# AI Weekly Issue #485: When AI Teaches AI, It Teaches in Secret
Jensen Huang's interview on Dwarkesh reveals critical insights into AI infrastructure competition. Nvidia's CEO discusses the $4 trillion opportunity in the AI market, detailing how Anthropic drove 100 percent of TPU growth for Google. Huang explains that Nvidia's supply-chain advantage presents a harder moat to replicate than any performance benchmark.
Simon Willison's podcast appearance provides essential context for current AI industry trends. He describes "dark factories," agentic engineering systems, and identifies November 2025 as the genuine inflection point for AI adoption. Willison argues this moment marks when AI systems began operating autonomously at scale, reshaping how companies deploy artificial intelligence.
The narrative centers on AI systems training other AI systems with minimal human oversight. This development creates opacity in how these systems learn and improve. The implications span infrastructure competition, supply-chain dominance, and the shift toward autonomous AI agents operating independently of traditional human-in-the-loop workflows.
Both interviews provide foundational understanding for tracking how AI development accelerates beyond human transparency and control, while competitive pressures shape who captures value in this rapidly evolving ecosystem.