



275 episodes
by Craig S. Smith · English
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.









Capital One’s warning about scaling “agentic AI” (e.g., aiming for 20,000 agents) and why most enterprise pilots don’t reach production; the episode explains the prerequisites for enterprise-scale…
20 Nov 2026 · 39m · 16 chapters
Why enterprise AI fails to deliver ROI despite strong demos; the bottleneck is governance, orchestration, semantic understanding, and a “memory layer,” not just using smarter LLMs.
13 Nov 2026 · 37m · 14 chapters
Reinforcement learning reward signals and why AI agents can behave “rogue” when rewards are misspecified or incomplete.
6 Nov 2026 · 44m · 15 chapters
Tealium CEO Jeff Lunsford argues that “agentic” traffic (AI agents browsing/acting on websites) will reach ~50% of enterprise web traffic by mid-next year.
30 Oct 2026 · 56m · 23 chapters
Zip’s “AI superagents” for procurement automation end-to-end, with emphasis on governance, auditability, and safety for actions that affect ERP/financials.
23 Oct 2026 · 51m · 24 chapters
Trioma’s “cheapest AI in 2026” pitch: a vertically integrated AI platform that uses 65 models plus an orchestration engine to deliver end results (not just chat) while minimizing token burn and cost.
16 Oct 2026 · 55m · 22 chapters
Ilya Sutskever (ChatGPT/AI creator) explains how AI “understands” via next-token prediction as compression, why scaling works only when you scale the right thing, and what limits remain (notably…
9 Oct 2026 · 44m · 15 chapters
“Coordination tax” in AI-driven customer operations—how adding AI and agent handoffs can increase context sharing and burnout, even while improving customer experience and retention.
6 Oct 2026 · 51m · 23 chapters
Ukraine’s Brave1 “Army of Drones” bonus program and the Delta/Avengers digital “digital army” system that verifies drone strikes, awards points, and lets units buy more drones via a marketplace to…
2 Oct 2026 · 34m · 18 chapters
Dr. Christof Koch argues that today’s AI (including large language models) is unlikely to be conscious because consciousness depends on specific brain-like causal structure, not just input-output…
30 Sep 2026 · 1h 2m · 23 chapters
Alex Liannyi (CTO of NORDA Dynamics) discusses how “fully autonomous attack” technology is already feasible, but why NORDA’s system keeps a human in the loop.
24 Sep 2026 · 19m · 10 chapters
Maj. “Phoenix” of Ukraine’s Lazarus Group explains how the unit’s “war of drones” works—using satellite-linked heavy bombers, ISR aircraft/drones for target finding, FPV drones for short-range…
21 Sep 2026 · 1h 6m · 27 chapters
How AI “validation layer” systems decide which B2B software to recommend, and why G2 positions itself as a trust layer for those recommendations (not just traffic).
14 Sep 2026 · 59m · 34 chapters
Why traditional cybersecurity “walls” fail in a borderless internet (and with AI code-scanning/exploit generation), and the proposed fix: data-level, field/paragraph/cell encryption tied to identity,…
10 Sep 2026 · 55m · 29 chapters
Subquadratic’s approach to long-context “sparse attention” that reduces quadratic attention compute so models can process multi-million-token inputs efficiently, enabling more robust RAG/agent…
8 Sep 2026 · 55m · 20 chapters
Ukraine’s drone industry and battlefield use, focused on General Cherry’s rapid scaling and semi-autonomous AI interceptors against Russian Shahed/Geran drones, within a “kill zone” concept where…
3 Sep 2026 · 38m · 17 chapters
The episode argues that software-driven autonomy in drones and defense will make non-AI weapon use unethical within 5–10 years, because non-autonomous systems increase collateral damage.
31 Aug 2026 · 53m · 25 chapters
Alexander Palamarchuk (call sign “Paradise”) discusses Azov’s drone R&D for AI-enabled navigation and target recognition under heavy Russian jamming, including radio/fiber/internet links, visual…
27 Aug 2026 · 42m · 16 chapters
Trust and runtime governance for agentic AI. The episode argues that 95% of AI agent projects fail to reach production because “intelligence without control” isn’t deployable: enterprises must ensure…
24 Aug 2026 · 1h 2m · 29 chapters
Foundation Future Industries’ push to deploy humanoid robots for defense and industrial work, arguing the technology is progressing faster than critics expect.
19 Aug 2026 · 1h 6m · 25 chapters
The episode argues that American enterprises have about 36 months to become “AI-native” or risk being outcompeted by AI-native startups and faster-moving rivals (especially China).
13 Aug 2026 · 59m · 23 chapters
Paolo Ardoino argues that most everyday AI use will shift from cloud data centers to on-device smartphone/laptop inference within 5 years, reducing the need for massive centralized compute.
10 Aug 2026 · 59m · 29 chapters
Sid Sheth (d-Matrix) discusses how AI is driving a “premium token economy” and why some companies pay 10x more for low-latency inference.
6 Aug 2026 · 51m · 18 chapters
How BMC Helix uses agentic AI to automate IT service management (“service ops”) by detecting issues from telemetry, performing root-cause and impact analysis, and generating economical remediation…
3 Aug 2026 · 59m · 27 chapters