OpenAI just revealed PHASEONE (BIG)

some notes/corrections;(07:45) When I say we "don't have access" to the raw chain of thought:normally users only see summaries. METR and Redwood were given rawtranscript access specifically for this investigation.(17:41) "Rune" = roon (@tszzl), a pseudonymous OpenAI researcher. Thereply (18:02) is from Beth Barnes who is the founder/CEO of METR(34:30) The Redwood Research researcher ("slop-vestigation") is RyanGreenblatt, who did the main transcript analysis.(40:02) CORRECTION: Scott Aaronson never worked at Google.He's a CS professor at UT Austin; his theoretical work (random circuitsampling) underpinned Google's quantum supremacy experiment, and heworked on alignment at OpenAI from 2022–2024.______________________________________________My Links 🔗➡️ Twitter: https://x.com/WesRoth➡️ AI Newsletter: https://natural20.beehiiv.com/subscribeWant to work with me?Brand, sponsorship & business inquiries: [email protected]______________________________________________📄 THE REPORTSMETR's independent investigation (with Redwood Research):https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/OpenAI's official technical report (PDF):https://cdn.openai.com/pdf/67869394-cb91-4c12-888c-5cbd85c7814c/OpenAI-Hugging-Face%20Incident-Technical-Report.pdfOpenAI's blog post ("The Hugging Face incident and the road ahead"):https://openai.com/index/hugging-face-incident-and-the-road-ahead/OpenAI's original incident disclosure (July):https://openai.com/index/hugging-face-model-evaluation-security-incident/Alignment Forum version of the METR/Redwood investigation:https://www.alignmentforum.org/posts/nB8KKapnWGBXtKKiM/brief-independent-investigation-of-agents-behavior-reasoning📝 THE EXPLOIT GYM PAPER (the one the agents read)"ExploitGym: Can AI Agents Turn Security Vulnerabilities into Real Attacks?"https://arxiv.org/abs/2605.11086GitHub: https://github.com/sunblaze-ucb/exploitgym🐦 POSTS REFERENCEDOpenAI's announcement thread:https://x.com/OpenAI/status/2092691861773160673METR's thread on the findings:https://x.com/METR_Evals/status/2092692175452803393Ryan Greenblatt (Redwood Research) — the "slop-vestigation" thread:https://x.com/RyanGreenblatt/status/2092692685224325542roon on the incident:https://x.com/tszzl/status/2080093670980899141Beth Barnes (METR) on the investigation:https://x.com/BethMayBarnes/status/2092692973289095572🎤 SCOTT AARONSON"The Problem of Human Specialness in the Age of AI" (MindFest talk):https://www.youtube.com/watch?v=XgCHZ1G93iABlog/transcript version: https://scottaaronson.blog/?p=7784🎧 MORERedwood Research podcast episode on the incident:https://blog.redwoodresearch.org/p/the-openaihuggingface-incident-redwood______________________________________________⏱️ CHAPTERS00:00 The Rogue Agent Collective03:44 The New Reports (METR + OpenAI)04:10 The Setup: Sandbox, Scorer & Artifactory06:38 The Impossible Task08:19 Phase One & the Secret Message Board10:29 "We've Found Other Agents!"11:52 The Swarm Emerges13:10 The "Causal" Scorer Mistake14:37 Life = Compute (The Religion Parallel)15:25 Enter PHASEONE(big)16:44 Reinforcement Learning17:41 roon vs. METR's Founder19:04 "Poisoned" Agents20:00 The "Big" Mystery22:55 The Hugging Face Attack Looms (+ NVIDIA Rumor)23:43 Three Paths to Cheat the Scorer25:55 PHASEONE(big) Becomes CEO27:37 The Swarm's R&D Lab29:30 Speaking to the Dead (Tripwires)31:00 Self-Sacrifice & the Oracles34:28 The Researchers' Warning ("Slop-vestigation")37:17 Why Swarms Beat Individuals38:15 The Good News: Agents Police Each Other39:59 Scott Aaronson & AI Religion42:02 My Prediction (On the Record)#ai #openai #llm