All Episodes

“Stopgap Measures to Address Immediate AI Security Threats” by Andrea_Miotti, Gabriel Alfour

“The Preference Cascade Is Only Getting Started” by Zvi

“The Horse” by Character#2736

“The Game is Set for a Targeted Memetic Attack on the AI Safety Community” by keltan

[Linkpost] “Three Hackers used Opus 5 to Hack Into OpenAI’s Core Codebase [WSJ]” by Linch

“Deep recurrent models are less robustly CoT-monitorable than normal CoT models in a toy setting” by Nick Kuhn, Alek Westover

“Superintelligence this Christmas” by Alexander Gietelink Oldenziel

“Grantmakers aren’t afraid to die” by dan.parshall

[Linkpost] “Pacing the Frontier: A Framework & Research Agenda” by CharlesD, technicalities, Raymond Douglas, Nowe Moore

“plzdontkillus Fellows Got ~2M AI Safety Views, Not 21M” by Josh Thorsteinson

“Obstacles to the scalable oversight of auto-alignment research” by Sam Martin, Dewi Gould, Cameron Holmes, Jacob Pfau

“AI #186: The World Takes Notice” by Zvi

[Linkpost] “callcongress.ai – the basic action US residents can take to help with AI risk” by Ruby, haglobah

“Did Galileo mistake Saturn’s rings for Jupiter’s Moons?” by Alfred Harwood

“Model organisms (sometimes) confess their misalignment when offered a deal” by Mark Keavney, Francis Rhys Ward

“Reducing the Resource Gap Between Lab and External Safety Researchers” by Alexandra Narin, Kyle O’Brien, Puria

“For Love of the Lightcone, Don’t Partisanize AI Safety” by DanB

“AI as orderly evacuation vs stampede” by Richard_Ngo

“Don’t trust Lean4 alone” by Milo Moses

“Microsoft AI’s “Humanist” CoC” by Stephen Martin

“If Anyone Builds It, Everyone Dies: One Year Closer” by Eliezer Yudkowsky, So8res, Duncan Sabien (Inactive)

“Trump Goes Full Hoax on AI Existential Risk” by Zvi

“Phantom transfer works via extremely subtle semantic cues” by Helena Casademunt, Anton de la Fuente, Josh Engels, Arthur Conmy

“We Should Assume We Have One Chance At AI Legislation” by Jamie Joyce

“Self Inoculation” by epicurus

“Shallow Beliefs: Midtraining does not inoculate against EM from reward hacking” by Jozdien, Julian Stastny

“Is METR A Meaningful Check On Anthropic?” by SE Gyges

“The Bad Guy With An AI Named Claude” by Zvi

“Quick notes from teaching technical profiles how to talk in public” by Camille B.

″[Cross-post] Palisade Podcast episode “How to Actually Influence AI Policy (No Law Degree Required) — with Matthew Lipka”” by davekasten

“Inoculation Midtraining with Learned Neologisms” by Kyle O’Brien, Edward James Young, Puria, Nathalie Kirch, Cam, Tomek Korbak, David Africa

“Cooperation with AIs seems to be a low-hanging fruit for better evals” by Clément Dumas

“Astra appears to perform belief-propagation-like inference without CoT” by MBaert

“OpenAI President Brockman says HuggingFace incident model had not been alignment-trained” by Caspar Oesterheld

“Model Weight Exfiltration Seems Overrated” by Vaniver

“OpenAI Says It’s Not Responsible for the Leading the Future Super PAC. But Only Its Employees Seem to Believe That.” by garrison

“Op-Ed: I Worked at Google DeepMind. You Should Listen to the Warnings About AI” by TurnTrout

“Current alignment training might be ineffective (and actively bad) in the age of RL” by Daniel Tan

“There is a channel to 900M weekly users. What goes in it?” by Charbel-Raphaël

“I am refusing to work on Cloud TPUs” by Yair Halberstadt

“Deployment” by Nina Panickssery

“Yet another concerning result on Astra’s no-CoT capabilities” by Christine Corry

“Alignment & Succession: The Two Bars of Alignment” by L Rudolf L

“Anthropic and OpenAI haven’t published a plan for aligning superintelligence” by Zephaniah Roe

“Consider how your global governance proposal is different from the EU Code of Practice” by David Matolcsi

“Brand New AI Solves a Millennium Prize” by Zvi

“Teleoperated Humans” by jefftk

“Mitigating Reward Hacking as Institutional Design” by beren

“The Talker Does Not Control The Doer (in Current AIs)” by Eliezer Yudkowsky

“GPT-6-Astra Can Do Ambitious Things” by Zvi

“On the origins of altruistic behaviour in the Hugging Face incident” by Fernando Rosas

“AI takeover is obviously bad, whether or not everyone dies” by Caleb Biddulph

″“Pacing the Frontier”: Dario Amodei Essay Linkpost” by fluxxrider

“OpenAI-HuggingFace: A Reproduction & Lessons for Alignment Testing” by Stewart Slocum

“Some ways AI could kill us all” by Ruby

“Post-AGI, we are all jobless aristocrats” by djbinder

“Caroline Ellison has joined Manifund” by Austin Chen, Carol N

“Astra’s no-CoT limits track speculative depth, not step count” by MBaert

“How a cold email got the Finnish government to respond on superintelligence regulation” by Josh Thorsteinson

“CoT controllability evals seem very under-elicited” by Jozdien

“Jacob Coxon Warns of Human Extinction and Triggers a Preference Cascade” by Zvi

“The Locally Optimal Discursive Posture” by deanball

“Astra is much better at reasoning with filler tokens than previous models” by Dylan Xu, SebastianP, Alek Westover

“To Thine Own AI Be Truthful: emergent misalignment in alignment research” by lumpenspace

“First Bill Introduced to Ban Superintelligent AI” by Andrea_Miotti

“What the Pro-Democracy Movement Knows About Quitting in Protest” by Maxwell Love

“Categorical taboos are much better than threshold taboos: neuralese edition” by Linch

[Linkpost] “Doom as a bad method not a utopia trade-off” by KatjaGrace

“The Geometry of Nonergodic Composition” by Adam Shai, Kyle Ray, Paul Riechers

“So where’s this AI thing going?” by Seth Herd

“An operationalization of opaque serial depth” by ryan_greenblatt, frisby, Alek Westover, Lukas Finnveden, Alexa Pan, Julian Stastny

“Proposal for tracking the effects of architecture on monitorability” by ryan_greenblatt, Alek Westover, Lukas Finnveden

“AI #185: Preference Cascade” by Zvi

“Astra can do a concerning amount with no chain of thought” by Neel Nanda

“One Billion Hemmingways” by Girard Dorney

“Can a superintelligence do THAT?” by Eliezer Yudkowsky

“Recommendations for People Getting into Technical AI Governance Research” by Aaron_Scher, yams, peterbarnett, Naci Cankaya

“Personal statement on joining the OpenAI board” by paulfchristiano

“GPT-6 Astra: The System Card, Alignment and What Comes Next” by Zvi

“Self Hosting” by Tomás B.

[Linkpost] “Estimating GPT-6 Astra’s no-CoT Time Horizon” by Francis Rhys Ward, Dewi Gould

“Exploration Hacking in AI Debate: Initial Empirics and Generalisation Splitting” by Jason R Brown, Nathalie Kirch, Joschka Braun, hyannakoudakis, David Lindner

“A Conceptual Framework for Reasoning about Exploration Hacking” by Jason R Brown, Nathalie Kirch, Joschka Braun, hyannakoudakis, David Lindner

“GPT-6 Astra can do a lot of multi-hop reasoning without chain of thought” by RohanS

“Pausing AI ASAP is preferable to agreeing to pause at some future time” by Connor Williams

“How good are slop-vestigators?” by Hasan Baig, OscarGilg, Hamzah

“Training on probes: What’s going on” by Charlie Steiner

“OpenAI have solved the Navier-Stokes Problem with a substantially more powerful model than Astra.” by fluxxrider

[Linkpost] “Frontier models still hack on simple variations of alignment evals from early 2025” by Dean Valentine

“Psychological Support for AI Safety Researchers Is Neglected and Easy to Provide” by Ihor Kendiukhov

“Astra Is Hard to Monitor” by Zvi

“Contra Piper on When Conversation Is Possible” by Zack_M_Davis

[Linkpost] “Where are the token-level LLM kill-switches?” by beyarkay (Boyd Kane)

“Machine Organizations” by Vaniver

“Dear God, Please Don’t Resign In Protest” by Kabir Kumar

“The Scramble: getting in position to pace the frontier” by Peter Wildeford

“The Magnus Challenge” by Taylor G. Lunt

“Most Anthropic equity that will ever be used for longtermist philanthropy should be sold ASAP and reinvested” by Zach Stein-Perlman

“The Wormtongue Test” by Drake Morrison

[Linkpost] ”“An Alien Mind” from OAI chief scientist seems newly cautious on alignment” by Seth Herd