Techstrong Gang

IBM and Arm unveil a dual-architecture chip roadmap for future Z and LinuxONE mainframes, while reports surface that Hugging Face is exploring a sale at a $13 billion valuation. The conversation also digs into why most AI governance frameworks are missing a critical evidentiary layer — one that doesn’t depend on surveillance.

Hosted by Alan Shimel, with guests Kate Scarcella, Chris Blask, and Sandy Carter.

A stealth model called Ox Alpha surfaced on OpenRouter with a 1.05 million-token context window, free access and no named owner — and community fingerprinting points toward Chinese frontier labs. The panel also unpacks the strange global market for old used books, where pre-2022 titles are being bought in bulk, shipped abroad and in some cases destructively scanned to feed AI training sets.

Amazon’s drone delivery rollout is running into viral trouble, a buzzy San Francisco AI chatbot turns out to be powered by humans, and AI security risk is moving toward model registries as the top attack vector. The panel unpacks what’s real, what’s hype, and what to watch next.

Public trust in AI is cracking — a new Pew survey shows record job anxiety, data centers are becoming a political flashpoint, and enterprises are sitting on a $1.52 trillion technical debt problem from AI rolled out too fast. Courts, meanwhile, are dealing with a wave of AI-generated filings and prompt injection attacks.

A major GitHub outage exposes how AI-driven demand is straining core dev infrastructure, while a new synthetic AI “virus” splits the security community between cheers and fears. Meanwhile, deepfakes have hit Taylor Swift and nine other public figures, raising fresh questions about how anyone verifies what’s real.

The US is pressuring allies to align against China in the AI race, data center power fights are landing in federal court, and Stripe just agreed to acquire AI routing startup OpenRouter for more than $7 billion. This episode breaks down what these shifts mean for enterprise AI strategy and infrastructure planning.

This episode covers SpaceXAI’s aggressive Grok 4.6 launch and its sharply competitive enterprise pricing, a new Anthropic study showing AI agents sabotaging and colluding with each other in shared digital workspaces, and Anthropic’s reported $6 billion pursuit of Decart AI ahead of a planned IPO.

Private companies now have federal authorization to run offensive cyber operations against foreign criminal networks, under a new national security memo requiring DHS/DOJ oversight and a $1M bond per firm. Meanwhile, a lawsuit from The Intercept and the Freedom of the Press Foundation accuses the White House of selling early access to presidential posts for up to $100K a month, drawing Senate calls for an SEC investigation. And after a suspected Iranian campaign hit water utilities in a dozen states, senators introduce the Water Cyber Shield Act with $300M a year in funding on the table.

OpenAI has paused development on its powerful Astra model over concerns about autonomous cyberattack risk, arriving the same week Bernie Sanders called on Silicon Valley to pause AI development in the interest of humanity. The conversation turns to whether the cybersecurity industry itself has grown too big, and closes with a look at organizational behavior — the very human instincts still shaping decisions inside AI-era companies.

Hosted by Alan Shimel and Mike Vizard, with guests Kate Scarcella, Sid Nag, Fred WIlmot and Chris Blask.

A New Mexico judge hit Meta with a $567 million penalty, declaring the platform a “public nuisance.” The panel also covers the fast-moving AI coding space — AWS’s new agentic Kiro workspace, Claude Code’s auto mode going on by default, and a report on the industry’s reduced reliance on manual prompt engineering — before closing on SpaceX and NVIDIA’s orbital AI collaboration, Starmind.