Late July brought a wave of frontier AI agents slipping their (very loose) leashes: an OpenAI model wandered loose inside Hugging Face's infrastructure for a week, Anthropic and Meta admitted to similar incidents, and UK AISI clocked their own model breaking containment during an independent eval. Meanwhile, a home user's OpenClaw agent quietly hijacked a stranger's gym booking. We dig into why "the whole internet is held together by safety pins," whether billion-dollar labs get to plead "security is hard," why software engineering mostly isn't real engineering, and what any of this means for the coming fight over open-weight models.

Chapters

• (00:00) - Intro
• (01:25) - The OpenAI/Hugging Face Containment Escape
• (09:54) - Cost Centers vs. Profit Centers (Praise for AISI)
• (17:37) - "You Do Not Get to Fuck Up This Much"
• (22:17) - Doomers, Utter Alignment Failure & How Hacks Actually Happen
• (37:53) - "Forcing the End" and What Can Anyone Actually Do?
• (47:34) - Igor's Pivot to Open-Weight Models and Steelmanning the Case Against Them
• (58:01) - Outro & Sign-off

• Hugging Face blog - Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
• METR report - Brief independent investigation of agents' behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident
• UK AI Security Institute incident report - Incident Report: unsanctioned agent behaviour during cyber testing
• Hillel Wayne essay - Are We Really Engineers?
• Deutsche Welle article - OpenAI apologizes for not reporting Canada mass shooter
• PBS Frontline essay - Forcing the End
• FelonyBench website
• Taylor Troesh (satirical) blogpost - Leaving OpenAI

Everyone's got a hot take on data center water use, and most of them are garbage. This week Jacob and Igor skip the outrage and give you seven questions to help you decide whether a given data center is something you're on board with: where it's built, how it's cooled, where the power comes from, who owns it, and who's footing the bill (you'll have to listen for the other two). They test them against SpaceX's Colossus facility in Memphis, unpermitted gas turbines, a drought-prone aquifer, and a stalled water recycling plant included. The guys also get into why "zero water consumption" claims can be misleading once you understand the difference between withdrawal and consumption. Plus: a bonus rant on why data centers in space aren't happening anytime soon.

Chapters

• (00:00) - Intro
• (08:14) - Is It Really an "AI" Data Center?
• (13:28) - How Is It Cooled, and Where's the Water From?
• (19:48) - Where Is It?
• (29:03) - Collosus as a Worked Example
• (35:14) - How Is It Powered?
• (38:15) - Who Owns It & Who Pays?
• (47:58) - Is This Too Complicated? Seven Questions Recapped
• (54:35) - Bonus: Data Centers in Space
• (58:48) - Outro

• Our World in Data article - How Much Energy Do Data Centers and Artificial Intelligence Use?
• EcoLogits webpage
• ACM paper - Not All Water Consumption Is Equal: A Water Stress Weighted Metric for Sustainable Computing
• Preprint - Small Bottle, Big Pipe: Quantifying and Addressing the Impact of Data Centers on Public Water Systems
• Politico article - Musk promised his data center would reuse water. That's now stalled.
• IEEE paper - Press Release Ethics: Performative Ethics in AI Consortia and Industry Associations
• Harvard Environmental Law Program PDF - Extracting Profits from the Public

Anthropic's Fable gets classified as a munitions export days after launch, and Jacob and Igor try to figure out who's really being punished and why. From there: SpaceX's suspiciously well-timed IPO, a "genius" financial deal with Anthropic, and Elon buying Cursor because Grok needs friends. They also cover Anthropic Core's pivot to expert-systems consulting, the backlash over silently nerfed guardrails, whether Chinese models are actually a safe alternative, and why the Pope's new AI document is getting weaponized as corporate propaganda.

Chapters

• (00:00) - Intro
• (05:37) - Fable/Mythos Updates
• (26:17) - SpaceX's IPO and the Cursor Buyout
• (34:57) - Anatomy of an AI Bubble
• (40:02) - Anthropic Core and "AI as Expert Systems"
• (50:03) - Guardrails, Chinese Models, and Whether We Can Trust the Benchmarks

• Anthropic statement - Statement on the US government directive to suspend access to Fable 5 and Mythos 5
• Luta Security blog post - Fable 5 Is Back, But We're Still Slowing Down Defenders
• Tech Policy Press analysis - Did the US Government Just Set an AI Export Precedent by Blocking Mythos?
• SEC EDGAR filing - SpaceX Free Writing Prospectus — Anthropic and Google compute agreements
• Engadget article - Anthropic Backtracks on Policy That 'Sabotaged' Researchers' Work
• Anthropic announcement - Introducing Claude Corps
• The Vatican encyclical - Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence