Eye On A.I.

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.

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.

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.

The AI chip market looks monolithic from the outside - NVIDIA dominates, and everyone else is fighting for scraps. But d-Matrix's CEO Sid Sheth argues that the market is quietly splitting into two distinct tiers, and the one that's exploding right now is the one NVIDIA's architecture isn't built for. In this episode, Sid joins Craig Smith to explain the "premium token economy": a new class of AI inference where interactivity is the product, users pay ten times more per million tokens for instant responses, and the memory bandwidth limits of GPU-based systems create a structural ceiling that purpose-built architectures don't have.
The conversation is unusually candid about what AI actually looks like at the executive level: Sid describes using Claude as a sounding board for M&A strategy, producing full integration plans in 15 minutes that used to require entire banking advisory teams, and watching AI shift from a tool that echoed his ideas back at him to one that genuinely disagrees, flags what he missed, and pushes back with enough confidence to be useful. He also makes the case that we're at the beginning of a shift from individual agents to what he calls "organizational AI" - teams of agents running entire company functions at a high level of abstraction - and that the infrastructure bet d-Matrix is making positions them directly in the path of that wave.
Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of 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.