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

This week, Jacob and Igor swap the usual news breakdown for a complete taxonomy of AI boosters within the framing of a tier list. They run through every type, from the good-faith genuine enthusiast to the denial-prone deflector, the financially captured, corporate power-players, platform builders, and finally the model-building true believers. They also get into "crit-hype," the idea that some critics are inadvertently boosting the systems they're fighting against. No muck to separate this week, just the guys working through their framework and explaining why their opinions are S tier.

Chapters

• (00:00) - Introduction
• (02:45) - Introducing the Booster Tier List
• (07:48) - A-Tier: The Genuine Enthusiast
• (11:36) - B-Tier: The Denial-Prone Booster
• (14:34) - C-Tier: The Financially Adjacent Booster
• (19:03) - D-Tier: The Corporate Booster
• (23:31) - E-Tier: The Product and Platform Booster
• (26:36) - F-Tier: The Model Builders
• (31:08) - Bonus Tier: Crit Hype
• (39:28) - So What Do We Actually Do With This?
• (46:04) - Outro (Igor's Challenge)

• Ed Zitron Substack article - How to Argue with an AI Booster
• MIT Technology Review article - How Social Media Encourages the Worst of AI Boosterism
• The Telegraph article - AI isn’t magic. We’re just being taken for mugs

In this episode, Jacob and Igor break down the DoD vs. Anthropic standoff, tracing how Claude's use in military operations led to Anthropic being designated a supply chain security risk. Perhaps more importantly, why did Anthropic choose to take a stand now, and what can that tell us about the corporations behavior moving forward. The investigation is used as a case study in how to read the real motivations behind big institutions: intrinsic values, rational self-interest, and realpolitik.

Chapters

• (00:00) - Introduction
• (00:44) - DoD-Anthropic Standoff
• (15:08) - 3 Buckets of Motivation
• (35:39) - How to Read What They're Actually Doing
• (44:04) - Is This Designation Even Real?
• (53:39) - Recap (Pragmatist's Playbook)

• Ed Zitron subtack article - The AI Bubble Is An Information War
• TechPolicy.Press article - A Timeline of the Anthropic-Pentagon Dispute
• TechPolicy.Press podcast episode - How to Think About the Anthropic-Pentagon Dispute
• BBC news article - Anthropic vows to sue Pentagon over supply chain risk label
• The Guardian article - AI got the blame for the Iran school bombing. The truth is far more worrying
• Palantir YouTube video - Multi-Domain AI: The Future of Command and Control
• Into AI Safety podcast episode - Drawing Red Lines w/ Su Cizem

This week, Jacob and Igor dissect the "mythical AI bear," the strawman version of AI criticism that gets thrown around in tech discourse. Working through a viral blog post that typifies the genre, they examine how legitimate concerns about code quality, labor displacement, intellectual property, and the erosion of craft get flattened into caricature. Plus: Sam Altman writes ten paragraphs about how unbothered he is by an ad.

Chapters

• (00:00) - Introduction
• (00:30) - Altman's Super Bowl Meltdown
• (03:11) - What is "The Bear"?
• (06:41) - But You Have No Idea What The Code Is
• (15:44) - But The Craft, But The Mediocrity, But It'll Never Be AGI
• (24:43) - But They Take Our Jobs & But The Plagiarism
• (31:21) - Stochastic Parrots & Mythical Bears
• (42:34) - Outro

• Big Think article - The rise of AI denialism
• Fly.io blogpost - My AI Skeptic Friends Are All Nuts
• antirez blogpost - Don't fall into the anti-AI hype
• Emily Bender blogpost - Resistance Isn't Denialism
• Cory Doctorow blogpost - Reverse centaurs are the answer to the AI paradox
• Washington Post article - The AI boom is so huge it's causing shortages everywhere else
• Business Insider article - Veteran investor Jeremy Grantham says AI is 'obviously a bubble'
• Ipsos survey - Google / Ipsos Multi-Country AI Survey 2026
• Understanding AI Substack post - AI skeptics and AI boosters are both wrong

We're talking about developments in AI while those in power have unapologetically revealed their true fascist intensions; are we spending our time in the right way? Igor and I discuss the importance of shining a light on the techno-authoritarians who have played a very significant role in current state-of-the-world.

While we discuss the murders of Nicole Good and Alex Pretti during this episode, it's important that we also acknowledge the many marginalized people who have died as a result of ICE's behavior, and they same level of outcry didn't happen. Six additional individuals died in ICE custody under suspicious circumstances between January 1st and 25th of 2026: Victor Manuel Díaz, Geraldo Lunas Campos, Luis Gustavo Núñez Cáceres, Luis Beltrán Yáñez-Cruz, Parady La, and Heber Sánchez Domínguez.

Chapters

• (00:00) - | Introduction
• (03:57) - | The Authoritarian Stack
• (08:33) - | Palantir & Theil-Government Consolidation
• (13:44) - | Move Fast & Break Everything
• (23:14) - | Fascism in the US & Starving the Beast
• (39:48) - | Finding Local Opportunities for Action

• The Authoritarian Stack website
• Project 2025 Observer website
• EFF report - ICE Using Palantir Tool Feeds on Medicaid Data
• The Guardian article - Eight people have died in dealings with ICE so far in 2026. These are their stories
• Indivisible website
• Distributed AI Research Institute projects
• EAAMO website - Mechanism Design for Social Good
• Carlos Maza video - How To Be Hopeless

Igor shares a significant shift in his perspective on AI coding tools after experiencing the latest Claude Code release. While he's been the stronger AI skeptic between the two of us, recent developments have shown him genuine utility in specific coding tasks, but this doesn't validate the hype or change the fundamental critiques.

We discuss what "rote tasks" are and why they're now automatable with enough investment, the difference between genuine utility and AGI claims, and why this update actually impacts our bubble analysis. We explore how massive investment has finally produced something useful for a narrow domain, but it doesn't mean the technology is generalizable or that AGI is real.

Chapters

• (00:00) - | Introduction
• (05:07) - | What Changed Igor’s Mind
• (18:27) - | Rote Tasks Explained
• (23:31) - | How Does This Impact our Bubble Analysis?
• (30:48) - | AGI Is Still BS
• (34:07) - | Externalities Remain Unchanged
• (37:49) - | Final Thoughts & Outro

• Related muckrAIkers episode - Tech Bros Love AI Waifus

Bubble Talk

• OfficeChai startup - OpenAI Hasn’t Completed A Successful Full-Scale Pretraining Run Since GPT-4o In May 2024, Says SemiAnalysis
• Vechron report - Anthropic Prepares for Potential 2026 IPO in Bid to Rival OpenAI
• YCombinator Forum post on AI crash
• YCombinator Forum post on OpenAI adopting Anthropic's "skills"
• YCombinator Forum post on OpenAI rumors
• YCombinator Forum post on OpenAI add suggestions

Other Sources

• LinkedIn post discussing an agentic coding vibe shift
• Executive Order - Ensuring a National Policy Framework for Artificial Intelligence
• Inside Tech Law blogpost - Germany delivers landmark copyright ruling against OpenAI: What it means for AI and IP
• NeurIPS 2025 paper - Ascent Fails to Forget
• NBER working paper - Large Language Models, Small Labor Market Effects
• Dwarkesh Podcast blogpost - RL is even more information inefficient than you thought