Eye On A.I.

Craig S. Smith

Eye on A.I. is a biweekly podcast, hosted by longtime New York Times correspondent Craig S. Smith. In each episode, Craig will talk to people making a difference in artificial intelligence. The podcast aims to put incremental advances into a broader context and consider the global implications of the developing technology. AI is about to change your world, so pay attention.

  1. hace 11 h

    86% of What Coding Agents Do Is Just Reading — Not Solving | Alexander Whedon of Subquadratic

    Every AI model in production today has the same hidden tax: doubling the context window quadruples the compute. That's what quadratic compute complexity means in practice, and it's the reason enterprises are spending most of their AI engineering budget on context management rather than on the actual problems they're trying to solve. Alexander Whedon, co-founder and CTO of Subquadratic, joins Craig Smith to explain how SubQ's sparse attention mechanism eliminates that tax, achieving 40 times faster inference and 64 times less compute than standard attention at one million tokens, and what becomes possible when that constraint disappears. The conversation covers striking benchmark findings: 86% of what frontier coding agents do is "read steps," just trying to gather and organize context before the actual problem-solving begins, and frontier models drop well below 50% accuracy on financial document analysis at 500,000 tokens, revealing how asymmetric long context capability actually is across industries. The most commercially important argument in this episode is about enterprise data. Most large organizations are sitting on hundreds of billions of tokens of data they've never been able to put to work in an AI product, told they need a $10 million data transformation project before they can even start building. Alex's core claim is that SubQ's architecture makes that barrier no longer necessary, enabling enterprises to process far more of their data with far less curation, at a fraction of the cost. He closes with what he describes as the most important and underexplored frontier in AI right now: we are still very far from understanding what users actually want from models reasoning over millions of tokens, and the product and alignment work needed to answer that question has barely begun. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

    86% of What Coding Agents Do Is Just Reading — Not Solving | Alexander Whedon of Subquadratic

  2. hace 5 d

    From 10 Drones a Month to Nearly 100,000 — Inside Ukraine's Largest Drone Manufacturer | Marko Kushnir, General Cherry

    In 2023, General Cherry started making 10 drones a month. Today they're approaching 100,000. Marko Kushnir, communications director of one of Ukraine's top-five drone manufacturers, joins Craig Smith for one of the most operationally specific conversations available about what drone warfare looks like at industrial scale, from the daily feedback loops with front-line units that drive product iteration, to the $2,000 interceptor drone that can destroy a $100,000 Shahed, to the on-device AI targeting model that guides an interceptor to impact at 70% accuracy after the operator activates it and steps back. The conversation's most important insights are structural rather than technical. Marko describes the fundamental asymmetry of the conflict with unusual precision: Ukraine's decentralized, startup-driven ecosystem produces new technologies faster than Russia's command economy, but Russia's vertical industrial structure copies and scales those technologies faster than Ukraine can stay ahead. He also expresses genuine alarm about fully autonomous AI drones, not from an ethical standpoint but from a practical one: any autonomous capability Ukraine deploys will be in Russian hands within weeks, making full autonomy a danger Ukraine would share immediately with its enemy. The episode closes with his most far-reaching argument: just as the internet era created a cybersecurity industry that every organization eventually had to build, the drone era is now beginning, and every government, police force, and major corporation will soon need a drone security department to function safely in a world where drones are as common as smartphones. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

    From 10 Drones a Month to Nearly 100,000 — Inside Ukraine's Largest Drone Manufacturer | Marko Kushnir, General Cherry

  3. 31 ago

    In 5 to 10 Years, Using Weapons Without AI Will Be Considered Unethical | Yaroslav Azhnyuk, The Fourth Law

    A Ukrainian entrepreneur who spent 14 years building cameras for pets pivoted to building cameras that down Shaheds, and is now building the autonomy software that could define how wars are fought for the next generation. Yaroslav Azhnyuk, co-founder of Fourth Law, joins Craig Smith in Kyiv to explain why Ukraine has become what he calls the Defense Valley or the Florence of Defense: a dense, fast-moving ecosystem of founders, engineers, and military operators who are building, testing, and iterating on autonomous drone systems in real combat conditions, with a feedback loop that no defense contractor in the West can currently match. The conversation covers Azhnyuk's five-level autonomy framework for drones, the eight-dimensional model for what a fully autonomous battlefield ecosystem requires, and the economic math that he believes makes global rearmament inevitable: a $500 drone that can already destroy a $5 million tank becomes roughly 10,000 times more capable when full autonomy is added for a few hundred dollars more. The competitive landscape is mapped with unusual candor, an "Apple vs. Android" comparison between Eric Schmidt's vertically integrated interceptors and Fourth Law's modular platform approach, alongside two arguments that cut against the mainstream narrative. First, that within 5 to 10 years it may become unethical to use weapons without AI, because non-AI weapons cause more collateral damage, not less. And second, that the real AGI risk isn't Skynet, it's the subtle transfer of power that happens when 100 smarter advisors gradually stop waiting for the President to decide, and nobody notices until the President is no longer the one making decisions. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

    In 5 to 10 Years, Using Weapons Without AI Will Be Considered Unethical | Yaroslav Azhnyuk, The Fourth Law

  4. 27 ago

    Inside Ukraine's Azov Drone R&D: The Engineer Building AI Weapons 18 km From the Front Line | Alexander Palamarchuk

    The most consequential arms race in the world right now isn't nuclear, it's software. Craig Smith speaks with Alexander Palamarchuk, an engineer in the R&D department of Ukraine's Azov Brigade, calling in from approximately 18 kilometers from the front line in the Pokrovsk region. What emerges is one of the most technically candid accounts available of what drone warfare actually looks like from the inside: how Ukraine went from homemade reconnaissance drones in 2014 to AI-guided systems being developed and tested in real combat today, how the jamming arms race has forced his unit to develop custom frequency systems spanning 100 to 3,000 megahertz in a constant search for clean windows Russia hasn't yet closed, and why the tank - once the defining weapon of land warfare - has been reduced on the modern battlefield to a mobile jamming platform. The most important distinction Alexander draws is one that rarely surfaces in mainstream coverage: AI already exists that can recognize and classify vehicles and people with high accuracy. The unsolved problem isn't recognition, it's discrimination, determining with certainty whether a recognized target is military or civilian. That gap is the only thing standing between today's AI-assisted drones and fully autonomous lethal systems, and Alexander puts the timeline for closing it at approximately two years. The US maintains an official policy of not developing fully autonomous lethal weapons for ethical reasons. On the battlefield 18 kilometers from where Alexander is speaking, that policy is being outpaced in real time, and he is clear that NATO's software advantage positions Western countries to win that race before anyone else gets there. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

    Inside Ukraine's Azov Drone R&D: The Engineer Building AI Weapons 18 km From the Front Line | Alexander Palamarchuk

  5. 24 ago

    95% of AI Agent Projects Fail to Reach Production. Here's Why | Manoj Saxena, TrustWise

    It takes a few hours to build an AI agent. It takes six to seven months to get it into production. Manoj Saxena - the executive who commercialized IBM Watson - built TrustWise around one thesis: intelligence without control is not deployable. In this episode, he joins Craig Smith to explain why 95% of enterprise AI agent projects are stalling between pilot and production, and why the answer has nothing to do with the quality of the underlying models. The bottleneck, Saxena argues, is the absence of an entirely new class of infrastructure, something that can evaluate every tool call, every action, every output of every agent at runtime, in milliseconds, against the full stack of alignment requirements that govern what an AI is actually allowed to do inside a real enterprise. TrustWise's AI Control Tower does that across all vendors and agent frameworks simultaneously, operating in live, sidecar, batch, or simulation mode and aligning agent behavior against six layers of requirements - from UN Human Rights frameworks down to individual customer SLA commitments - in 10 to 300 milliseconds per decision. The conversation covers demonstrated results (83% cost reduction, 40% safety improvement), the token consumption paradox that's making agentic AI far more expensive than expected even as token costs fall, and a milestone Saxena compares to the moment data traffic surpassed voice on AT&T's network: last month, for the first time ever, agent traffic on the internet exceeded human traffic. The episode closes with a preview of Genesis agents, TrustWise's next product, designed not just to prevent bad outcomes but to surface beneficial hypotheses by looking 95 moves deep into enterprise data, in domains like fraud detection and revenue leakage, in the way Deep Blue looked 95 moves deep in chess. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

    95% of AI Agent Projects Fail to Reach Production. Here's Why | Manoj Saxena, TrustWise

  6. 19 ago

    From Zero to 150 Robots in Just 20 Months | Mike LeBlanc, Foundation Future Industries

    Most humanoid robot companies are still running curated demos in replica environments. Foundation Future Industries is running 150 robots on real automotive production lines in Georgia, and heading to Ukraine this year to deploy on the battlefield. Mike LeBlanc, the co-founder of Foundation Future Industries - currently the only company supplying humanoid robots to the US Department of Defense, with contracts across the Army, Navy, and Air Force, joins Craig Smith to explain why the race is moving faster than almost anyone in the industry believes, and why the companies that are moving cautiously are about to be left behind. His frame is striking: he keeps a framed 1906 New York Times article on his office wall predicting that human flight would take between one million and ten million years. It was published three months before the Wright Brothers flew. He thinks humanoids are in exactly that moment right now. The conversation covers the full operational picture: how Foundation trains robots using video rather than simulation; why the fry-cook robot that couldn't open the bag of fries is a perfect metaphor for everything wrong with how most companies approach go-to-market in this space; why the human form factor isn't a philosophical preference but an empirical fact,  humans are still doing every job in every factory that other robots can't, and that's the proof of concept; and why Mike LeBlanc isn't particularly worried about competing against Boston Dynamics backed by Google DeepMind, because they're still demoing in replica sites while Foundation is deploying on production lines. The episode ends with a bet: LeBlanc tells Craig that in twelve months, he'll be back to report 10,000 robots deployed in the world. Craig says he remains cautious. One of them is going to be right. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

    From Zero to 150 Robots in Just 20 Months | Mike LeBlanc, Foundation Future Industries

  7. 13 ago

    American Companies Have 36 Months to Go AI-Native or Get Left Behind | Drew Cukor, TWG AI

    The same tools that slowed the U.S. military down in Afghanistan (PowerPoint, Excel, email, and Word) are now slowing American businesses down in the AI race. Drew Cukor spent 30 years as a Marine intelligence officer, helped build Project Maven into a battlefield command and control system, served as Chief Data Officer at JP Morgan, and is now leading AI transformation at TWG AI. In this episode, he joins Craig Smith to make a case that most enterprise AI strategies are fundamentally broken, not because the technology isn't there, but because companies are storing their data in Microsoft file folders where it becomes inaccessible to AI, appointing AI officers who block progress rather than enable it, and mistaking chatbot deployments for transformation. Cukor's prescription is specific: take a company's core workflows apart, how it acquires customers, delivers services, handles back office operations, and rebuild them from scratch with AI embedded throughout, protected inside Palantir Foundry, delivered within 36 months, with the CEO owning the outcome rather than delegating it to a CTO or a made-up AI officer role. The stakes, he argues, are not abstract: China is going AI-native from the start without the legacy infrastructure that's slowing American enterprise, token spend is approaching the cost of a human salary making poorly designed AI workflows as expensive as bad hiring decisions, and the window for acting is closing. The most important video he recommends any business leader watch isn't one where the AI wins, it's the footage of Lee Sedol losing to AlphaGo and realizing mid-game that he no longer understands how the game works. That moment, Cukor says, is coming for every legacy business that doesn't move now. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

    American Companies Have 36 Months to Go AI-Native or Get Left Behind | Drew Cukor, TWG AI

  8. 10 ago

    In 5 Years, 90% of What You Use AI For Will Run on Your Smartphone | Paolo Ardoino, Tether

    Hundreds of billions of dollars are flowing into AI data centers right now, and Paolo Ardoino, CEO of Tether - the company behind the world's most widely used stablecoin with 573 million users - thinks that investment is going to age very badly. In this episode, he joins Craig Smith to explain QVAC, Tether's open-source platform for running AI on smartphones, laptops, and edge devices, and to make a case that within five years, 90% of what ordinary people use AI for will run entirely on consumer hardware, without touching a data center. The evidence is already there: Tether's team built a 4-billion-parameter medical AI model that outperforms Google's 27-billion-parameter MedGemma, running on a good smartphone, and a 1.7-billion-parameter version that runs on the average $80 smartphone available in Africa. The deeper argument in this conversation is philosophical as much as technical. Ardoino applies the same disintermediation logic that made sending dollars to the world's unbanked free - zero transaction fees, revenue from treasury bill interest - to AI: "not your AI, not your intelligence." If you don't control how your AI runs and your data never leaves your device, the AI is genuinely yours. If it does, someone else is getting smarter with your information. He also makes a pointed economic argument: the AI companies currently charging $200 for subscriptions that cost $1,000 to $5,000 to deliver are subsidizing growth while private, and when they go public, retail investors will absorb the gap. His prescription isn't to stop building, it's to build differently, toward millions of small interacting models rather than trillion-parameter monoliths, toward devices that think locally rather than systems that route everything through Ireland and back. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.

    In 5 Years, 90% of What You Use AI For Will Run on Your Smartphone | Paolo Ardoino, Tether
Ver todo (378)

Valoraciones y reseñas

4,3
de 5
3 valoraciones

Información

Eye on A.I. is a biweekly podcast, hosted by longtime New York Times correspondent Craig S. Smith. In each episode, Craig will talk to people making a difference in artificial intelligence. The podcast aims to put incremental advances into a broader context and consider the global implications of the developing technology. AI is about to change your world, so pay attention.

Ficha técnica

  • Creado por
    Craig S. Smith
  • Años en activo
    2018 - 2026
  • Episodios
    378
  • Clasificación
    Apto
  • Derechos de autor
    © Eye On A.I.
  • Sitio web del podcast
    Eye On A.I.

Quizá también te guste