meaning of the word as opposed to the exact semantic form the user typed in. And that was, you know, 2001, very much pre LLM, but really it was about softening the, the strict definition of what the user typed in order to get at the meaning.Alessio Fanelli [00:26:47]: What are like principles that you use to like design the systems, especially when you have, I mean, in 2001, the internet is like. Doubling, tripling every year in size is not like, uh, you know, and I think today you kind of see that with LLMs too, where like every year the jumps in size and like capabilities are just so big. Are there just any, you know, principles that you use to like, think about this? Yeah.Jeff Dean [00:27:08]: I mean, I think, uh, you know, first, whenever you're designing a system, you want to understand what are the sort of design parameters that are going to be most important in designing that, you know? So, you know, how many queries per second do you need to handle? How big is the internet? How big is the index you need to handle? How much data do you need to keep for every document in the index? How are you going to look at it when you retrieve things? Um, what happens if traffic were to double or triple, you know, will that system work well? And I think a good design principle is you're going to want to design a system so that the most important characteristics could scale by like factors of five or 10, but probably not beyond that because often what happens is if you design a system for X. And something suddenly becomes a hundred X, that would enable a very different point in the design space that would not make sense at X. But all of a sudden at a hundred X makes total sense. So like going from a disk space index to a in memory index makes a lot of sense once you have enough traffic, because now you have enough replicas of the sort of state on disk that those machines now actually can hold, uh, you know, a full copy of the, uh, index and memory. Yeah. And that all of a sudden enabled. A completely different design that wouldn't have been practical before. Yeah. Um, so I'm, I'm a big fan of thinking through designs in your head, just kind of playing with the design space a little before you actually do a lot of writing of code. But, you know, as you said, in the early days of Google, we were growing the index, uh, quite extensively. We were growing the update rate of the index. So the update rate actually is the parameter that changed the most. Surprising. So it used to be once a month.Shawn Wang [00:28:55]: Yeah.Jeff Dean [00:28:56]: And then we went to a system that could update any particular page in like sub one minute. Okay.Shawn Wang [00:29:02]: Yeah. Because this is a competitive advantage, right?Jeff Dean [00:29:04]: Because all of a sudden news related queries, you know, if you're, if you've got last month's news index, it's not actually that useful for.Shawn Wang [00:29:11]: News is a special beast. Was there any, like you could have split it onto a separate system.Jeff Dean [00:29:15]: Well, we did. We launched a Google news product, but you also want news related queries that people type into the main index to also be sort of updated.Shawn Wang [00:29:23]: So, yeah, it's interesting. And then you have to like classify whether the page is, you have to decide which pages should be updated and what frequency. Oh yeah.Jeff Dean [00:29:30]: There's a whole like, uh, system behind the scenes that's trying to decide update rates and importance of the pages. So even if the update rate seems low, you might still want to recrawl important pages quite often because, uh, the likelihood they change might be low, but the value of having updated is high.Shawn Wang [00:29:50]: Yeah, yeah, yeah, yeah. Uh, well, you know, yeah. This, uh, you know, mention of latency and, and saving things to this reminds me of one of your classics, which I have to bring up, which is latency numbers. Every programmer should know, uh, was there a, was it just a, just a general story behind that? Did you like just write it down?Jeff Dean [00:30:06]: I mean, this has like sort of eight or 10 different kinds of metrics that are like, how long does a cache mistake? How long does branch mispredict take? How long does a reference domain memory take? How long does it take to send, you know, a packet from the U S to the Netherlands or something? Um,Shawn Wang [00:30:21]: why Netherlands, by the way, or is it, is that because of Chrome?Jeff Dean [00:30:25]: Uh, we had a data center in the Netherlands, um, so, I mean, I think this gets to the point of being able to do the back of the envelope calculations. So these are sort of the raw ingredients of those, and you can use them to say, okay, well, if I need to design a system to do image search and thumb nailing or something of the result page, you know, how, what I do that I could pre-compute the image thumbnails. I could like. Try to thumbnail them on the fly from the larger images. What would that do? How much dis bandwidth than I need? How many des seeks would I do? Um, and you can sort of actually do thought experiments in, you know, 30 seconds or a minute with the sort of, uh, basic, uh, basic numbers at your fingertips. Uh, and then as you sort of build software using higher level libraries, you kind of want to develop the same intuitions for how long does it take to, you know, look up something in this particular kind of.Shawn Wang [00:31:21]: I'll see you next time.Shawn Wang [00:31:51]: Which is a simple byte conversion. That's nothing interesting. I wonder if you have any, if you were to update your...Jeff Dean [00:31:58]: I mean, I think it's really good to think about calculations you're doing in a model, either for training or inference.Jeff Dean [00:32:09]: Often a good way to view that is how much state will you need to bring in from memory, either like on-chip SRAM or HBM from the accelerator. Attached memory or DRAM or over the network. And then how expensive is that data motion relative to the cost of, say, an actual multiply in the matrix multiply unit? And that cost is actually really, really low, right? Because it's order, depending on your precision, I think it's like sub one picodule.Shawn Wang [00:32:50]: Oh, okay. You measure it by energy. Yeah. Yeah.Jeff Dean [00:32:52]: Yeah. I mean, it's all going to be about energy and how do you make the most energy efficient system. And then moving data from the SRAM on the other side of the chip, not even off the off chip, but on the other side of the same chip can be, you know, a thousand picodules. Oh, yeah. And so all of a sudden, this is why your accelerators require batching. Because if you move, like, say, the parameter of a model from SRAM on the, on the chip into the multiplier unit, that's going to cost you a thousand picodules. So you better make use of that, that thing that you moved many, many times with. So that's where the batch dimension comes in. Because all of a sudden, you know, if you have a batch of 256 or something, that's not so bad. But if you have a batch of one, that's really not good.Shawn Wang [00:33:40]: Yeah. Yeah. Right.Jeff Dean [00:33:41]: Because then you paid a thousand picodules in order to do your one picodule multiply.Shawn Wang [00:33:46]: I have never heard an energy-based analysis of batching.Jeff Dean [00:33:50]: Yeah. I mean, that's why people batch. Yeah. Ideally, you'd like to use batch size one because the latency would be great.Shawn Wang [00:33:56]: The best latency.Jeff Dean [00:33:56]: But the energy cost and the compute cost inefficiency that you get is quite large. So, yeah.Shawn Wang [00:34:04]: Is there a similar trick like, like, like you did with, you know, putting everything in memory? Like, you know, I think obviously NVIDIA has caused a lot of waves with betting very hard on SRAM with Grok. I wonder if, like, that's something that you already saw with, with the TPUs, right? Like that, that you had to. Uh, to serve at your scale, uh, you probably sort of saw that coming. Like what, what, what hardware, uh, innovations or insights were formed because of what you're seeing there?Jeff Dean [00:34:33]: Yeah. I mean, I think, you know, TPUs have this nice, uh, sort of regular structure of 2D or 3D meshes with a bunch of chips connected. Yeah. And each one of those has HBM attached. Um, I think for serving some kinds of models, uh, you know, you, you pay a lot higher cost. Uh, and time latency, um, bringing things in from HBM than you do bringing them in from, uh, SRAM on the chip. So if you have a small enough model, you can actually do model parallelism, spread it out over lots of chips and you actually get quite good throughput improvements and latency improvements from doing that. And so you're now sort of striping your smallish scale model over say 16 or 64 chips. Uh, but as if you do that and it all fits in. In SRAM, uh, that can be a big win. So yeah, that's not a surprise, but it is a good technique.Alessio Fanelli [00:35:27]: Yeah. What about the TPU design? Like how much do you decide where the improvements have to go? So like, this is like a good example of like, is there a way to bring the thousand picojoules down to 50? Like, is it worth designing a new chip to do that? The extreme is like when people say, oh, you should burn the model on the ASIC and that's kind of like the most extreme thing. How much of it? Is it worth doing an hardware when things change so quickly? Like what was the internal discussion? Yeah.Jeff Dean [00:35:57]: I mean, we, we have a lot of interaction between say the TPU chip design architecture team and the sort of higher level modeling, uh, experts, because you really want to take advantage of being able to co-design what should future TPUs look like based on where we think the sort of ML research puck is going, uh, in some sense, because, uh, you know, as a hardware designer for ML and in particular, you're trying to design a chip starting today and that design might take two years before it even lands in a data center. And then it has to sort of be a reasonable lifetime of the chip to take you three, four or five years. So you're trying to predict two to six years out where, what ML computations will people want to run two to six years out in a very fast changing field. And so having people with interest. Interesting ML research ideas of things we think will start to work in that timeframe or will be more important in that timeframe, uh, really enables us to then get, you know, interesting hardware features put into, you know, TPU N plus two, where TPU N is what we have today.Shawn Wang [00:37:10]: Oh, the cycle time is plus two.Jeff Dean [00:37:12]: Roughly. Wow. Because, uh, I mean, sometimes you can squeeze some changes into N plus one, but, you know, bigger changes are going to require the chip. Yeah. Design be earlier in its lifetime design process. Um, so whenever we can do that, it's generally good. And sometimes you can put in speculative features that maybe won't cost you much chip area, but if it works out, it would make something, you know, 10 times as fast. And if it doesn't work out, well, you burned a little bit of tiny amount of your chip area on that thing, but it's not that big a deal. Uh, sometimes it's a very big change and we want to be pretty sure this is going to work out. So we'll do like lots of carefulness. Uh, ML experimentation to show us, uh, this is actually the, the way we want to go. Yeah.Alessio Fanelli [00:37:58]: Is there a reverse of like, we already committed to this chip design so we can not take the model architecture that way because it doesn't quite fit?Jeff Dean [00:38:06]: Yeah. I mean, you, you definitely have things where you're going to adapt what the model architecture looks like so that they're efficient on the chips that you're going to have for both training and inference of that, of that, uh, generation of model. So I think it kind of goes both ways. Um, you know, sometimes you can take advantage of, you know, lower precision things that are coming in a future generation. So you can, might train it at that lower precision, even if the current generation doesn't quite do that. Mm.Shawn Wang [00:38:40]: Yeah. How low can we go in precision?Jeff Dean [00:38:43]: Because people are saying like ternary is like, uh, yeah, I mean, I'm a big fan of very low precision because I think that gets, that saves you a tremendous amount of time. Right. Because it's picojoules per bit that you're transferring and reducing the number of bits is a really good way to, to reduce that. Um, you know, I think people have gotten a lot of luck, uh, mileage out of having very low bit precision things, but then having scaling factors that apply to a whole bunch of, uh, those, those weights. Scaling. How does it, how does it, okay.Shawn Wang [00:39:15]: Interesting. You, so low, low precision, but scaled up weights. Yeah. Huh. Yeah. Never considered that. Yeah. Interesting. Uh, w w while we're on this topic, you know, I think there's a lot of, um, uh, this, the concept of precision at all is weird when we're sampling, you know, uh, we just, at the end of this, we're going to have all these like chips that I'll do like very good math. And then we're just going to throw a random number generator at the start. So, I mean, there's a movement towards, uh, energy based, uh, models and processors. I'm just curious if you've, obviously you've thought about it, but like, what's your commentary?Jeff Dean [00:39:50]: Yeah. I mean, I think. There's a bunch of interesting trends though. Energy based models is one, you know, diffusion based models, which don't sort of sequentially decode tokens is another, um, you know, speculative decoding is a way that you can get sort of an equivalent, very small.Shawn Wang [00:40:06]: Draft.Jeff Dean [00:40:07]: Batch factor, uh, for like you predict eight tokens out and that enables you to sort of increase the effective batch size of what you're doing by a factor of eight, even, and then you maybe accept five or six of those tokens. So you get. A five, a five X improvement in the amortization of moving weights, uh, into the multipliers to do the prediction for the, the tokens. So these are all really good techniques and I think it's really good to look at them from the lens of, uh, energy, real energy, not energy based models, um, and, and also latency and throughput, right? If you look at things from that lens, that sort of guides you to. Two solutions that are gonna be, uh, you know, better from, uh, you know, being able to serve larger models or, you know, equivalent size models more cheaply and with lower latency.Shawn Wang [00:41:03]: Yeah. Well, I think, I think I, um, it's appealing intellectually, uh, haven't seen it like really hit the mainstream, but, um, I do think that, uh, there's some poetry in the sense that, uh, you know, we don't have to do, uh, a lot of shenanigans if like we fundamentally. Design it into the hardware. Yeah, yeah.Jeff Dean [00:41:23]: I mean, I think there's still a, there's also sort of the more exotic things like analog based, uh, uh, computing substrates as opposed to digital ones. Uh, I'm, you know, I think those are super interesting cause they can be potentially low power. Uh, but I think you often end up wanting to interface that with digital systems and you end up losing a lot of the power advantages in the digital to analog and analog to digital conversions. You end up doing, uh, at the sort of boundaries. And periphery of that system. Um, I still think there's a tremendous distance we can go from where we are today in terms of energy efficiency with sort of, uh, much better and specialized hardware for the models we care about.Shawn Wang [00:42:05]: Yeah.Alessio Fanelli [00:42:06]: Um, any other interesting research ideas that you've seen, or like maybe things that you cannot pursue a Google that you would be interested in seeing researchers take a step at, I guess you have a lot of researchers. Yeah, I guess you have enough, but our, our research.Jeff Dean [00:42:21]: Our research portfolio is pretty broad. I would say, um, I mean, I think, uh, in terms of research directions, there's a whole bunch of, uh, you know, open problems and how do you make these models reliable and able to do much longer, kind of, uh, more complex tasks that have lots of subtasks. How do you orchestrate, you know, maybe one model that's using other models as tools in order to sort of build, uh, things that can accomplish, uh, you know, much more. Yeah. Significant pieces of work, uh, collectively, then you would ask a single model to do. Um, so that's super interesting. How do you get more verifiable, uh, you know, how do you get RL to work for non-verifiable domains? I think it's a pretty interesting open problem because I think that would broaden out the capabilities of the models, the improvements that you're seeing in both math and coding. Uh, if we could apply those to other less verifiable domains, because we've come up with RL techniques that actually enable us to do that. Uh, effectively, that would, that would really make the models improve quite a lot. I think.Alessio Fanelli [00:43:26]: I'm curious, like when we had Noam Brown on the podcast, he said, um, they already proved you can do it with deep research. Um, you kind of have it with AI mode in a way it's not verifiable. I'm curious if there's any thread that you think is interesting there. Like what is it? Both are like information retrieval of JSON. So I wonder if it's like the retrieval is like the verifiable part. That you can score or what are like, yeah, yeah. How, how would you model that, that problem?Jeff Dean [00:43:55]: Yeah. I mean, I think there are ways of having other models that can evaluate the results of what a first model did, maybe even retrieving. Can you have another model that says, is this things, are these things you retrieved relevant? Or can you rate these 2000 things you retrieved to assess which ones are the 50 most relevant or something? Um, I think those kinds of techniques are actually quite effective. Sometimes I can even be the same model, just prompted differently to be a, you know, a critic as opposed to a, uh, actual retrieval system. Yeah.Shawn Wang [00:44:28]: Um, I do think like there, there is that, that weird cliff where like, it feels like we've done the easy stuff and then now it's, but it always feels like that every year. It's like, oh, like we know, we know, and the next part is super hard and nobody's figured it out. And, uh, exactly with this RLVR thing where like everyone's talking about, well, okay, how do we. the next stage of the non-verifiable stuff. And everyone's like, I don't know, you know, Ellen judge.Jeff Dean [00:44:56]: I mean, I feel like the nice thing about this field is there's lots and lots of smart people thinking about creative solutions to some of the problems that we all see. Uh, because I think everyone sort of sees that the models, you know, are great at some things and they fall down around the edges of those things and, and are not as capable as we'd like in those areas. And then coming up with good techniques and trying those. And seeing which ones actually make a difference is sort of what the whole research aspect of this field is, is pushing forward. And I think that's why it's super interesting. You know, if you think about two years ago, we were struggling with GSM, eight K problems, right? Like, you know, Fred has two rabbits. He gets three more rabbits. How many rabbits does he have? That's a pretty far cry from the kinds of mathematics that the models can, and now you're doing IMO and Erdos problems in pure language. Yeah. Yeah. Pure language. So that is a really, really amazing jump in capabilities in, you know, in a year and a half or something. And I think, um, for other areas, it'd be great if we could make that kind of leap. Uh, and you know, we don't exactly see how to do it for some, some areas, but we do see it for some other areas and we're going to work hard on making that better. Yeah.Shawn Wang [00:46:13]: Yeah.Alessio Fanelli [00:46:14]: Like YouTube thumbnail generation. That would be very helpful. We need that. That would be AGI. We need that.Shawn Wang [00:46:20]: That would be. As far as content creators go.Jeff Dean [00:46:22]: I guess I'm not a YouTube creator, so I don't care that much about that problem, but I guess, uh, many people do.Shawn Wang [00:46:27]: It does. Yeah. It doesn't, it doesn't matter. People do judge books by their covers as it turns out. Um, uh, just to draw a bit on the IMO goal. Um, I'm still not over the fact that a year ago we had alpha proof and alpha geometry and all those things. And then this year we were like, screw that we'll just chuck it into Gemini. Yeah. What's your reflection? Like, I think this, this question about. Like the merger of like symbolic systems and like, and, and LMS, uh, was a very much core belief. And then somewhere along the line, people would just said, Nope, we'll just all do it in the LLM.Jeff Dean [00:47:02]: Yeah. I mean, I think it makes a lot of sense to me because, you know, humans manipulate symbols, but we probably don't have like a symbolic representation in our heads. Right. We have some distributed representation that is neural net, like in some way of lots of different neurons. And activation patterns firing when we see certain things and that enables us to reason and plan and, you know, do chains of thought and, you know, roll them back now that, that approach for solving the problem doesn't seem like it's going to work. I'm going to try this one. And, you know, in a lot of ways we're emulating what we intuitively think, uh, is happening inside real brains in neural net based models. So it never made sense to me to have like completely separate. Uh, discrete, uh, symbolic things, and then a completely different way of, of, uh, you know, thinking about those things.Shawn Wang [00:47:59]: Interesting. Yeah. Uh, I mean, it's maybe seems obvious to you, but it wasn't obvious to me a year ago. Yeah.Jeff Dean [00:48:06]: I mean, I do think like that IMO with, you know, translating to lean and using lean and then the next year and also a specialized geometry model. And then this year switching to a single unified model. That is roughly the production model with a little bit more inference budget, uh, is actually, you know, quite good because it shows you that the capabilities of that general model have improved dramatically and, and now you don't need the specialized model. This is actually sort of very similar to the 2013 to 16 era of machine learning, right? Like it used to be, people would train separate models for lots of different, each different problem, right? I have, I want to recognize street signs and something. So I train a street sign. Recognition recognition model, or I want to, you know, decode speech recognition. I have a speech model, right? I think now the era of unified models that do everything is really upon us. And the question is how well do those models generalize to new things they've never been asked to do and they're getting better and better.Shawn Wang [00:49:10]: And you don't need domain experts. Like one of my, uh, so I interviewed ETA who was on, who was on that team. Uh, and he was like, yeah, I, I don't know how they work. I don't know where the IMO competition was held. I don't know the rules of it. I just trained the models, the training models. Yeah. Yeah. And it's kind of interesting that like people with these, this like universal skill set of just like machine learning, you just give them data and give them enough compute and they can kind of tackle any task, which is the bitter lesson, I guess. I don't know. Yeah.Jeff Dean [00:49:39]: I mean, I think, uh, general models, uh, will win out over specialized ones in most cases.Shawn Wang [00:49:45]: Uh, so I want to push there a bit. I think there's one hole here, which is like, uh. There's this concept of like, uh, maybe capacity of a model, like abstractly a model can only contain the number of bits that it has. And, uh, and so it, you know, God knows like Gemini pro is like one to 10 trillion parameters. We don't know, but, uh, the Gemma models, for example, right? Like a lot of people want like the open source local models that are like that, that, that, and, and, uh, they have some knowledge, which is not necessary, right? Like they can't know everything like, like you have the. The luxury of you have the big model and big model should be able to capable of everything. But like when, when you're distilling and you're going down to the small models, you know, you're actually memorizing things that are not useful. Yeah. And so like, how do we, I guess, do we want to extract that? Can we, can we divorce knowledge from reasoning, you know?Jeff Dean [00:50:38]: Yeah. I mean, I think you do want the model to be most effective at reasoning if it can retrieve things, right? Because having the model devote precious parameter space. To remembering obscure facts that could be looked up is actually not the best use of that parameter space, right? Like you might prefer something that is more generally useful in more settings than this obscure fact that it has. Um, so I think that's always attention at the same time. You also don't want your model to be kind of completely detached from, you know, knowing stuff about the world, right? Like it's probably useful to know how long the golden gate be. Bridges just as a general sense of like how long are bridges, right? And, uh, it should have that kind of knowledge. It maybe doesn't need to know how long some teeny little bridge in some other more obscure part of the world is, but, uh, it does help it to have a fair bit of world knowledge and the bigger your model is, the more you can have. Uh, but I do think combining retrieval with sort of reasoning and making the model really good at doing multiple stages of retrieval. Yeah.Shawn Wang [00:51:49]: And reasoning through the intermediate retrieval results is going to be a, a pretty effective way of making the model seem much more capable, because if you think about, say, a personal Gemini, yeah, right?Jeff Dean [00:52:01]: Like we're not going to train Gemini on my email. Probably we'd rather have a single model that, uh, we can then use and use being able to retrieve from my email as a tool and have the model reason about it and retrieve from my photos or whatever, uh, and then make use of that and have multiple. Um, you know, uh, stages of interaction. that makes sense.Alessio Fanelli [00:52:24]: Do you think the vertical models are like, uh, interesting pursuit? Like when people are like, oh, we're building the best healthcare LLM, we're building the best law LLM, are those kind of like short-term stopgaps or?Jeff Dean [00:52:37]: No, I mean, I think, I think vertical models are interesting. Like you want them to start from a pretty good base model, but then you can sort of, uh, sort of viewing them, view them as enriching the data. Data distribution for that particular vertical domain for healthcare, say, um, we're probably not going to train or for say robotics. We're probably not going to train Gemini on all possible robotics data. We, you could train it on because we want it to have a balanced set of capabilities. Um, so we'll expose it to some robotics data, but if you're trying to build a really, really good robotics model, you're going to want to start with that and then train it on more robotics data. And then maybe that would. It's multilingual translation capability, but improve its robotics capabilities. And we're always making these kind of, uh, you know, trade-offs in the data mix that we train the base Gemini models on. You know, we'd love to include data from 200 more languages and as much data as we have for those languages, but that's going to displace some other capabilities of the model. It won't be as good at, um, you know, Pearl programming, you know, it'll still be good at Python programming. Cause we'll include it. Enough. Of that, but there's other long tail computer languages or coding capabilities that it may suffer on or multi, uh, multimodal reasoning capabilities may suffer. Cause we didn't get to expose it to as much data there, but it's really good at multilingual things. So I, I think some combination of specialized models, maybe more modular models. So it'd be nice to have the capability to have those 200 languages, plus this awesome robotics model, plus this awesome healthcare, uh, module that all can be knitted together to work in concert and called upon in different circumstances. Right? Like if I have a health related thing, then it should enable using this health module in conjunction with the main base model to be even better at those kinds of things. Yeah.Shawn Wang [00:54:36]: Installable knowledge. Yeah.Jeff Dean [00:54:37]: Right.Shawn Wang [00:54:38]: Just download as a, as a package.Jeff Dean [00:54:39]: And some of that installable stuff can come from retrieval, but some of it probably should come from preloaded training on, you know, uh, a hundred billion tokens or a trillion tokens of health data. Yeah.Shawn Wang [00:54:51]: And for listeners, I think, uh, I will highlight the Gemma three end paper where they, there was a little bit of that, I think. Yeah.Alessio Fanelli [00:54:56]: Yeah. I guess the question is like, how many billions of tokens do you need to outpace the frontier model improvements? You know, it's like, if I have to make this model better healthcare and the main. Gemini model is still improving. Do I need 50 billion tokens? Can I do it with a hundred, if I need a trillion healthcare tokens, it's like, they're probably not out there that you don't have, you know, I think that's really like the.Jeff Dean [00:55:21]: Well, I mean, I think healthcare is a particularly challenging domain, so there's a lot of healthcare data that, you know, we don't have access to appropriately, but there's a lot of, you know, uh, healthcare organizations that want to train models on their own data. That is not public healthcare data, uh, not public health. But public healthcare data. Um, so I think there are opportunities there to say, partner with a large healthcare organization and train models for their use that are going to be, you know, more bespoke, but probably, uh, might be better than a general model trained on say, public data. Yeah.Shawn Wang [00:55:58]: Yeah. I, I believe, uh, by the way, also this is like somewhat related to the language conversation. Uh, I think one of your, your favorite examples was you can put a low resource language in the context and it just learns. Yeah.Jeff Dean [00:56:09]: Oh, yeah, I think the example we used was Calamon, which is truly low resource because it's only spoken by, I think 120 people in the world and there's no written text.Shawn Wang [00:56:20]: So, yeah. So you can just do it that way. Just put it in the context. Yeah. Yeah. But I think your whole data set in the context, right.Jeff Dean [00:56:27]: If you, if you take a language like, uh, you know, Somali or something, there is a fair bit of Somali text in the world that, uh, or Ethiopian Amharic or something, um, you know, we probably. Yeah. Are not putting all the data from those languages into the Gemini based training. We put some of it, but if you put more of it, you'll improve the capabilities of those models.Shawn Wang [00:56:49]: Yeah.Jeff Dean [00:56:49]:

What does it take to lead with clarity, compassion, and courage through the stormy waters of educational transformation?In this inspiring episode of Voices for Excellence, Dr. Michael Conner welcomes Dr. Lloyd Jackson, Superintendent of the Texarkana Arkansas School District and a proud alumnus of the very community he now leads. Known for his signature calm and purposeful leadership, Dr. Jackson joins the Black Excellence Series to share the deeply human work of transforming systems, with his trademark humility, clarity of vision, and unwavering belief that “a change is gonna come.”Dr. Jackson walks us through the intentional steps he's taken to evolve his district from a collection of schools into a coherent, student-centered system. With laser focus on three districtwide priorities, literacy, behavior, and chronic absenteeism, he shares how collective action, data-informed leadership, and outcome-driven partnerships can create conditions where every student thrives. From restructuring assessment practices to leveraging AI as a force multiplier, Dr. Jackson models what it means to be a lead learner committed to the future of education.What you'll learn:Bold simplicity: How three focused priorities, literacy, attendance, behavior, transformed culture, coherence, and performance.Data as a conversation: How moving from data compliance to data literacy empowers teachers and drives change.Mission-aligned partnerships: The why and how of building community alliances that deliver real outcomes for students. AI with purpose: How artificial intelligence is being ethically integrated to reduce workload, increase instructional quality, and drive innovation.Student-centered systems: Why human relationships must remain at the center of tech-enabled education. Vision for 2080: How today's kindergarteners will retire mid-century, and what we must do now to prepare them for that world.Dr. Jackson's leadership isn't only about strategy, it's about soul. From community-rooted reforms to outcome-based contracts and personalized learning systems, his vision challenges all of us to lead with dignity, data, and deep purpose.This episode is a masterclass in how to build the systems our future demands, boldly, equitably, and with excellence.Subscribe and share to continue driving the future of education for all.

What hardens a heart? This week the Rabbi Blum Pastor Kahlil Carmichael shows us the causes of a hardened heart and how to be pure in heart. We begin in Exodus and Proverbs 4 with the message, “A Unified Service with Rabbi Blum and Pastor Khalil Carmichael.”Live Well, your spiritual family, gathers every Sunday at 11 a.m. at 51 Church Street, Robbinsville, Windsor, NJ. We look forward to welcoming you and sharing this faith journey with you.Thank you for givingpushpay.com/g/itiswellchurchVisit our website at livewellchurch.orgFollow us on Facebook @pastorkahlilFind us on Instagram @livewellwithpastorkahlil

Last time we spoke about The Battle of Suixian–Zaoyang-Shatow. Following the brutal 1938 capture of Wuhan, Japanese forces aimed to solidify their hold by launching an offensive against Chinese troops in the 5th War Zone, a rugged natural fortress in northern Hubei and southern Henan. Under General Yasuji Okamura, the 11th Army deployed three divisions and cavalry in a pincer assault starting May 1, 1939, targeting Suixian and Zaoyang to crush Nationalist resistance and secure flanks. Chinese commander Li Zongren, leveraging terrain like the Dabie and Tongbai Mountains, orchestrated defenses with over 200,000 troops, including Tang Enbo's 31st Army Group. By May 23, they recaptured Suixian and Zaoyang, forcing a Japanese withdrawal with heavy losses, over 13,000 Japanese casualties versus 25,000 Chinese, restoring pre-battle lines. Shifting south, Japan targeted Shantou in Guangdong to sever supply lines from Hong Kong. In a massive June 21 amphibious assault, the 21st Army overwhelmed thin Chinese defenses, capturing the port and Chao'an despite guerrilla resistance led by Zhang Fakui. Though losses mounted, Japan tightened its blockade, straining China's war effort amid ongoing attrition.   #188 From Changkufeng to Nomonhan Welcome to the Fall and Rise of China Podcast, I am your dutiful host Craig Watson. But, before we start I want to also remind you this podcast is only made possible through the efforts of Kings and Generals over at Youtube. Perhaps you want to learn more about the history of Asia? Kings and Generals have an assortment of episodes on history of asia and much more  so go give them a look over on Youtube. So please subscribe to Kings and Generals over at Youtube and to continue helping us produce this content please check out www.patreon.com/kingsandgenerals. If you are still hungry for some more history related content, over on my channel, the Pacific War Channel where I cover the history of China and Japan from the 19th century until the end of the Pacific War. Well hello again, and yes you all have probably guessed we are taking another detour. Do not worry I hope to shorten this one a bit more so than what became a sort of mini series on the battle of Changkufeng or Battle of Lake Khasan. What we are about to jump into is known in the west as the battle of khalkin Gol, by the Japanese the Nomohan incident. But first I need to sort of set the table up so to say. So back on August 10th, 1938 the Litvinov-Shigemitsu agreement established a joint border commission tasked with redemarcating the disputed boundary between the Soviet Union and Japanese-controlled Manchukuo. However, this commission never achieved a mutually agreeable definition of the border in the contested area. In reality, the outcome was decided well before the group's inaugural meeting. Mere hours after the cease-fire took effect on the afternoon of August 11, General Grigory Shtern convened with a regimental commander from Japan's 19th Division to coordinate the disengagement of forces. With the conflict deemed "honorably" concluded, Japan's Imperial General Headquarters mandated the swift withdrawal of all Japanese troops to the west bank of the Tumen River. By the night of August 13, as the final Japanese soldier crossed the river, it effectively became the de facto border. Soviet forces promptly reoccupied Changkufeng Hill and the adjacent heights—a move that would carry unexpected and profound repercussions. Authoritative Japanese military analyses suggest that if negotiations in Moscow had dragged on for just one more day, the 19th Division would likely have been dislodged from Changkufeng and its surrounding elevations. Undoubtedly, General Shtern's infantry breathed a sigh of relief as the bloodshed ceased. Yet, one can't help but question why Moscow opted for a cease-fire at a juncture when Soviet troops were on the cusp of total battlefield triumph. Perhaps Kremlin leaders deemed it wiser to settle for a substantial gain, roughly three-quarters of their objectives, rather than risk everything. After all, Japan had mobilized threatening forces in eastern Manchuria, and the Imperial Army had a history of impulsive, unpredictable aggression. Moreover, amid the escalating crisis over Czechoslovakia, Moscow may have been wary of provoking a broader Asian conflict. Another theory posits that Soviet high command was misinformed about the ground situation. Reports of capturing a small segment of Changkufeng's crest might have been misinterpreted as control over the entire ridge, or an imminent full takeover before midnight on August 10. The unexpected phone call from Foreign Minister Maxim Litvinov to the Japanese embassy that night—proposing a one-kilometer Japanese retreat in exchange for a cease-fire along existing lines—hints at communication breakdowns between Shtern's headquarters and the Kremlin. Ironically, such lapses may have preserved Japanese military honor, allowing the 19th Division's evacuation through diplomacy rather than defeat. Both sides endured severe losses. Initial Japanese press reports claimed 158 killed and 740 wounded. However, the 19th Division's medical logs reveal a grimmer toll: 526 dead and 914 injured, totaling 1,440 casualties. The true figure may have climbed higher, possibly to 1,500–2,000. Following the armistice, the Soviet news agency TASS reported 236 Red Army fatalities and 611 wounded. Given Shtern's uphill assaults across open terrain against entrenched positions, these numbers seem understated. Attackers in such scenarios typically suffered two to three times the defenders' losses, suggesting Soviet casualties ranged from 3,000 to 5,000. This aligns with a Soviet Military Council investigation on August 31, 1938, which documented 408 killed and 2,807 wounded. Japanese estimates placed Soviet losses even higher, at 4,500–7,000. Not all victims perished in combat. Marshal Vasily Blyukher, a decorated Soviet commander, former warlord of the Far East, and Central Committee candidate, was summoned to Moscow in August 1938. Relieved of duty in September and arrested with his family in October, he faced charges of inadequate preparation against Japanese aggression and harboring "enemies of the people" within his ranks. On November 9, 1938, Blyukher died during interrogation a euphemism for torture-induced death.Other innocents suffered as well. In the wake of the fighting, Soviet authorities deported hundreds of thousands of Korean rice farmers from the Ussuri region to Kazakhstan, aiming to eradicate Korean settlements that Japanese spies had allegedly exploited. The Changkufeng clash indirectly hampered Japan's Wuhan offensive, a massive push to subdue China. The influx of troops and supplies for this campaign was briefly disrupted by the border flare-up. Notably, Kwantung Army's 2nd Air Group, slated for Wuhan, was retained due to the Soviet threat. Chiang Kai-shek's drastic measure, breaching the Yellow River dikes to flood Japanese advance routes—further delayed the assault. By October 25, 1938, when Japanese forces captured Hankow, Chiang had relocated his capital to distant Chungking. Paradoxically, Wuhan's fall cut rail links from Canton inland, heightening Chiang's reliance on Soviet aid routed overland and by air from Central Asia. Japan secured a tactical win but missed the decisive blow; Chinese resistance persisted, pinning down a million Japanese troops in occupation duties. What was the true significance of Changkufeng? For General Koiso Suetaka and the 19th Division, it evoked a mix of bitterness and pride. Those eager for combat got their share, though not on their terms. To veterans mourning fallen comrades on those desolate slopes, it might have felt like senseless tragedy. Yet, they fought valiantly under dire conditions, holding firm until a retreat that blended humiliation with imperial praise, a bittersweet inheritance. For the Red Army, it marked a crucial trial of resolve amid Stalin's purges. While Shtern's forces didn't shine brilliantly, they acquitted themselves well in adversity. The U.S. military attaché in Moscow observed that any purge-related inefficiencies had been surmounted, praising the Red Army's valor, reliability, and equipment. His counterpart in China, Colonel Joseph Stilwell, put it bluntly: the Soviets "appeared to advantage," urging skeptics to rethink notions of a weakened Red Army. Yet, by World War II's eve, many British, French, German, and Japanese leaders still dismissed it as a "paper tiger." Soviet leaders appeared content, promoting Shtern to command the Transbaikal Military District and colonel general by 1940, while honoring "Heroes of Lake Khasan" with medals. In a fiery November 7, 1938, speech, Marshal Kliment Voroshilov warned that future incursions would prompt strikes deep into enemy territory. Tokyo's views diverged sharply. Many in the military and government saw it as a stain on Imperial Army prestige, especially Kwantung Army, humiliated on Manchukuo soil it swore to protect. Colonel Masanobu Tsuji Inada, however, framed it as a successful reconnaissance, confirming Soviet border defense without broader aggression, allowing the Wuhan push to proceed safely. Critics, including Major General Gun Hashimoto and historians, questioned this. They argued IGHQ lacked contingency plans for a massive Soviet response, especially with Wuhan preparations underway since June. One expert warned Japan had "played with fire," risking Manchuria and Korea if escalation occurred. Yet, Japanese commanders gleaned few lessons, downplaying Soviet materiel superiority and maintaining disdain for Red Army prowess. The 19th Division's stand against outnumbered odds reinforced this hubris, as did tolerance for local insubordination—attitudes that would prove costly. The Kremlin, conversely, learned Japan remained unpredictable despite its China quagmire. But for Emperor Hirohito's intervention, the conflict might have ballooned. Amid purges and the Czech crisis, Stalin likely viewed it as a reminder of eastern vulnerabilities, especially with Munich advancing German threats westward. Both sides toyed with peril. Moderation won in Tokyo, but Kwantung Army seethed. On August 11, Premier Fumimaro Konoye noted the need for caution. Kwantung, however, pushed for and secured control of the disputed salient from Chosen Army by October 8, 1938. Even winter's chill couldn't quench their vengeful fire, setting the stage for future confrontations. A quick look at the regional map reveals how Manchukuo and the Mongolian People's Republic each jut into the other's territory like protruding salients. These bulges could be seen as aggressive thrusts into enemy land, yet they also risked encirclement and absorption by the opposing empire. A northward push from western Manchuria through Mongolia could sever the MPR and Soviet Far East from the USSR's heartland. Conversely, a pincer movement from Mongolia and the Soviet Maritime Province might envelop and isolate Manchukuo. This dynamic highlights the frontier's strategic volatility in the 1930s. One particularly tense sector was the broad Mongolian salient extending about 150 miles eastward into west-central Manchukuo. There, in mid-1939, Soviet-Japanese tensions erupted into major combat. Known to the Japanese as the Nomonhan Incident and to the Soviets and Mongolians as the Battle of Khalkhin Gol, this clash dwarfed the earlier Changkufeng affair in scale, duration, and impact. Spanning four months and claiming 30,000 to 50,000 casualties, it amounted to a small undeclared war, the modern era's first limited conflict between great powers. The Mongolian salient features vast, semiarid plains of sandy grassland, gently rolling terrain dotted with sparse scrub pines and low shrubs. The climate is unforgivingly continental: May brings hot days and freezing nights, while July and August see daytime highs exceeding 38°C (100°F in American units), with cool evenings. Swarms of mosquitoes and massive horseflies necessitate netting in summer. Rainfall is scarce, but dense morning fogs are common in August. Come September, temperatures plummet, with heavy snows by October and midwinter lows dipping to –34°C. This blend of North African aridity and North Dakotan winters supports only sparse populations, mainly two related but distinct Mongol tribes. The Buriat (or Barga) Mongols migrated into the Nomonhan area from the northwest in the late 17th to early 18th centuries, likely fleeing Russian expansion after the 1689 Treaty of Nerchinsk. Organized by Manchu emperors between 1732 and 1735, they settled east of the river they called Khalkhin Gol (Mongolian for "river"), in lands that would later become Manchukuo. The Khalkha Mongols, named for the word meaning "barrier" or "shield," traditionally guarded the Mongol Empire's northern frontiers. Their territories lay west of the Buriats, in what would become the MPR. For centuries, these tribes herded livestock across sands, river crossings, and desert paths, largely oblivious to any formal borders. For hundreds of years, the line dividing the Mongolian salient from western Manchuria was a hazy administrative divide within the Qing Empire. In the 20th century, Russia's detachment of Outer Mongolia and Japan's seizure of Manchuria transformed this vague boundary into a frontline between rival powers. The Nomonhan Incident ignited over this contested border. Near the salient's northeastern edge, the river, called Khalkhin Gol by Mongols and Soviets, and Halha by Manchurians and Japanese, flows northwest into Lake Buir Nor. The core dispute: Was the river, as Japan asserted, the historic boundary between Manchukuo and the MPR? Soviet and MPR officials insisted the line ran parallel to and 10–12 miles east of the river, claiming the intervening strip. Japan cited no fewer than 18 maps, from Chinese and Japanese sources, to support the river as the border, a logical choice in such barren terrain, where it served as the sole natural divider. Yet, Soviets and Mongolians countered with evidence like a 1919 Chinese postal atlas and maps from Japanese and Manchukuoan agencies (1919–1934). Unbeknownst to combatants, in July 1939, China's military attaché in Moscow shared a 1934 General Staff map with his American counterpart, showing the border east of the river. Postwar Japanese studies of 18th-century Chinese records confirm that in 1734, the Qing emperor set a boundary between Buriat and Khalkha Mongols east of the river, passing through the hamlet of Nomonhan—as the Soviets claimed. However, Kwantung Army Headquarters dismissed this as non-binding, viewing it as an internal Qing affair without Russian involvement. Two former Kwantung Army officers offer a pragmatic explanation: From 1931 to 1935, when Soviet forces in the Far East were weak, Japanese and Manchukuoan authorities imposed the river as the de facto border, with MPR acquiescence. By the mid- to late 1930s, as Soviet strength grew, Japan refused to yield, while Mongolians and Soviets rejected the river line, sparking clashes. In 1935, Kwantung Army revised its maps to align with the river claim. From late that year, the Lake Buir Nor–Halha sector saw frequent skirmishes between Manchukuoan and MPR patrols. Until mid-1938, frontier defense in northwestern Manchukuo fell to the 8th Border Garrison Unit , based near Hailar. This 7,000-man force, spread thin, lacked mobility, training, and, in Kwantung Army's eyes, combat readiness. That summer, the newly formed 23rd Division, under Kwantung Army, took station at Hailar, absorbing the 8th BGU under its command, led by Lieutenant General Michitaro Komatsubara. At 52, Komatsubara was a premier Russian specialist in the Imperial Army, with stints as military attaché in the USSR and head of Kwantung's Special Services Agency in Harbin. Standing 5'7" with a sturdy build, glasses, and a small mustache, he was detail-oriented, keeping meticulous diaries, writing lengthy letters, and composing poetry, though he lacked combat experience. Before departing Tokyo in July 1938, Komatsubara received briefings from Colonel Masazumi Inada, AGS Operations Section chief. Amid planning for Changkufeng, Inada urged calm on the Manchukuo-MPR border given China's ongoing campaigns. Guidelines: Ignore minor incidents, prioritize intelligence on Soviet forces east of Lake Baikal, and study operations against the Soviet Far East's western sector. Familiar with the region from his Harbin days, Komatsubara adopted a low-key approach. Neither impulsive nor aggressive, he kept the green 23rd Division near Hailar, delegating patrols to the 8th BGU. An autumn incident underscores his restraint. On November 1, 1938, an 8th BGU patrol was ambushed by MPR forces. Per Japanese accounts, the three-man team, led by a lieutenant, strayed too close to the border and was attacked 50 meters inside Manchukuo. The lieutenant escaped, but his men died. Komatsubara sent an infantry company to secure the site but forbade retaliation. He pursued body recovery diplomatically, protested to MPR and Soviet officials, and disciplined his officers: garrison leaders got five days' confinement for poor troop training, the lieutenant thirty days. Despite this caution, pressures at AGS and KwAHQ were mounting, poised to thrust the 23rd Division into fierce battle. Modern militaries routinely develop contingency plans against potential adversaries, and the mere existence of such strategies doesn't inherently signal aggressive intentions. That said, shifts in Japan's operational planning vis-à-vis the Soviet Union may have inadvertently fueled the Nomonhan Incident. From 1934 to 1938, Japanese war scenarios emphasized a massive surprise assault in the Ussuri River region, paired with defensive holding actions in northwestern Manchuria. However, between mid-1938 and early 1939, a clandestine joint task force from the Army General Staff  and Kwantung Army's Operations Departments crafted a bold new blueprint. This revised strategy proposed containing Soviet forces in the east and north while unleashing a full-scale offensive from Hailar, advancing west-northwest toward Chita and ultimately Lake Baikal. The goal: sever the Transbaikal Soviet Far East from the USSR's core. Dubbed Plan Eight-B, it gained Kwantung Army's endorsement in March 1939. Key architects—Colonels Takushiro Hattori and Masao Terada, along with Major Takeharu Shimanuki—were reassigned from AGS to Kwantung Army Headquarters to oversee implementation. The plan anticipated a five-year buildup before execution, with Hattori assuming the role of chief operations staff officer.  A map review exposes a glaring vulnerability in Plan Eight-B: the Japanese advance would leave its southern flank exposed to Soviet counterstrikes from the Mongolian salient. By spring 1939, KwAHQ likely began perceiving this protrusion as a strategic liability. Notably, at the outbreak of Nomonhan hostilities, no detailed operational contingencies for the area had been formalized. Concurrently, Japan initiated plans for a vital railroad linking Harlun Arshan to Hailar. While its direct tie to Plan Eight-B remains unclear, the route skirted perilously close to the Halha River, potentially heightening KwAHQ's focus on the disputed Mongolian salient. In early 1939, the 23rd Division intensified reconnaissance patrols near the river. Around this time, General Grigory Shtern, freshly appointed commander of Soviet Far Eastern forces, issued a public warning that Japan was gearing up for an assault on the Mongolian People's Republic. As Plan Eight-B took shape and railroad proposals advanced, KwAHQ issued a strikingly confrontational set of guidelines for frontier troops. These directives are often cited as a catalyst for the Nomonhan clash, forging a chain linking the 1937 Amur River incident, the 1938 Changkufeng debacle, and the 1939 conflict.Resentment had festered at KwAHQ over perceived AGS meddling during the Amur affair, which curtailed their command autonomy. This frustration intensified at Changkufeng, where General Kamezo Suetaka's 19th Division endured heavy losses, only for the contested Manchukuoan territory to be effectively ceded. Kwantung Army lobbied successfully to wrest oversight of the Changkufeng salient from Chosen Army. In November 1938, Major Masanobu Tsuji of KwAHQ's Operations Section was sent to survey the site. The audacious officer was dismayed: Soviet forces dominated the land from the disputed ridge to the Tumen River. Tsuji undertook several winter reconnaissance missions. His final outing in March 1939 involved leading 40 men to Changkufeng's base. With rifles slung non-threateningly, they ascended to within 200 yards of Soviet lines, formed a line, and urinated in unison, eliciting amused reactions from the enemy. They then picnicked with obentos and sake, sang army tunes, and left gifts of canned meat, chocolates, and whiskey. This theatrical stunt concealed Tsuji's real aim: covert photography proving Soviet fortifications encroached on Manchukuoan soil. Tsuji was a singular figure. Born of modest means, he embodied a modern samurai ethos, channeling a sharp intellect into a frail, often ailing body through feats of extraordinary daring. A creative tactician, he thrived in intelligence ops, political scheming, aerial scouting, planning, and frontline command—excelling across a tumultuous career. Yet, flaws marred his brilliance: narrow bigotry, virulent racism, and capacity for cruelty. Ever the ambitious outsider, Tsuji wielded outsized influence via gekokujo—Japan's tradition of subordinates steering policy from below. In 1939, he was a major, but his pivotal role at Nomonhan stemmed from this dynamic. Back in Hsinking after his Changkufeng escapade, Tsuji drafted a response plan: negotiate border "rectification" with the Soviets; if talks failed, launch an attack to expel intruders. Kwantung Army adopted it. Deputy Chief of Staff Major General Otozaburo Yano flew to Tokyo with Tsuji's photos, seeking AGS approval. There, he was rebuffed—Changkufeng was deemed settled, and minor violations should be overlooked amid Tokyo's aversion to Soviet conflict. Yano's plea that leniency would invite aggression was countered by notes on Europe's tensions restraining Moscow. Yano's return sparked outrage at KwAHQ, seen as AGS thwarting their imperial duty to safeguard Manchukuo. Fury peaked in the Operations Section, setting the stage for Tsuji's drafting of stringent new frontier guidelines: "Principles for the Settlement of Soviet-Manchukuoan Border Disputes." The core tenet: "If Soviet troops transgress the Manchukuoan frontiers, Kwantung Army will nip their ambitions in the bud by completely destroying them." Specific directives for local commanders included: "If the enemy crosses the frontiers … annihilate him without delay, employing strength carefully built up beforehand. To accomplish our mission, it is permissible to enter Soviet territory, or to trap or lure Soviet troops into Manchukuoan territory and allow them to remain there for some time… . Where boundary lines are not clearly defined, area defense commanders will, upon their own initiative, establish boundaries and indicate them to the forward elements… . In the event of an armed clash, fight until victory is won, regardless of relative strengths or of the location of the boundaries. If the enemy violates the borders, friendly units must challenge him courageously and endeavor to triumph in their zone of action without concerning themselves about the consequences, which will be the responsibility of higher headquarters." Major Tsuji Masanobu later justified the new guidelines by pointing to the "contradictory orders" that had hamstrung frontier commanders under the old rules. They were tasked with upholding Manchukuo's territorial integrity yet forbidden from actions that might spark conflict. This, Tsuji argued, bred hesitation, as officers feared repercussions for decisive responses to incursions. The updated directives aimed to alleviate this "anxiety," empowering local leaders to act boldly without personal liability. In truth, Tsuji's "Principles for the Settlement of Soviet-Manchukuoan Border Disputes" were more incendiary than conciliatory. They introduced provocative measures: authorizing commanders to unilaterally define unclear boundaries, enforce them with immediate force "shoot first, ask questions later", permit pursuits into enemy territory, and even encourage luring adversaries across the line. Such tactics flouted both government policy and official army doctrine, prioritizing escalation over restraint. The proposals sparked intense debate within Kwantung Army's Operations Section. Section chief Colonel Takushiro Hattori and Colonel Masao Terada outranked Tsuji, as did Major Takeharu Shimanuki, all recent transfers from the Army General Staff. Tsuji, however, boasted longer tenure at Kwantung Army Headquarters since April 1936 and in Operations since November 1937, making him the de facto veteran. Hattori and Terada hesitated to challenge the assertive major, whose reputation for intellect, persuasion, and deep knowledge of Manchuria commanded respect. In a 1960 interview, Shimanuki recalled Tsuji's dominance in discussions, where his proactive ideas often swayed the group. Unified, the section forwarded Tsuji's plan to Kwantung Army Command. Commander Lieutenant General Kenkichi Ueda consulted Chief of Staff General Rensuke Isogai and Vice Chief General Otozaburo Yano, seasoned leaders who should have spotted the guidelines' volatility. Yet, lingering grudges from AGS "interference" in past incidents like the Amur River and Changkufeng clouded their judgment. Ueda, Isogai, and Tsuji shared history from the 1932 Shanghai Incident: Tsuji, then a captain, led a company in the 7th Regiment under Colonel Isogai, with Yano as staff officer and Ueda commanding the 9th Division. Tsuji was wounded there, forging bonds of camaraderie. This "clique," which grew to include Hattori, Terada, and Shimanuki, amplified Tsuji's influence. Despite Isogai's initial reservations as the group's moderate voice, the guidelines won approval. Ueda issued them as Kwantung Army Operations Order 1488 on April 25, 1939, during a division commanders' conference at KwAHQ. A routine copy reached AGS in Tokyo, but no formal reply came. Preoccupied with the China War and alliance talks with Germany, AGS may have overlooked border matters. Colonel Masazumi Inada, AGS Operations head, later noted basic acceptance of Order 1488, with an informal expectation—relayed to Hattori and Terada—of prior consultation on violations. KwAHQ dismissed this as another Tokyo intrusion on their autonomy. Some Japanese analysts contend a stern AGS rejection might have prevented Nomonhan's catastrophe, though quelling Kwantung's defiance could have required mass staff reassignments, a disruptive step AGS avoided. Tsuji countered that permitting forceful action at Changkufeng would have deterred Nomonhan altogether, underscoring the interconnectedness of these clashes while implicitly critiquing the 1939 battle's location. Undeniably, Order 1488's issuance on April 25 paved the way for conflict three weeks later. Japanese records confirm that Khalkha Mongols and MPR patrols routinely crossed the Halha River—viewed by them as internal territory, 10 miles from the true border. Such crossings passed uneventfully in March and April 1939. Post-Order 1488, however, 23rd Division commander General Michitaro Komatsubara responded aggressively, setting the stage for escalation. The Nomonhan Incident ignited with a border clash on May 11–12, 1939, that rapidly spiraled into a major conflict. Over a dozen "authoritative" accounts exist, varying in viewpoint, focus, and specifics. After cross-referencing these sources, a coherent timeline emerges. On the night of May 10–11, a 20-man Mongolian People's Republic border patrol crossed eastward over the Halha River (known as Khalkhin Gol to Mongols and Soviets). About 10 miles east, atop a 150-foot sandy hill, lay the tiny hamlet of Nomonhan, a cluster of crude huts housing a few Mongol families. Just south flowed the Holsten River, merging westward into the broader Halha. By morning on May 11, Manchukuoan forces spotted the MPR patrol north of the Holsten and west of Nomonhan. In the MPR/Soviet perspective, Nomonhan Hill marked the Mongolia-Manchuria border. To Manchukuoans and Japanese, it sat 10 miles inside Manchukuo, well east of the Halha. A 40-man Manchukuoan cavalry unit repelled the Mongolians back across the river, inflicting initial casualties on both sides—the Manchukuoans drawing first blood. The MPR patrol leader exaggerated the attackers as 200 strong. The next day, May 12, a 60-man MPR force under Major P. Chogdan evicted the Manchukuoans from the disputed zone, reestablishing positions between the Halha and Nomonhan. The Manchukuoans, in turn, reported facing 700 enemies. Sporadic skirmishes and maneuvering persisted through the week. On May 13, two days post-clash, the local Manchukuoan commander alerted General Michitaro Komatsubara's 23rd Division headquarters in Hailar. Simultaneously, Major Chogdan reported to Soviet military command in Ulaanbaatar, Mongolia's capital. What began as a Mongolian-Manchukuoan spat was poised to draw in Soviet and Japanese patrons. Attributing the May 10–11 violation hinges on border interpretations: both sides claimed the Halha-Nomonhan strip. Yet, most accounts concur that Manchukuoan forces initiated the fighting. Post-May 13 notifications to Moscow and Tokyo clarify the record thereafter. Midday on May 13, Komatsubara was leading a staff conference on the newly issued Kwantung Army Operations Order 1488—Major Tsuji Masanobu's aggressive border guidelines. Ironically, the first Nomonhan combat report arrived mid-discussion. Officers present recall Komatsubara deciding instantly to "destroy the invading Outer Mongolian forces" per Order 1488. That afternoon, he informed Kwantung Army Headquarters of the incident and his intent to eradicate the intruders, requesting air support and trucks. General Kenkichi Ueda, Kwantung commander, approved Komatsubara's "positive attitude," dispatching six scout planes, 40 fighters, 10 light bombers, two anti-aircraft batteries, and two motorized transport companies. Ueda added a caveat: exercise "extreme caution" to prevent escalation—a paradoxical blend of destruction and restraint, reflective of KwAHQ's fervent mood. Ueda relayed the details to Tokyo's Army General Staff, which responded that Kwantung should handle it "appropriately." Despite Kwantung's impulsive reputation, Tokyo deferred, perhaps trusting the northern strategic imbalance, eight Japanese divisions versus 30 Soviet ones from Lake Baikal to Vladivostok, would enforce prudence. This faith proved misguided. On May 14, Major Tsuji flew from KwAHQ for aerial reconnaissance over Nomonhan, spotting 20 horses but no troops. Upon landing, a fresh bullet hole in his plane confirmed lingering MPR presence east of the Halha. Tsuji briefed 23rd Division staff and reported to Ueda that the incident seemed minor. Aligning with Order 1488's spirit, Komatsubara deployed a force under Lieutenant Colonel Yaozo Azuma: an armored car company, two infantry companies, and a cavalry troop. Arriving at Nomonhan on May 15, Azuma learned most MPR forces had retreated westward across the Halha the prior night, with only token elements remaining, and those withdrawing. Undeterred, he pursued. The advance met scant resistance, as foes had crossed the river. However, Japanese light bombers struck a small MPR concentration on the west bank, Outpost Number 7, killing two and wounding 15 per MPR reports; Japanese claimed 30–40 kills. All agree: the raid targeted undisputed MPR territory. Hearing of May 15's events, Komatsubara deemed the Mongolians sufficiently rebuked and recalled Azuma to Hailar on May 16. KwAHQ concurred, closing the matter. Soviet leaders, however, saw it differently. Mid-May prompted Soviet support for the MPR under their 1936 Mutual Defense Pact. The Red Army's 57th Corps, stationed in Mongolia, faced initial disarray: Commander Nikolai Feklenko was hunting, Chief of Staff A. M. Kushchev in Ulan Ude with his ill wife. Moscow learned of clashes via international press from Japanese sources, sparking Chief of Staff Boris Shaposhnikov's furious inquiry. Feklenko and Kushchev rushed back to Ulaanbaatar, dispatching a mixed force—a battalion from the 149th Infantry Regiment (36th Division), plus light armor and artillery from the 11th Tank Brigade—to Tamsag Bulak, 80 miles west of the Halha. Led by Major A. E. Bykov, it bolstered the MPR's 6th Cavalry Division. Bykov and Cavalry Commander Colonel Shoaaiibuu inspected the site on May 15, post-Azum's departure. The cavalry arrived two days later, backed by Bykov (ordered to remain west of the river and avoid combat if possible). Some MPR troops recrossed, occupying the disputed zone. Clashes with Manchukuoan cavalry resumed and intensified. Notified of renewed hostilities, Komatsubara viewed it as defiance, a personal affront. Emboldened by Order 1488, he aimed not just to repel but to encircle and annihilate. The incident was on the verge of major expansion. I would like to take this time to remind you all that this podcast is only made possible through the efforts of Kings and Generals over at Youtube. Please go subscribe to Kings and Generals over at Youtube and to continue helping us produce this content please check out www.patreon.com/kingsandgenerals. If you are still hungry after that, give my personal channel a look over at The Pacific War Channel at Youtube, it would mean a lot to me. The ghosts of the Changufeng incident have come back to haunt both the USSR and Japan. Those like Tsuji Masanobu instigated yet another border clash that would erupt into a full blown battle that would set a precedent for both nations until the very end of WW2. 

Juan shares his opinion on Shakur Stevenson, Keyshawn Davis, Xander Zayas, and the latest news in the sport of boxing.

On the Flyover Conservatives Show, we sat down with Che Ahn, a pastor, revival leader, and gubernatorial candidate who believes California's crisis is spiritual before it is political. He shares the supernatural moments that led him into the race, the Supreme Court battle that put him against state power, and the growing movement of unified Christians mobilizing through prayer and civic action. We explore whether a united Church could succeed where politics, money, and government programs have failed.TO WATCH ALL FLYOVER CONTENT: www.theflyoverapp.comFollow and Subscribe on YouTube: https://www.youtube.com/@TheFlyoverConservativesShow Che AhnWEBSITE: www.che4ca.comX: https://x.com/che_ahn Che Ahn is a pastor, global ministry leader, and revival voice who has spent decades mobilizing prayer movements and church networks across the United States and in more than 70 nations. He is the founder of Harvest International Ministry and has helped lead large-scale national gatherings focused on prayer, repentance, and spiritual awakening. After successfully challenging state shutdown orders in a Supreme Court case defending religious freedom, he is now running for Governor of California, calling for spiritual renewal, parental rights, fiscal responsibility, and a return to constitutional and biblical principles.-------------------------------------------

Pedersen v. Unified Patents, LLC

In this episode of the Board Drill Podcast, Kyle and Matt sit down with Coach Tom Yashinsky, head coach at Onalaska High School in Wisconsin, to dive into the nuts and bolts of building a consistent and competitive football program from the ground up. Coach Yashinsky brings nearly two decades of experience and shares detailed insight on aligning your program 9–12, structuring staff, developing coaches, and building buy-in across your community.Whether you're a head coach, coordinator, or position coach, this one's packed with practical wisdom you can apply immediately. Don't miss the breakdown on practice structure, middle school alignment, and how Onalaska turned a struggling program into a perennial playoff team.Timestamps:0:00 – Intro and how Coach Yashinsky joined the show1:10 – Onalaska football background and program context3:04 – Taking over the program under tough circumstances6:35 – Aligning your staff across all levels (9–12)9:58 – Year-round planning and development focus12:45 – The importance of coach retention and mentorship16:12 – Practice organization and seasonal phases20:41 – Lifting schedules and coordinating with multi-sport athletes24:09 – Offseason program and coach roles28:57 – Building culture: “Be Early, Be Loud, Be Invested”32:44 – Middle school and youth integration strategies37:11 – Handling success: sustaining performance and expectations42:36 – Advice for new head coaches46:58 – Final thoughts and how to connect with Coach YashinskySubscribe, like, and share if you're a coach looking for real-world solutions and ideas you can bring into your own program tomorrow.

Unified

Must humanity unite to colonize space, or can rivalry and diversity be our greatest strengths among the stars?Get Nebula using my link for 50% off an annual subscription: https://go.nebula.tv/isaacarthurCheck out Mad Kings: https://nebula.tv/madkings?ref=isaacarthurWatch my exclusive video Chronoengineering: https://nebula.tv/videos/isaacarthur-chronoengineering-manipulating-time-as-technology

Must humanity unite to colonize space, or can rivalry and diversity be our greatest strengths among the stars?Get Nebula using my link for 50% off an annual subscription: https://go.nebula.tv/isaacarthurCheck out Mad Kings: https://nebula.tv/madkings?ref=isaacarthurWatch my exclusive video Chronoengineering: https://nebula.tv/videos/isaacarthur-chronoengineering-manipulating-time-as-technology

Alec Patton talks to Beverley Jenkins and Kate Hogan of the System Improvement Leads Networked Improvement Community and Nicole Leveille of Cloverdale Unified School District about how Cloverdale dramatically increased the percentage of students with IEPs in the general education population, and cut chronic absenteeism among students with disabilities in half. Every other week, we publish a newsletter with great resources like this one, sign up for it here! What are you waiting for, register for the National Summit for Improvement in Education before you miss out! Episode Notes: You can read an article by Kate Hogan and Sandra Park about this improvement network on Unboxed here ! To learn more about the System Improvement Leads (SIL) team and their supports, visit  systemimprovement.org To learn more about California's Compliance and Improvement Monitoring process, visit  caltan.info In partnership with Cloverdale, SIL has published a strategy handout  linked here Learn more about the High Tech High Graduate School of Education

Automatic Script Execution In Visual Studio Code Visual Studio Code will read configuration files within the source code that may lead to code execution. https://isc.sans.edu/diary/Automatic%20Script%20Execution%20In%20Visual%20Studio%20Code/32644 Cisco Unified Communications Products Remote Code Execution Vulnerability A vulnerability in Cisco Unified Communications Manager (Unified CM), Cisco Unified Communications Manager Session Management Edition (Unified CM SME), Cisco Unified Communications Manager IM & Presence Service (Unified CM IM&P), Cisco Unity Connection, and Cisco Webex Calling Dedicated Instance could allow an unauthenticated, remote attacker to execute arbitrary commands on the underlying operating system of an affected device. https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-voice-rce-mORhqY4b Zoom Vulnerability A Command Injection vulnerability in Zoom Node Multimedia Routers (MMRs) before version 5.2.1716.0 may allow a meeting participant to execute remote code on the MMR via network access. https://www.zoom.com/en/trust/security-bulletin/zsb-26001/ Possible new SSO Exploit (CVE-2025-59718) on 7.4.9 https://www.reddit.com/r/fortinet/comments/1qibdcb/possible_new_sso_exploit_cve202559718_on_749/ SANS SOC Survey The 2026 SOC Survey is open, and we need your input to create a meaningful report. Please share your experience so we can advocate for what actually works in the trenches. https://survey.sans.org/jfe/form/SV_3ViqWZgWnfQAzkO?is=socsurveystormcenter

In part two of the "Come Back to God" podcast mini-series, Lisa Whittle challenges the culture of "revival hype" and invites listeners to consider revival's deep, holy roots. While stadiums and tents may symbolize revival, Lisa emphasizes that true revival is about heart transformation—marked by conviction, discomfort, and unity. She reflects on biblical examples like Pentecost and the Moravian Revival to highlight the importance of unified prayer, confession, and God's presence. Lisa encourages us to seek authentic revival—not for show, but for lasting change—and invites listeners to reflect on how they can foster revival within their own hearts and communities. The episode concludes with practical steps to pursue revival, including prayer, reconciliation, and justice.Listen in to learn more:(1:10) Revival hype vs. true revival(1:45) The sacredness of personal revival and its unseen work(3:47) Rocky analogy: Revival isn't for public show, it's deep and gritty(7:42) How Global Christian Relief is sparking revival(8:32) Biblical foundation: Acts 1 & 2, unity in prayer, and the early church(10:00) The Moravian Revival of 1727: A 100-year prayer movement that sparked global missions(11:33) Elements of authentic revival: Unified hunger, extraordinary prayer, confession and reconciliation, sincere worship, centered on God's presence(16:15) The role of justice and righteous living in revival (Amos 5:24)(16:50) Practical steps for fostering revival in your life and community(17:44) Closing encouragement: Pray, unite, invite others, and seek revival nowMentioned in the episode:Come Back to God Bible Study: https://www.lisawhittle.com/comebackGlobal Christian Relief: http://link.globalchristianrelief.org/lisaCoaching with Lisa: https://www.lisawhittle.com/coaching-with-lisa Connect with Lisa:Website: https://www.lisawhittle.comSubstack: https://letsbeclear.substack.comYouTube: https://www.youtube.com/@lisawhittleofficialInstagram: https://www.instagram.com/lisawhittleFacebook: https://www.facebook.com/lisawhittleofficial