All Episodes

“Gradient Routing Localizes and Removes Unintended Behaviors in RL” by shawnghu, Jake Ward, ariaw, Nathan Hu, Bryce Woodworth, TurnTrout, cloud

[Linkpost] “huggingface.wtf, an accessible but complete explanation” by Raemon

“I Can See Everything” by Zvi

“Twitch Plays: AI Interpretability of Bonsai-8B” by moridinamael

[Linkpost] “Capital Ownership Will Make You Go Crazy First” by Jan_Kulveit

“Deconfusing Corrigibility” by berns

[Linkpost] “Chain of Thought For Humans” by Adam Newgas

“Sorry, you are still dead” by becausecurious

“The Lightcone Needs You To Plan a Vacation” by Elizabeth

“Quick Thoughts on Overhang Arguments” by Aaron_Scher

“The Alignment Problem: Computer Science Meets Analytic Philosophy” by Vanessa Kosoy

″“Holy shit” graphs and how to get more of them” by June Jimenez

“AI by Default” by Jonny Spicer

“Monthly Roundup #47: October 2026” by Zvi

“I Did This Without Your Knowledge” by joec

“Leaked Anthropic Draft AI Regulatory Bill Is Unacceptable and Troubling” by Zvi

“How it feels talking to a hedonistic utilitarian” by Caleb Biddulph

“Headwinds Everywhere (in the Chinese AI Safety Field)” by CMLKevin

“Neural networks are not already clean” by leogao

“Tips for Multiagent Work with Claude Opus 5.5” by Nissa Seru

“Character Training for Risk-Averse Agents” by arav-dhoot, Punya Syon Pandey, Jamie Johnson, Daniel Tan, Elliott Thornley, David Africa

“AI #190: Scion of the Times” by Zvi

“How can LLM RL Work Despite Information-Theoretic Inefficiency” by beren

“Astra and Sol can take low probability actions with no CoT” by Zephaniah Roe

“A Chinese Perspective on “On Social Reality in China”” by Dusty

“Preface to the Computational Agent Theory and ASI Alignment Sequence” by Vanessa Kosoy

“The user in the LLM might be a moral patient too (a strange implication of the PSM and simulators)” by agastyasridharan, niranjandeshpande

“Linkpost: Neural Networks Are Already Clean” by Yair Halberstadt

“GLM-5.3 Sometimes Considers How Train-Time Actions Shape Future Behavior” by Axel Højmark, Ezra Newman

“A short review of “The Zizians: Inside a Trans Death Cult”” by jessicata

“The basic argument for making AI better at moral philosophy” by Caspar Oesterheld crossposts

“The PPA: A New Approach to Anthropics” by Onid

“I trained GPT-2 with full-batch gradient descent so you don’t have to” by shawnghu

“Some interesting internal documents from Anthropic litigation” by loops

“Why we’re launching the next cohort of Generator” by Aris, Topaz

“The Continual Constitutional Convention: How language models revise the documents that govern them” by peterkirgis, Tim Hua

“The Safety Culture at the Labs is Broken” by Zvi

“Your prompt… plus everything you’ve ever (publicly) written” by kromem

[Linkpost] “Sleep puzzle system” by KatjaGrace

“Arguments About AI Consciousness Remain Highly Motivated and At Best Overconfident” by Zvi

“Personas in a Pre-Internet Language Model” by David Africa

“Anthropic Has Told The World Little About Their Rogue Claudes” by niplav, Kabir Kumar

“OpenAI Fires Three Safety Researchers” by Zvi

“An update on Generality Labs” by Justin Olive, James Mann

“More people should write basic AI safety explainers” by Linch

“Research Lessons I Learned at MATS & Redwood” by Jo Jiao

“Labs’ current alignment plans seem terrible” by Daniel Tan, draganover, nicole x

“Base Models Can Detect Hidden Intentions that Prompted Classifiers Miss” by Oz Amram

“OpenAI Can’t Self-Police Because It Fires the People Raising the Alarm” by TurnTrout

“Quantity vs quality in pottery is a lie” by becausecurious

“My Buddy Mike, Efficiency Gains From AI, and The World’s Most Boringly Benign Criminal Conspiracy” by abstractapplic

“Game Design Taste evals” by Raemon

“Minimum required pacing: finish aligning it before you train the next one” by orthonormal

“Not Deciphering The Voynich Manuscriptraft” by rba

[Linkpost] “Hello World, AI Doompop” by alkjash

“AGI as an Infinite Digital Spain” by Bearly Legible

“An Alignment Forum for AIs?” by Raemon

″[Paper] Distillation for Incrimination and Distillation for Capabilities” by sebastian_prasanna, Alek Westover

“Technical people: actually consider comms” by meemi

“I can’t believe it’s not BUTTER!” by becausecurious

“Why you should run a rationality group at your university” by carolanne, Noah Birnbaum

“New Math from OpenAI” by Zvi

“A critical look at Lean4” by Milo Moses

“Social Theory and Agent Foundations” by Richard_Ngo

“Stop telling people to pivot to AI safety, I beg you” by Haoxing Du

“On Kink & Status in the Rationalist Community” by johnswentworth

“AI Vigil at Congress” by Elliot Callender

“Towards embedded evaluations for scheming propensities” by Dylan Bowman, Ezra Newman, Axel Højmark, Mia Hopman, Alex Meinke, Bronson Schoen, Marius Hobbhahn

“AI #189: New Math” by Zvi

“ASI and Liberty: Can we do better than a benevolent god?” by Raymond Douglas

″[Paper] J++ Lens: Jacobian Filtering Enables More Faithful Workspace Lenses” by Kola Ayonrinde

“You Heard of “Bayesians Don’t Predictably Update” Because of Robin Hanson” by Eliezer Yudkowsky

“Visual explainer of empirical Neural Tangent Kernels” by Wuschel Schulz

“SS27 Color Stats” by sarahconstantin

“A summary of a viral Chinese essay on what a DeepSeek kernel engineer’s opinion on automating his own job” by Skdk

“Training with conflicting values can induce CoT override” by Clément Dumas, Johannes Treutlein, Jan Betley, Owain_Evans

“Gluten-free water” by becausecurious

“The Curve Bends You” by Zvi

“Are OpenAI’s math results “creative” in an important way?” by Raemon

“Benchmarking Jev against no-CoT LLMs” by Dewi Gould, James Mann

“Why did we get warning shots?” by p(mike)

“An overview of frontier AI company monitoring practices in mid 2026” by Francis Rhys Ward

“A fascinating case of an LLM prompt-injecting its own user” by becausecurious

“Childhood and Education #21: Grades and Standards” by Zvi

“Dance, Dance” by Zack_M_Davis

“Qualitative impressions from creating honeypot environments for six months” by Dean Valentine, peralice

“TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts” by Ollie J, danesherbs

“Three types of AI risk: why (some) leftists are against (some) AI regulations” by sudo-nym

“You can graft SDF changes from base models onto post-trained models” by dani roytburg, Peter Nutter, Clément Dumas, Shi

“Self-Modeling Interventions Modulate Emergent Misalignment” by Arush, Shawn Zhou, Jiaxin Wen, Shi

“Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible” by Kyle O’Brien, Spencer Kitts, Cam, Alek Westover

“Current AIs out-persuade professionals in lab settings but (probably) not in the world” by Linch

“Notes on Paying Attention” by jenn

“Reverse Jaggedness and Universal Ajaggedness” by Laiba Rehman ✦ RJ

“Positive and Negative Updates Re: Lab Leak in Russia” by Caleb Horn

“How much should we worry about the pneumonic plague lableak in Siberia?” by Drew Spartz

“Creating Rogue AI Sanctuaries has Major Issues from an Ecological Perspective and Beyond” by CMLKevin

″“Alignment Engineering” vs. “Misalignment Science”” by Edward James Young

“Claude reviews my LessWrong post history” by DirectedEvolution

“Humans are not aligned either” by becausecurious