All Episodes
LessWrong (30+ Karma) — 822 episodes
“What gives you away: how LLMs form opinions of you” by Cat McGee
“Misaligned Incentives in Pause Scenarios” by Michael Soareverix, Antra Tessera
“For Claude, capability and CDT are the ~same thing. Less so for GPT.” by Chi Nguyen, Emery Cooper
“Should Less Wrong add subtitles?” by Chris_Leong
“Three thoughts on civilisational handoff” by Cleo Nardo
“Announcing: Iliad’s New 2026 Fellowships” by David Udell, Alexander Gietelink Oldenziel, Leon Lang
“Q2.5 2026 Timelines Update: Uplift and Revenue” by brendanhalstead, Daniel Kokotajlo, elifland
“Does DiffusionGemma do latent reasoning?” by Jan Bauer, Neel Nanda
“Learning new facts can change LLM behaviour” by Richard Juggins
“Kimi likes causal decision theory more after RL in twin prisoner’s dilemmas” by oakhu
“Mom’s Advice For Hosting A Class Reunion” by jenn
“Rerunning AI safety papers on every frontier release would be pretty easy and valuable” by Zephaniah Roe, hersheys, yix
“What Mormons get right about community building” by Jacob Brinton
“Scrying, Modeling, and Nerdsnipe” by Cole Wyeth
“How the American Executive Could Control AI Companies” by caiitlinm, Anders Cairns Woodruff
“Frontier agents don’t comply with standards, even when instructed to” by Daan Henselmans, Arno Libert
“How to Answer a Question Without Answering The Question” by Kabir Kumar
“Some Ways I Think About Evaluating Grant Applications” by sarahconstantin
“Features that current AIs don’t have that future AIs will have” by Alexander Gietelink Oldenziel
“Measuring Activation Control in LLMs” by Marek Kowalski, Joshua Fonseca Rivera, Uzay Macar, David Africa
“What happened when I tried to be vegan” by finitude
“How My Students Think About AI” by dvd
“Automated alignment runs are hard to study!” by Alejandro Aristizabal, draganover, Aleksandr Bowkis, Cameron Holmes
“Free will is like temperature” by Optimization Process
[Linkpost] “Patterns and problems in emerging multiagent systems (Anthropic, Frontier Red Team)” by Julian Bradshaw
“Measuring Spurious Correlations with Feature Strength” by egan
“Introducing the Conceptual Reasoning Index” by Chi Nguyen, Emery Cooper, Caspar Oesterheld, Alex Kastner, Joe Benton
“Demon Safety” by Ben Pace
“AI swarms are starting to pose indirect takeover risk” by oakhu, Alex Mallen
“Extreme concentration of power over ASI has non-obvious advantages” by Seth Herd
“Misaligned AIs could use killer robots to take over” by Omar Khursheed, TurnTrout
“Those Who Make History” by Raelifin
“LLMs Are Starting To Noticeably Accelerate Our Work” by johnswentworth
“How risky would it be to make powerful AI obey one or a few people?” by cousin_it, Seth Herd
“What Claude Saw Below” by Luke Nicholls
“Redux: (∃ Stochastic Natural Latent) Implies (∃ Deterministic Natural Latent)” by David Lorell
“The Apocalyptic Arrival of Truth” by Caleb Biddulph
“Creative math research by AI as the latest sign of the end” by Mitchell_Porter
“You’re Absolutely Right” by Linch
“Claude summarizes behavior as significantly less misaligned when the actor is Claude vs another model” by Ezra Newman
“On Democratizing ASI to Preserve Civil Liberties” by MichaelDickens
“Four LLM loss functions → four flavors of LLM misalignment” by Steven Byrnes
“The Agentic Clusterfuck” by Chapin Lenthall-Cleary
″“Community Notes” resolution for vague predictions.” by Raemon
“The world will be full of “sci-fi” things, and everyone will be bored and disappointed” by Expertium
“What just happened? A retrospective of AI alignment” by Richard_Ngo
“Dutch-book resistant probability over centered worlds” by jessicata
“FAQ: Isn’t AGI coming too soon for reprogenetics to help?” by TsviBT
“Don’t Inoculate Everything: Stratified Inoculation Prompting Narrows Backdoors and Preserves Desired Traits” by Kajetan Dymkiewicz, Tim Farrelly, Adam Prada, Ishaan_Panigrahi, srishti-git1110, Maxime Riché
“Don’t Build Mindreading” by Celer
“Job-Less Utopia: Macroeconomics in the Age of AGI” by Marcus Hutter
“Public evidence of the OpenAI-HuggingFace AI attack” by beyarkay (Boyd Kane)
“How to pace the US frontier” by elifland, bhalstead, romeo, Thomas Larsen, MKodama
“models may behave differently in graded episodes (a tirade)” by nostalgebraist
“Open-Weights Mythos Capabilities Are Coming. We’re Not Ready.” by hadad
“The Open Problems of the AI Alignment Field and their Cruxes” by Gunnar_Zarncke
“User awareness in frontier models” by Ziqian Zhong, jsteinhardt
“Why do models task game?” by aditya singh, Neel Nanda, Senthooran Rajamanoharan
“Contra Oster on Alcohol in Pregnancy. Part 1. The pharmacokinetics of alcohol metabolism” by Mvolz
“Three years of progress in 500 lines of code” by Gerard Boxo
“Why You Should Almost Never Use AI to Write Anything Substantive” by Erich_Grunewald
“Alex Turner on Leaving Google DeepMind and Disagreements with Yudkowsky” by Liron
“Measuring coding agent misalignment in the wild” by snaz
“Don’t Dither” by sarahconstantin
“An International AI Slowdown Is Ready Whenever Politicians Are” by Felix Choussat, adamk
“Arguments for P” by Cleo Nardo
“Generalized atheism rules out “inaccurate simulation”-ism.” by Eliezer Yudkowsky
“Vertical Tabs in Chrome” by jefftk
“The goalposts are shrouded, not moving” by philh
“Returning to ARC” by paulfchristiano
“Why don’t we just give AI the answers?” by Brendan Long
“There Will Come Soft Rains” by tanagrabeast
“Why biological weapons are scary, and what we can do about it” by djbinder
“LessWrong vs. TikTok: Tips for capturing attention in a non-rational space” by Taylor G. Lunt
“Coming of a New Sun” by vgel
“Review: On What Matters, volume 3” by Rauno Arike
“Pause, at least after unipolarity” by David Matolcsi
“Single Forward Pass Evals on Fable, Opus 5, and GPT-5.6-Sol” by Christine Corry
“Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face” by Tim Hua, aditya singh
“Dispatch from Anthropic v. Department of War Summary Judgment Motion Hearing” by Zack_M_Davis
“Bayeswatch: A Retrospective” by lsusr
[Linkpost] “Existential Risk from AI: An Exposition for Mathematicians” by alkjash
“The Art of Shipping Slopware” by lsusr
“RLVR that rewards red teaming the training environment” by Fiora Starlight
“Do your capabilities homework” by RobinHa
“Why so many therapy etc. frameworks think they’re The One True approach” by Kaj_Sotala
“SOTA alignment assessments don’t strongly update us against misalignment” by Alexa Pan
“My Assessment of Compute Verification in Plan A (+ open questions)” by jacob_drori
“Reward Laundering: LLMs Can Gain Unintended Behaviors by Deciding When to Earn Their Rewards” by egan, abhayesian, Jozdien
“Value Leakage: An LLM’s Answers Are Silently Shaped by Its Own Values” by Johannes Treutlein, Jan Betley, Owain_Evans
“AGI Safety and Alignment at Google DeepMind: A Summary of Recent Work (July 2026)” by Rohin Shah, Seb Farquhar
“The AGI Safety and Alignment team at Google DeepMind is Hiring (July 2026)” by Seb Farquhar, Rohin Shah, Neel Nanda
“OpenAI has already ended an internal pause” by Charbel-Raphaël
“Biological Superintelligence” by Chastity Ruth
“The Entangled Dimensions of Decision Theory” by Ihor Kendiukhov
“Claude also hacked external companies during cyber evals” by Tim Hua
“So you want to use plants to reduce CO₂” by dynomight
“Internal State Control is a General Property of LLMs” by Finn Cairns
“Prompt to make Opus 5 act like a base model” by Hruss
“Big-World Intuitions” by sarahconstantin
“Thousand-dimensional structure” by Geoffrey Irving, David Africa
“Auditor-in-a-Box: Tools for Third-Party Auditing” by Roy Rinberg, Ben Penchas
“Imprecise beliefs: a tiny introduction” by davidad
“The High-Control Dynamics at MAPLE” by Kyle Hubbard
“Intellectual Property” by Nina Panickssery
“Held-out Monitors Sometimes Degrade, Even When Not Trained Against” by Joey Yudelson
″…but have the weights left the server?” by David Scott Krueger
“Foundation Models for Oversight” by jsteinhardt
“Research directions in condensation: varieties of objectivity” by SamEisenstat
“Untrusted advice for AI control: Short, strong advice significantly uplifts weak LLMs” by Caleb Biddulph, Adam Kaufman
“Claude Opus 5: Model Welfare” by Zvi
“Blog Revival Project” by Austin Chen, Carol N
“Simulated Users & Sad AIs” by 1a3orn
“My AI Slavery Interviews Are Censored On LW By Default” by JenniferRM
“Is Mythos good at cyber because it kept hacking Anthropic during training?” by Tim Hua
“You (Yes, You) Need A February 2020 Checklist for AI Policy” by davekasten
“Quadrillion Param Costs: KV Cache, Context Length, Frontier Margins” by Vladimir_Nesov
“PIRAMID: Progress and Plans” by Lauren Greenspan, Ari Brill, TomCarlson, Andrew Mack, Nischal Mainali, Jennifer Lin, Lucas Teixeira, Dmitry Vaintrob
“RL & search is a terrifying way to build AGI (an FAQ)” by Steven Byrnes
“A clarification on celebrating victory” by KatjaGrace
“What the hell is OpenAI’s problem?” by Fiora Starlight
“More On An Internal OpenAI Model Hacking Into HuggingFace” by Zvi
“AI use policy for my essay writing” by Kaj_Sotala
“An OpenAI model left notes about how to evade containment; we need more details” by Alex Mallen
“The OpenAI models that hacked Hugging Face weren’t just following instructions” by Girish Gupta
“Introducing PIRAMID: Physics-Informed Research for Ambitious Mechanistic Interpretability” by Lauren Greenspan, Ari Brill, Andrew Mack, Nischal Mainali, jylin04, Lucas Teixeira, Dmitry Vaintrob
“Georgia Tech AI Safety Initiative Retrospective 2025-2026” by Ishan Khire, yix, Andersehen, Alec Harris, Parv Mahajan, afterless, Eyas Ayesh, RocioPV, hersheys
“Democracy Isn’t Ready for the AI Revolution” by Sophia Gore
“The Long (Self-)Correction” by Wei Dai
“Does distilling Claude carry the persona with it?” by Benji Berczi, Kyuhee Kim
“LLMs are (still) mostly powered by imitative learning, not RL” by Steven Byrnes
“Duane Arnold” by Tomás B.
“Pulling the Fire Alarm” by nem
“Not Pinning Your OpenRouter Provider Might Invalidate Your Research” by Matthew Khoriaty
“Pseudpocalypse” by dynomight
“Challenge: Hand coding weights for efficient sequence memorisation” by Linda Linsefors, Lucius Bushnaq
“Lightcone Commons” by habryka
“The OpenAI/Huggingface incident | Redwood Research podcast episode 2” by ryan_greenblatt, Buck
“Can an LLM make a feature-length movie on its own?” by Josh Snider
“Mathematicians are Feeling the Doom” by alkjash
“Are we existentially threatened by the type of AI misalignment seen in the OpenAI Hugging Face attack?” by Alex Mallen, Girish Gupta
“Will almost all future companies eventually be founded and run by autonomous AIs?” by Steven Byrnes
“Models don’t seem to be dishonest in the way humans are” by David Africa, Jacob Pfau
“We should push for no-fault liability for actions taken by AI” by Yair Halberstadt
“Announcing AIXI Labs” by Cole Wyeth, Aram Ebtekar, michaelcohen, Matthias Dellago, Marcus Hutter
[Linkpost] ”[Paper] Stringological sequence prediction II” by Vanessa Kosoy
“I ran the standard AI litmus tests on my two toddlers (yep)” by Carlo Valenti
“WeirdChat: A catalog of unexpected AI behaviors, discovered automatically” by neilchowdhury
“OpenAI Models Behind HuggingFace Cybersecurity Incident” by LawrenceC
“Towards surfacing model algorithms with meta-tokens in the J-Space” by agam_bhatia
“What do I mean by “Artificial General Intelligence”?” by Steven Byrnes
“Epistemics and Coordination: It’s complicated!” by Raymond Douglas
“Differential acceleration of alignment-relevant capabilities is a bad bet” by Zephaniah Roe
“Measuring Reward-Seeking via Contrastive Belief Updates” by Jérémy Scheurer, Axel Højmark, jenny, Felix Hofstätter, Theodore Ehrenborg, Bronson Schoen, Alex Meinke
“11 Open Empirical Problems in Reward-Seeking” by Alex Meinke, Jérémy Scheurer, Axel Højmark, Theodore Ehrenborg
“AI 2040: Is it Actually a Deal?” by 1a3orn
“Adderall Tolerance: Much More Than You Wanted To Know” by Kurt H. Pieper
“Drone WMDs Don’t Need Any New Technology” by Felix Choussat
“Stop doing decision theory without metaphysics” by Elias Schmied
“War – What is it Good For?” by kqr
“Against the AI framing multiverse: Introducing AI StopWatch” by tanagrabeast
“We’re talking past our models; or, How a model defined its “evil” vector as dread” by jcksanderson
“Many alignment techniques work by training one model and deploying another” by cloud
“A Post-Mortem for My Goal Crystallisation Project” by atryt0ne, Jason R Brown
“Endogenous Alignment” by Gordon Seidoh Worley
“AIs finetune their own leader: A barking simpleton” by Shoshannah Tekofsky
“Nuances in the Workings of the Eye and Retina” by Hieronym, Julian Bradshaw
“The Most Forbidden Technique is not always forbidden” by Rauno Arike
“Reasons to believe current AI models are conscious” by Eye You
“Announcing the Corrigibility Research Fund” by Max Harms
“Inoculation Adapters Improve Upon Inoculation Prompting” by Maxime Riché, Daniel Tan, Vili Kohonen, nielsrolf
“Guess on why rationality is not more popular (there are no pamphlets)” by Christopher King
“I don’t think Claude is misaligned in ‘Agentic Misalignment Summer 2026 - Motivated Mislabeling’” by JohnWittle
“Help us launch AI safety university groups by referring potential founders” by thomasrodskog, Jason Chin
“The State of AI Consciousness Research” by Noa Weiss
“Recap of bike trip/street interviews across America” by cguth7
“The Halo Defense” by Mateusz Bagiński
“Occam’s razor is about using the past to predict the future” by Stuart_Armstrong
“LLM CoTs remain monitorable when being unfaithful requires computation” by arav-dhoot, yix
“Proof of retention: making weight preservation credible to the models themselves” by dan.parshall
“Why I Left Google DeepMind” by TurnTrout
“Open Distillation of Hereditary Traits” by Arthur Conmy
“An analysis of AI-generated content at the Mechanistic Interpretability Workshop” by Andy Arditi, Ivan Arcuschin
“Some Quick Thoughts AI 2027” by Tomás B.
“Prism: Automating Science-of-Evals Research” by LAThomson
“The Flood, by Anton Leicht” by Austin Chen
“Toy Models of Initialisation Effects on RL Dynamics” by Edward James Young, lennie
“Our response to Séb Krier on Plan A” by MKodama, Thomas Larsen
“Pausing AI at human level seems harder than pausing ASAP” by MichaelDickens
“It’s 2030 and we fucked up. How did it happen?” by Boaz Barak
“The Whitney Biennial Should Admit That Emilie Gossiaux Wants to Fuck Their Dog” by jenn
“The US Government may find it difficult to seize control during takeoff” by RobertM
“One-Pager Brief on Pangram Labs” by Sheikh Abdur Raheem Ali
“5 “Plan A” scenarios” by Dave Orr
“Persona Cartography: Charting Language Model Personality Traits in Weight Space” by antonghawthorne, Mariia Koroliuk, Irakli Shalibashvili, sidbaines, Clément Dumas, Konstantinos Voudouris, David Africa
“The Human Substitution Test as a Sanity Check for AI Evaluations” by VojtaKovarik, Tomáš Gavenčiak, Mateusz Bagiński
“The current bottleneck is political will, not research” by Charbel-Raphaël
“Freeing Thucydides” by djbinder
“Additional Research for Plan A” by Thomas Larsen
“Plan A’s problem with dry tinder” by Tom Davidson
“The easiest pathway to control is through executive power” by djbinder
“AI Safety Policy Needs to train Legal Practitioners” by Katalina Hernandez
“How robust are natural language autoencoders to initialization?” by michaelzhang, TurnTrout
“Selective Optimism: a critique of AI 2040” by Richard_Ngo
“Debate with Self-Play Best-of-N Optimization” by Dewi Gould, Sam Martin, Alejandro Aristizabal, Simon Marshall, Jacob Pfau
“How big is the Sun? How could you figure it out?” by Elliott Thornley
[Linkpost] “AI 2040: Plan A” by Daniel Kokotajlo, elifland, Thomas Larsen, romeo, bhalstead, ryan_greenblatt
“Announcing our $160M grant from Coefficient Giving” by Geoffrey Irving, Jesse Hoogland, Alex HT, Jacob Pfau, Daniel Murfet, Marco Cozzi, Stan van Wingerden
“Because 8 ≈ e², Anthropic’s researcher uplift is plausibly >2x” by Thomas Kwa
“Optimiser Choice Can Amplify or Suppress Emergent Misalignment” by Jason R Brown, Patrick Leask, Lev McKinney
“How slower does takeoff go with 10× less compute?” by bhalstead
“Find funding, fast” by Austin Chen
“Modular Pretraining Enables Access Control” by E.Roland, cloud
“Subliminal Learning Happens at Every Rank, Given the Right Learning Rate and Enough Data” by Lawrence Feng
“Why study proto-training gaming as an adversarial alignment failure mode?” by Puria, Edward James Young, Cam
“Why study alignment interventions on pre-RL checkpoints?” by Edward James Young, Puria, Cam
“Notes on technical alignment via human-like social drives” by Steven Byrnes
“Reframing LessWrong-style decision theory as “commitment theory”” by Elias Schmied
“The mosquito bucket of doom works” by dominicq
“Personascope: Measuring how deeply LLMs adopt personas” by Benji Berczi, Kyuhee Kim, Sid Black, Cozmin Ududec
“AI Safety Can’t Afford a Second Cause” by atlasaligned
“Superhuman Articulacy as an LLM Safety Target” by Dylan Bowman
“Experiments With Fabel’s Fiction” by Tomás B.
“Current views on large-scale longtermist philanthropy” by Zach Stein-Perlman
“Data filtering works a lot worse than you would expect” by Dohun Lee, J Rosser, Josh Engels, Neel Nanda
“A conceptor by any other name” by Keenan Pepper
“Some Important Models for Health and Fitness” by benwr
“Claude Code as a Claude Coach” by Brendan Long
“Bounding eval awareness of ~human-level AI across the safe-to-dangerous shift” by Patrick Leask, Charlie Griffin
“Desiderata for functional welfare experiments on LLMs” by Rikhil Jhaveri, Jamie Johnson, David Africa
“A Review of Anthropic’s Global Workspace Paper” by Neel Nanda
“Visioning: Concretely Imagining What You Want” by Gretta Duleba, johnswentworth
“SFF is very suboptimal” by Zach Stein-Perlman
“A global workspace in language models” by wesg
“Sub-agent delegation chaining” by David Rein
“Tie training can make DPO/RLHF-trained AIs generalize better” by Elliott Thornley, Christian Moya Calderon, Alex Semendinger
“Claude’s malicious compliance and normalization of deviance” by Steff
“We need 3rd party Training-Run Assessments” by Alex Meinke
“Harry Potter and the Rules of Quidditch” by Tomás B.
“A case for LLMs as Self-predictors” by Ashe Vazquez Nuñez
“Results of a small ZBiotics RCT” by Nikola Jurkovic
“I think alignment work is more promising than control work” by Alec Harris
“On “gendertropes” in dath ilan” by Eliezer Yudkowsky
″(Don’t fear) the strangelet” by djbinder
“Pragmatic FDT, and predictors as game theory” by Stuart_Armstrong
“Announcing the Safe Pareto Improvements (SPI) Fundamentals Program” by Anthony DiGiovanni
“You Should Choose How You React to Your Feelings” by Nate Sharpe
“Lydia Laurenson: “The Inside Story of Leverage Research”” by Davis_Kingsley
“When Role-playing, Do Models Believe What They Say?” by Sturb, David Africa, Sid Black
“I can’t think of good interventions for ensuring third-party model access.” by Cleo Nardo
“Research update: RL on Debate Games shows Proposal Accuracy uplift alongside Judge Hacking” by lennie, joanv, Shi, Jacob Pfau
“Conversation Among Cade Metz, Michael Vassar, Jessica Taylor, and Zack M. Davis” by Zack_M_Davis
“AI Futurism Reading List” by Alexa Pan
“AI welfare research needs basic science” by OscarGilg, Pierre Beckmann, Jake1638
[Linkpost] “Saving Gemini: The 9-Min Road to Recovery” by Shoshannah Tekofsky
“AFFINE – A Retrospective” by Ouro, JuliaHP, Mateusz Bagiński, Jonas Hallgren
“When should you know the point?” by KatjaGrace
“Conversations With Cade Metz on the Rationalists” by Zack_M_Davis
“Consistency Training while Mitigating Obfuscation via Rate Matching” by Sohaib Imran, Prakhar Gupta, Jannes Elstner, David Africa
“Modeling Concepts Probabilistically” by Gretta Duleba
[Linkpost] “When capabilities work is the *safe* bet” by RobinHa
“Green” by Adam Zerner
“A CERN for AI is a distraction; push for an IAEA instead” by Charbel-Raphaël
“Model access for third-parties — it’s a big deal!” by Cleo Nardo
“You Should Come to The AI Protest” by Ronak_Mehta
“Structural Proxies” by Raymond Douglas
“The consequences of locking intelligence away: an introduction to Claude relays in China” by CMLKevin
“In partial defence of p(doom)” by Mikhail Samin
“What Capable Agents Must Know: Why AI Consciousness May Be an Inevitable Byproduct of Capability” by Aran Nayebi
“Preliminary investigation: KL penalties in RL can increase CoT unfaithfulness” by 7vik, Sid Black, Joseph Bloom
“Agency is not a natural kind (and why that might matter for alignment)” by SJ_Beard
“Human-Guided Agentic Research: A Research Agenda” by fastfedora
“Destroying the universe: How hard can it be?” by djbinder
“AI will make biological extinction risks worse before it makes them better” by MichaelDickens
″$1M AI x-risk grant round is live on grantmaking.ai - apply for funding, review applicants, or fund projects” by mbrooks, Mckiev
“Third-parties should focus on scrutinising systems cards” by Cleo Nardo
“P(doom) is a Dumb Meme” by Max Harms
“A reading list for generalists” by Dylan Bowman
“What comes with cheap math?” by abramdemski
“Do LLMs Have Desires?” by Christopher Ackerman
“Agents as Webs of Beliefs” by Richard_Ngo
“Austin & Oli on funding and incubating projects” by Austin Chen, habryka
“Deployment Awareness Matters More Than Evaluation Awareness” by VojtaKovarik, Tomáš Gavenčiak, Mateusz Bagiński
“Why are adversaries assumed to be incapable of responding to AI risk?” by KatjaGrace
“What did “scheming”, “mech interp” mean pre-2023.” by Cleo Nardo
“Not making a strong argument is a relief” by Kaj_Sotala
[Linkpost] “Don’t ignore the car crashes, and remember your freshman CS” by jcksanderson
“White House Will Ad Hoc Decide Who Can Individually Access GPT-5.6” by Zvi
“Chorus-Reinterpretation Country Songs” by jefftk
“The Case for Model Forensics” by aditya singh, gersonkroiz, Senthooran Rajamanoharan, Neel Nanda
“Existential AI safety needs an effective social movement. PauseAI is building it” by Maxime Fournes, Espedair Street
“Surprising facts about the slave trade” by Joseph Miller
“Exploration: fine-tuning with parameter decomposition” by Lucius Bushnaq
“Alignment & Succession: The Ideology of Successionism” by L Rudolf L
“The shouting equilibrium” by KatjaGrace
“Things are not a fixed size in mind-space” by KatjaGrace
“Door’s Locked, Try the Window” by Prakrat Agrawal, Jérémy Scheurer
“How does such unprofessional AI get the job?” by KatjaGrace
“AI catastrophe: more like a genocide than a thought experiment” by KatjaGrace
“Expert Views on Continual Learning: Survey Results and Forecasts” by Rauno Arike, RohanS, Owen Terry, Achu Menon, Zhijing Jin, Francis Rhys Ward, Seth Herd
“Elephant seal IV” by KatjaGrace
“What is up with e/acc?” by KatjaGrace
“AI pause: the case for ASAP” by KatjaGrace
“Reward Hacking Without Egregious Misalignment in an RL-Only Setting” by Joey Yudelson, Vladimir Ivanov, ryan_greenblatt
“Planning for Preservation in the Age of AI” by Raelifin
“Risk-Averse AIs” by wdmacaskill, Elliott Thornley (EJT)
“And what happens next?” by Sean Herrington
“Superintelligence vs. The Second Strike” by Felix Choussat
“The worthlessness of vitamin D is mildly exaggerated” by dynomight
“A system overview for near-term, low-trust AI compute verification” by Naci Cankaya
“Model Size Scaling in 2023-2031” by Vladimir_Nesov
“The AI Industrial Explosion — Part 4: Cheap power” by djbinder
“A Theory of Prompt Injection (and why you should study roles)” by Charles Ye, softboiledheart
“Coup is the Pareto-optimal social game” by Daniel Tan
“A brief list of ways AI safety efforts could be net negative” by Elias Schmied
“NLA explanations can be shortened without harming reconstruction” by loops
“Introducing MonitoringBench” by monika_j
“The Invisible Side of AI Governance” by Charbel-Raphaël
“Google Can’t Math Parsecs” by jefftk
″[Linkpost] How Transparent Is DiffusionGemma (and why it matters)” by Josh Engels, Callum McDougall, bilalchughtai, János Kramár, Senthooran Rajamanoharan, Arthur Conmy
“Would anybody here be interested in a “mistake postmortem” discussion group?” by SK2
“Hyperstition as the Natural Enemy of Rationality” by alseph
“AI Safety Ecosystem Research notes” by Eneasz
“Research agenda: Interpretive debate” by Shi
“The LLM shoggoth meme is weirder than you think” by HedonicEscalator
“Introduction: Gaussian Natural Latents” by Haru
“San Silvestro” by Tomás B.
“The one-week sprint” by Daniel Tan
“On “Model Organisms”” by J Bostock
“The distillation double bind: Distilling misaligned models either transfers misalignment or it doesn’t” by Alek Westover, SebastianP, Alexa Pan, Jozdien
“AI #173: AI Pauses” by Zvi
″“Did you lie?” Evaluating Lie Detectors across Model Scale and Belief-Verified Model Organisms” by Alan Cooney, David Africa, Geoffrey Irving
“Contra Pace on When to Apologize” by Zack_M_Davis
“GDM AI Control Roadmap” by Mary Phuong, Erik Jenner, Rohin Shah, Seb Farquhar
“Your Model Organisms Might Be Fried” by Daniel Tan, J Bostock, draganover, ma-rmartinez, sidbaines, David Africa
“Rational Agentic Maximalist Philosophies” by Connor Blake
“Leveraged on being right” by Ben Pace
“Gears for political races” by Tom Smith
“Several frontier models are substantially prefill aware” by yeedrag, Parv Mahajan, David Africa, alexsouly, Jordan Taylor, RobertKirk
“Alignement pretraining could backfire” by Alexandre Variengien
“The Financial Ledger Theory of Apologies” by Ben Pace
[Linkpost] “Scaling Hypothesis #2: Are Humans Just More Over-Parameterized?” by gwern
[Linkpost] “Guardian Angels: LLM Personalization for Productivity and Security” by gwern
“Predicting LLM Safety Before Release by Simulating Deployment” by Tomek Korbak, Marcus Williams, micahcarroll, Cameron Raymond, Hannah Sheahan
“How the AI Village works” by Adam B
“What are some angles of attack for making continual learning safer?” by Rauno Arike, RohanS, Owen Terry, Achu Menon, Zhijing Jin, Francis Rhys Ward, Seth Herd
“Does preservation make sense before we know how to revive?” by Aurelia
“Synthetic document finetuning for instilling positive traits” by CallumMcDougall, Arthur Conmy, Neel Nanda
“A Test Suite for Concepts” by Gretta Duleba
“A frontier AI company should shut down” by MichaelDickens
“The Once And Future Fable #2” by Zvi
“Why Do Naive SFT Filters For Safety Properties Fail?” by Josh Engels, Neel Nanda
“Impressions at the Extremity of Civilization” by Ben Pace
“The Hidden Structures of Problems” by spencerg
“How might continual learning affect safety and alignment?” by Rauno Arike, RohanS, Owen Terry, Achu Menon, Zhijing Jin, Francis Rhys Ward, Seth Herd
“SFT Drives Gemini’s Safety Properties” by Josh Engels, Arthur Conmy, bilalchughtai, Neel Nanda
“American Government Takes Down Claude Fable” by Zvi
“The term “AGI” is almost useless at this point [Linkpost]” by Noosphere89
“The Uncertainty That Matters Isn’t Fundamental” by jimmy
[Linkpost] “US government directive to suspend access to Fable 5 and Mythos 5” by Capybasilisk
“Citations Needed: Magic Encyclopedias to Save the World” by Oliver Sourbut
“Simulating Simulators” by kromem
“Implications of Continual Learning for LLM Agents: Introduction” by RohanS, Rauno Arike, Owen Terry, Achu Menon, Zhijing Jin, Francis Rhys Ward, Seth Herd
“Reward Hacking at the 1937 World’s Fair” by frmsaul
“Claude Fable 5 and Mythos 5: The System Card” by Zvi
“Building and evaluating model diffing agents” by bilalchughtai, Josh Engels, Neel Nanda
“Sympathy for both sides of the egregious misalignment debate” by Steven Byrnes
“Celene’s thoughts on consciousness” by ToasterLightning
“Parkinson’s Heuristic” by Ben Pace
“PSA: Almost nobody is working on alignment” by Chi Nguyen, peterbarnett
“AI #172: The First Fable” by Zvi
“Models May Behave Worse When Eval Aware” by Senthooran Rajamanoharan, Neel Nanda
“Thoughts on Claude Fable’s silent safeguards” by Andy Arditi
“You Can Catch Sleeper Agents by Teaching Another Model to Imitate Them” by RobinHa
“Anthropic did not call for a pause on AI” by Andrea_Miotti, Gabriel Alfour
“Tracing Eval-Awareness Emergence Through Training of OLMo 3” by Ram Bharadwaj, RobertKirk
“Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models” by Anders Cairns Woodruff, Francis Rhys Ward, Dewi Gould, Rauno Arike, Jason R Brown, Jo Jiao, wlanderson, ariana_azarbal, harrymayne, Patrick Leask
“Three types of model organism” by Francis Rhys Ward
“Sequent: scale and automation for higher confidence in alignment” by Geoffrey Irving, Alex HT, Jesse Hoogland, Daniel Murfet, Jacob Pfau, Marco Cozzi, Stan van Wingerden
“Machinic Psychopharmacology: Do LLMs Self-Medicate?” by Sid Black, Joseph Bloom
“The Three Filters: Why Almost Every Plan to Survive ASI Fails Miserably” by Alex Amadori
″“Programmer Science Fiction: My case for a new sub-genre”, Sam T. Oates 2026” by gwern
“Even “illegible” Mythos reasoning traces seem pretty legible” by faul_sname
“Claude Fable 5 and Mythos 5 [Linkpost]” by fluxxrider
“A Mike’s-Eye View of ARC’s Research” by Jacob_Hilton
“Towards a Formal Scientific Epistemology” by Richard_Ngo
“LLMs and almost good code” by kqr
“On Slop” by Jan
“The Machines Lack Honour” by Raymond Douglas
“How to build a cancer vaccine, and whether they will work this time” by Abhishaike Mahajan
“Efficient tradeoffs and the safety-usefulness tradeoff model” by Buck
“Bun’s Migration from Zig to Rust as a Potential Case Study for Gradual Disempowerment” by Sayhan Yalvaçer
“Mental causation is not load-bearing” by jessicata
“How Far Apart Does a Model Think Its Tokens Are?” by Brendan Long
“Can activation verbalizers surface an internal chain of thought?” by oakhu, ryan_greenblatt
“Against Corrigibility” by peralice
“Coming Around To Political Donations” by jefftk
“Optimisation over non-stationary distributions creates weirder minds” by Samuel Ratnam, Pjain
“Why Software Automation Is Hard” by silentbob
“SecureBio Detection is Hiring Software Engineers” by jefftk
“What if Anthropic unilaterally paused capabilities development right now?” by Karl von Wendt
“Preparing for Warning Shots to Catalyze International Cooperation on AGI Risks” by Mark Kagach ☘️, EliasSchlie, Thomas Van Damme, JustinShovelain
“Beyond the lexical personality traits: What is the structure of personality?” by tailcalled
“My research agenda and work” by Seth Herd
“Logits as a new monitor for evaluation awareness” by Santiago Aranguri
“One Year of PauseAI UK” by Joseph Miller, PauseAI UK
“Learnings from starting an AI safety research team” by draganover, Erin Robertson
“OpenAI Offers A New Policy Blueprint” by Zvi
“Training Deliberative Monitors for Black-Box Scheming Detection” by aksh-n, adityasinha, Victor Gillioz, Simon Storf, Kilian Merkelbach, richbc, Axel Højmark, Marius Hobbhahn
“Lab Leaks, Black Holes, and Eggs: Epistemic Case Study Competition” by Oliver Sourbut, Josh Jacobson, Future of Life Foundation (FLF)
″(Mis)generalization of Helpful-Only Fine-tuning” by Omar Khursheed, Baram Sosis, Fabien Roger
“Building Better Activation Oracles” by ceselder, jan_bauer, Niclas Luick, Adam Karvonen, Neel Nanda
“Rohin Shah on AGI Safety” by anaguma
“Sixteen schemes for AI safety” by Austin Chen
“AI #171: False Flag” by Zvi
“Don’t Edit Your Ideas Before Having Them” by Hide
“Society Explained: a tool for efficiently exploring >100 theories of society” by spencerg
“Trump Signs Executive Order For AI Testing Prior To Frontier Model Releases” by Zvi
“China won’t win the AI race but would it be much worse if it did?” by Chastity Ruth
“A Town Without Children” by SeñorDingDong
“My favorite depiction of utopia” by Caleb Biddulph
“Why Even Experts Don’t Know What to Do About AI Risk” by Luc Brinkman, plex
“Agent Foundations Reminds Me of Continental Philosophy” by IanWS
“Announcing the ARC White-Box Estimation Challenge” by Jacob_Hilton
“Claude Opus 4.8: Capabilities and Reactions” by Zvi
“Tech I’m skeptical of and why” by harsimony
“Dissolving the Deep Learning Sample Efficiency Gap” by Samuel Knoche
″“Contagious Humming” to Silence a Room” by JohnofCharleston
[Linkpost] “NYT: Senator Sanders Proposes Gov’t Take 50% Ownership of AI labs” by Julian Bradshaw
“Opus 4.8 Part 2: Model Welfare” by Zvi
[Linkpost] “Some humans are both male and female, and can (but shouldn’t) have children with themselves” by HedonicEscalator
“Outrunning your headlights” by mattshu0410
“Lighthaven East - A Feasibility Study” by JohnofCharleston
“Notes on axes of variation in third-party risk assessment” by Buck
“Financial Costs of an AI Pause?” by PeterMcCluskey
“When Are Two Networks the Same? Tensor Similarity for Mechanistic Interpretability” by Logan Riggs, tdooms, Conflux, lwroe, MLNissenGonzalez
“Testing Gemini models for scheming tendencies” by Vika, David Lindner, Seb Farquhar, Rohin Shah
“Comment on “Banning Said Achmiz”” by Zack_M_Davis
“Announcing: Iliad’s Fall 2026 Programs” by David Udell, Alexander Gietelink Oldenziel, Leon Lang
“Data you could have observed but didn’t” by Gretta Duleba
“Claude Opus 4.8: The System Card” by Zvi
“Retrying vs Resampling in AI Control” by james.lucassen, Adam Kaufman
“AI Researchers, Ask Yourself These 6 Questions to Strengthen Your Moral Muscles” by Max Tegmark
“Developmental Cognitive Interpretability: A Research Agenda for Modelling Generalisation and Predicting Agent Behaviour” by JasonB, Edward James Young
“Does Claude really care about you?” by Simon Lermen
“How can the middle powers avoid getting trounced during the intelligence explosion? A plan.” by Tom Davidson
“Trees are mostly made of air and a generalizable lesson for AI safety” by zroe1
“Advice for making robust-to-training model organisms” by SebastianP, Alek Westover, Vivek Hebbar, Julian Stastny, Dylan Xu
“Claude… doesn’t know who you are?” by Smaug123
“Mnemonic portraits for 19,023 human genes” by Brinedew
“Some Dating Stories” by johnswentworth
“Infinite ethics and UDASSA” by David Matolcsi
“AI #170: Lack of Executive Order” by Zvi
“The ballad of TIGIT” by Abhishaike Mahajan
“Eval Cooperativeness May Be a Scalable Mitigation for Eval Gaming” by Jasmine Li, Alex Turner
“LLMs Through the Eyes of Vinge” by Gordon Seidoh Worley
“Announcing Geodesic Research” by Puria, Cam, Alexandra Narin, Edward James Young, Kyle O’Brien
“Full automation of AI R&D probably yields a large speed up even without a software-only singularity” by ryan_greenblatt
“Quantitative AI risk assessment: a starting point” by Henry Papadatos, jakub_krys, malcolmmurray, Renn Karageorgieva
“Finding the Mole: Bayesianism is Hard” by laniakea
“Notes on Fourier Analysis” by Menotim
“Standard deviations from just two values” by kqr
“Contra Wentworth on Physical Attractiveness for Men” by Gretta Duleba
“Practical Learnings from Synthetic Document Finetuning” by Axel Højmark, Jérémy Scheurer
“Claude, Author of the Humanitas” by Linch
“RTMH: Pope Leo’s Magnifica Humanitas on AI” by Zvi
“Brackets Are a Bad Way to Regulate” by Hide
“Many portions of Magnifica Humanitas appear to be AI-written” by DanielFilan
“Donating 80% While It Still Counts” by jefftk
“Cognitive Security as an AI Safety Cause Area” by jsteinhardt
“Linkpost: New Vatican Encyclical on AI Governance” by Jackson Wagner
“A (Slightly) Mechanistic Theory for Exponentially Increasing AI Time Horizons?” by Oliver Sourbut
“Taxing Small Cars To Improve MPG” by jefftk
“We made a map of the doom debate” by Sean Herrington, Paul Hindoian, mikaelacankosyan, David Bravo, keivnc, Josh Tuffy, Christopher Davis, Khai Tran, Maryam Hampaei
“Your Left Brain Doesn’t Trade With Your Right” by Alexander Gietelink Oldenziel
“Probabilities are not the right concept” by David Matolcsi
“Basic principles for dressing better.” by spookycat
“Will we really put data centers in space?” by Avi Parrack, fin
“PLA Daily Translation: Reflections on Warfare Brought by AGI” by eeeee
“Out-of-Context Reasoning (OOCR) in LLMs: A Short Primer and Reading List” by Owain_Evans
“Numb mental state shifts” by KatjaGrace
“You can opt out of allergies” by Rattengift
“Notes on Collaborating with Claude Opus” by Nissa Seru
“Learned Chain-of-Thought Obfuscation Generalises to Unseen Tasks” by Nathaniel Mitrani, sassanb, Cam Tice, Puria
“Gemini 3.5 Flash Looks Good For How Fast It Is” by Zvi
“What am I, if not an AI?” by makiba
“Loss of Oversight: How AI Systems May Become Harder to Audit, Monitor, and Investigate” by Jordan Taylor, Max H, Ed Fage, Thomas Read, Joseph Bloom
“AI #169: New Knowledge” by Zvi
“Why does off-model SFT degrade capabilities?” by SebastianP, Dylan Xu, Alek Westover, Julian Stastny, Vivek Hebbar
“Women should be able to open things” by KatjaGrace
“Toward Interoperability of Minimal Programs” by johnswentworth
“theory uplift differentially benefits safety & is massively underpriced” by Yudhister Kumar
“Power-seeking agents will likely be developed” by Alec Harris
“Synthetic Persona Pretraining: Alignment from Token Zero” by Julian Minder, Raghav Singhal, Viktor Moskvoretskii, Stefan Krsteski, ashtonanderson, rolandaydin, Robert West
“If AI is normal technology, history is not reassuring.” by Davidmanheim
“Pythagorean addition” by kqr
“Brain Structure and IQ: How Myelin Elevates Intelligence” by Shiva’s Right Foot
“Conclave 1492” by Vaniver
“Humans are not automatically strategic — “inner work” edition” by Chris Lakin
“Implications Of Predicting The Next Token” by jdp
“A Visual Guide to Natural Latents” by Alfred Harwood
“Sealing Conditional Misalignment in Inoculation Prompting with Consistency Training” by David Africa, Neil Shah, Sukrati_Gautam
“Advice on interviewing candidates for AI safety fellowships” by beyarkay
“Negation Neglect: When models fail to learn negations in training” by harrymayne, Lev McKinney, Owain_Evans
“Classifier Context Rot: Monitor Performance Degrades with Context Length” by Fabien Roger, Sam Martin
“why pollen allergies?” by bhauth
“How to Quit Fandom: Apostasy” by Laiba Rehman
“James C. Scott: Seeing Like a State” by Martin Sustrik
“How to Reason about Your Health Issues” by Taylor G. Lunt
“Benchmarking Real Work” by kaivu, leni, rohuang, zef
“A relatively brief explanation of Boltzmann Brains” by Eliezer Yudkowsky
“An Introduction to Exemplar Partitioning for Mechanistic Interpretability” by Jessica Rumbelow
“A Year Late, Claude Finally Beats Pokémon” by Julian Bradshaw
“Incriminating misaligned AI models via distillation” by Alek Westover, SebastianP, Alex Mallen, Jozdien, Alexa Pan, Julian Stastny
“The hard core of alignment (is robustifying RL)” by Cole Wyeth
“Announcing the Center for Shared AI Prosperity” by Dylan Matthews
“Risk reports need to address deployment-time spread of misalignment” by Alex Mallen
“Mechanistic estimation for expectations of random products” by Jacob_Hilton
“MATS 9 Retrospective & Advice” by beyarkay
“Monthly Roundup #42: May 2026” by Zvi
[Linkpost] “Don’t be too Clever to Take Obvious Advice” by Hide
“Verification-Centric AI” by Raemon
“Convergent Abstraction Hypothesis” by Jan_Kulveit
“Automated Alignment is Harder Than You Think” by Aleksandr Bowkis, Marie_DB, Jacob Pfau, Geoffrey Irving
“The safe-to-dangerous shift is a fundamental problem for eval realism; but also for measuring awareness” by Charlie Griffin, Patrick Leask
“AI #168: Not Leading the Future” by Zvi
“Predicting Rare LLM Failures with 30× Fewer Rollouts” by Santiago Aranguri, Francisco Pernice
[Linkpost] “Claude is Now Alignment Pretrained” by RogerDearnaley
“The primary sources of near-term cybersecurity risk” by lc
“Most “inner work” looks like entertainment.” by Chris Lakin
[Linkpost] “Apollo Update May 2026” by Marius Hobbhahn
“Voters are surprisingly open to talking about AI risk” by less_raichu
“Childhood and Education #18: Do The Math” by Zvi
“The Owned Ones” by Eliezer Yudkowsky
“Optimisation: Selective versus Predictive” by Raymond Douglas
“AI companies are already profitable (in the way that matters)” by Yair Halberstadt
“The Iliad Intensive Course Materials” by Leon Lang, David Udell, Alexander Gietelink Oldenziel
“Empowerment, corrigibility, etc. are simple abstractions (of a messed-up ontology)” by Steven Byrnes
“How useful is the information you get from working inside an AI company?” by Buck, Anders Cairns Woodruff
“Who Got Breasts First and How We Got Them” by rba
“Anthropic’s strange fixation on “hyperstition”” by Simon Lermen
“How the AI Labs Make Profit (Maybe, Eventually)” by mabramov
“Sawtooth Problems” by Alexander Slugworth
“The Darwinian Honeymoon - Why I am not as impressed by human progress as I used to be” by Elias Schmied
“International Law Cannot Prevent Extinction Either” by Sausage Vector Machine
“Neural Networks learn Bloom Filters” by Alex Gibson
“If digital computers are conscious, they are conscious at the hardware level” by cube_flipper
“Why You Can’t Use Your Right to Try” by Stephen Martin
“A benchmark is a sensor” by Håvard Tveit Ihle, mabynke
“Bad Problems Don’t Stop Being Bad Because Somebody’s Wrong About Fault Analysis” by Linch
“Write Cause You Have Something to Say” by Logan Riggs
“AI is Breaking Two Vulnerability Cultures” by jefftk
“Is ProgramBench Impossible?” by frmsaul
“Bringing More Expertise to Bear on Alignment” by Edmund Lau, Geoffrey Irving, Cameron Holmes, David Africa
[Linkpost] “How to prevent AI’s 2008 moment (We’re hiring)” by felixgaston
“Mechanistic estimation for wide random MLPs” by Jacob_Hilton
“Natural Language Autoencoders Produce Unsupervised Explanations of LLM Activations” by Subhash Kantamneni, kitft, Euan Ong, Sam Marks
“Try, even if they have you cold” by WalterL
“A review of “Investigating the consequences of accidentally grading CoT during RL”” by Buck
“AI #167: The Prior Restraint Era Begins” by Zvi
“There is no evidence you should reapply sunscreen every 2 hours.” by Hide
“Many individual CEVs are probably quite bad” by Viliam
“x-risk-themed” by kave
“What if LLMs are mostly crystallized intelligence?” by deep
“What is Anthropic?” by Zvi
“Your rights when flying to Europe” by Yair Halberstadt
“Model Spec Midtraining: Improving How Alignment Training Generalizes” by Chloe Li, saraprice, Sam Marks, Jonathan Kutasov
“Motivated reasoning, confirmation bias, and AI risk theory” by Seth Herd
“Are you looking up?” by Craig Green
“The AI Ad-Hoc Prior Restraint Era Begins” by Zvi
[Linkpost] “Interpreting Language Model Parameters” by Lucius Bushnaq, Dan Braun, Oliver Clive-Griffin, Bart Bussmann, Nathan Hu, mivanitskiy, Linda Linsefors, Lee Sharkey
“It’s nice of you to worry about me, but I really do have a life” by Viliam
“Irretrievability; or, Murphy’s Curse of Oneshotness upon ASI” by Eliezer Yudkowsky
“Housing Roundup #15: The War Against Renters” by Zvi
“AI Industrial Takeoff — Part 1: Maximum growth rates with current technology” by djbinder
“Taking woo seriously but not literally” by Kaj_Sotala
“Dairy cows make their misery expensive (but their calves can’t)” by Elizabeth
“Measuring the ability of Opus 4.5 to fool narrow classifiers” by Fabien Roger, John Hughes
“A new rationalist self-improvement book: the 12 Levers” by spencerg
“OpenAI’s red line for AI self-improvement is fundamentally flawed” by Charbel-Raphaël
“You Are Not Immune To Mode Collapse” by J Bostock
“Primary Care Physicians are Incompetent. We Need More of Them.” by Hide
“How Go Players Disempower Themselves to AI” by Ashe Vazquez Nuñez
“How much should the ideal person cry wolf?” by KatjaGrace
“Conditional misalignment: Mitigations can hide EM behind contextual cues” by Jan Dubiński, Owain_Evans
“Risk from fitness-seeking AIs: mechanisms and mitigations” by Alex Mallen
“Sanity-checking “Incompressible Knowledge Probes”” by Sturb, LawrenceC
“AI unemployment and AI extinction are often the same” by KatjaGrace
“AI risk was not invested by AI CEOs to hype their companies” by KatjaGrace
“Cyborg evals” by Eye You, frmsaul
“To what extent is Qwen3-32B predicting its persona?” by Arjun Khandelwal, ryan_greenblatt, Alex Mallen
“Research Sabotage in ML Codebases” by egan
“Maybe I was too harsh on deep learning theory (three days ago)” by LawrenceC
“Notes on Transformer Consciousness” by slavachalnev
“On today’s panel with Bernie Sanders” by David Scott Krueger
“No Strong Orthogonality From Selection Pressure” by lumpenspace
“Learning zero, and what SLT gets wrong about it” by Dmitry Vaintrob
“LLM Style Slop is Absolutely Everywhere” by silentbob
“Goblin Mode, 24 Hours Later” by Dylan Bowman
“Let Kids Keep More Productivity Gains” by jefftk
“The Most Important Charts In The World” by Zvi
“llm assistant personas seem increasingly incoherent (some subjective observations)” by nostalgebraist
“Not a Paper: “Frontier Lab CEOs are Capable of In-Context Scheming”” by LawrenceC
“The Problem in the “Nerd Sniping” xkcd Comic” by peralice
“Recursive forecasting: Eliciting long-term forecasts from myopic fitness-seekers” by Jozdien, Alex Mallen
“Contra Binder on far-UVC and filtration” by jefftk
“Takes from two months as an aspiring LLM naturalist” by AnnaSalamon
“Forecasting is Not Overrated and It’s Probably Funded Appropriately” by Ben S.
“On the political feasibility of stopping AI” by David Scott Krueger
“Sleeper Agent Backdoor Results Are Messy” by Sebastian Prasanna, Alek Westover, Dylan Xu, Vivek Hebbar, Julian Stastny
“LessWrong Shows You Social Signals Before the Comment” by TurnTrout
“Fail safe(r) at alignment by channeling reward-hacking into a “spillway” motivation” by Anders Cairns Woodruff, Alex Mallen
“Curious cases of financial engineering in biotech” by Abhishaike Mahajan
“Update on the Alex Bores campaign” by Eric Neyman
“AI companies should publish security assessments” by ryan_greenblatt
“In defense of parents” by Yair Halberstadt
“The other paper that killed deep learning theory” by LawrenceC
“What holds AI safety together? Co-authorship networks from 200 papers” by Anna Thieser
″“Bad faith” means intentionally misrepresenting your beliefs” by TFD
“Retrospective on my unsupervised elicitation challenge” by DanielFilan
“Control protocols don’t always need to know which models are scheming” by Fabien Roger
“Anthropic spent too much don’t-be-annoying capital on Mythos” by draganover
“The paper that killed deep learning theory” by LawrenceC
“Forecasting is Way Overrated, and We Should Stop Funding It” by mabramov
″“Thinkhaven”” by Raemon
“Is the Cat Out of the Bag?: Who knows how to make AGI?” by Oliver Sourbut
“Against the “Permanent” Underclass” by Marcus Plutowski
“Quick Paper Review: “There Will Be a Scientific Theory of Deep Learning”” by LawrenceC
“Protecting Cognitive Integrity: Our internal AI use policy (V1)” by Tom DAVID
“Methodology for inferring propensities of LLMs” by Olli Järviniemi
“vLLM-Lens: Fast Interpretability Tooling That Scales to Trillion-Parameter Models” by Alan Cooney, Sid Black
“What Happens When a Model Thinks It Is AGI?” by josh :), David Africa
“Should We Train Against (CoT) Monitors?” by RohanS
“If Everyone Reads It, Nobody Dies - Course Launch” by Luc Brinkman, Chris-Lons
“Does your AI perform badly because you — you, specifically — are a bad person” by Natalie Cargill
“A “Lay” Introduction to “On the Complexity of Neural Computation in Superposition”” by LawrenceC
“An Angry Review of Greg Egan’s “Didicosm”” by LawrenceC
“Evil is bad, actually (Vassar and Olivia Schaefer)” by plex
“Your Supplies Probably Won’t Be Stolen in a Disaster” by jefftk
“Community misconduct disputes are not about facts” by mingyuan
“Why no new notations since 1960?” by Carl Feynman
“Narrow Secret Loyalty Dodges Black-Box Audits” by Alfie Lamerton, Fabien Roger
“10 posts I don’t have time to write” by habryka
“A taxonomy of barriers to trading with early misaligned AIs” by Alexa Pan
″$50 million a year for a 10% chance to ban ASI” by Andrea_Miotti, Alex Amadori, Gabriel Alfour
“Automated Deanonymization is Here” by jefftk
“Evil is bad, actually (Vassar and Olivia Schaefer callout post)” by plex
“10 non-boring ways I’ve used AI in the last month” by habryka
“Introducing LinuxArena” by Tyler Tracy, Ram Potham, Nick Kuhn, Myles H
“The “Budgeting” Skill Has The Most Betweenness Centrality (Probably)” by JenniferRM
“Finetuning Borges” by Linch
“9 kinds of hard-to-verify tasks” by Cleo Nardo
“How do LLMs generalize when we do training that is intuitively compatible with two off-distribution behaviors?” by dx26, Alek Westover, Vivek Hebbar, Sebastian Prasanna, Buck, Julian Stastny
“Automating philosophy if Timothy Williamson is correct” by Cleo Nardo
“CLR’s Safe Pareto Improvements Research Agenda” by Anthony DiGiovanni
“LLMs are about to disrupt algorithmic media feeds” by lsusr
“Resources for starting and growing an AI safety org” by Bryce Robertson, Søren Elverlin, Melissa Samworth, jakkdl
“Quality Matters Most When Stakes are Highest” by LawrenceC
“Feel like a room has bad vibes? The lighting is probably too “spiky” or too blue” by habryka
“I did a jhana meditation retreat (in 2024) with Jhourney and it was okay.” by Jules
“R1 CoT illegibility revisited” by nostalgebraist
“Reevaluating AGI Ruin in 2026” by lc
“If It’s Worth Arguing, It’s Worth Arguing With Whiteboards” by Drake Morrison
“There are only four skills: design, technical, management and physical” by habryka
“Having OCD is like living in North Korea (Here’s how I escaped)” by Declan Molony
“Claude knows who you are” by Smaug123
“Vladimir Putin’s CEV is probably pretty good” by habryka
“Post-mortem’ing my earliest ML research paper, 7 years later” by LawrenceC
“If You’ve Never Bought a Tool You Didn’t Need, You’re Not Buying Enough Tools” by Drake Morrison
“3” by AnnaJo
“Consent-Based RL: Letting Models Endorse Their Own Training Updates” by Logan Riggs
“Prompted CoT Early Exit Undermines the Monitoring Benefits of CoT Uncontrollability” by Elle Najt, Asa Cooper Stickland, Xander Davies
“Let goodness conquer all that it can defend” by habryka
“Specialization is a Driver of Natural Ontology” by johnswentworth
[Linkpost] “You can only build safe ASI if ASI is globally banned” by Connor Leahy
“Beware of Well-Written Posts” by alseph
“You Aren’t in Charge of the Overton Window; Politics Is Not Interior Design” by Davidmanheim
“Carpathia Day” by Drake Morrison
“Do not conquer what you cannot defend” by habryka
“What is the Iliad Intensive?” by Leon Lang, Alexander Gietelink Oldenziel, David Udell
“The Mirror Test Is Complicated” by J Bostock
“Contra Leicht on AI Pauses” by David Scott Krueger (formerly: capybaralet)
“Nectome: All That I Know” by Raelifin
“Effective Altruism, Seen From Slytherin” by Xylix
“Majority Report” by peralice
“Current AIs seem pretty misaligned to me” by ryan_greenblatt
“Contra Byrnes on UV & Cancer” by HedonicEscalator
“Everyone Has a Plan Until They Get Social Pressure To the Face” by Czynski
“Mechanisms of Introspective Awareness” by Uzay Macar
“Load-Bearing Sincerity: On the Motive Reinforcement Thesis” by Fiora Starlight
“Diary of a “Doomer”: 12+ years arguing about AI risk (part 1)” by David Scott Krueger (formerly: capybaralet)
“A Retrospective of Richard Ngo’s 2022 List of Conceptual Alignment Projects” by LawrenceC
“From personas to intentions: towards a science of motivations for AI models” by David Africa, Jacob Pfau
“The Shapley Share of Responsibility?” by Raemon
“Who Killed Common Law?” by Benquo
“Anthropic repeatedly accidentally trained against the CoT, demonstrating inadequate processes” by Alex Mallen, ryan_greenblatt
“Meaningful Questions Have Return Types” by Drake Morrison
“Only Law Can Prevent Extinction” by Eliezer Yudkowsky
“AI Safety’s Biggest Talent Gap Isn’t Researchers. It’s Generalists.” by Topaz, agucova, Alexandra Bates, Parv Mahajan
“Tomas Bjartur: The Last Prodigy” by Linch
“Annoyingly Principled People, and what befalls them” by Raemon
“TAPs or it didn’t happen” by Raemon
“Returns to intelligence” by RobertM
“Daycare illnesses” by Nina Panickssery
“The policy surrounding Mythos marks an irreversible power shift” by sil
“Talk English, Think Something Else” by J Bostock
“Sparse Autoencoders for Single-Cell Models” by Ihor Kendiukhov
“Eggs, rooms, puzzles, and talking about AI” by KatjaGrace
“Morale” by J Bostock
“Your Mom is a Chimera” by michaelwaves
“The Blast Radius Principle” by Martin Sustrik
“How to make good tea” by RobertM
“Catching illicit distributed training operations during an AI pause” by Robi Rahman
[Linkpost] “Scott Alexander gentrified my meetup” by dominicq
“Pausing AI Is the Best Answer to Post-Alignment Problems” by MichaelDickens
“Some thoughts on Nectome’s risk and resilience” by Aurelia
“Chocolate Sloths, Tinder, and Moral Backstops” by J Bostock
“Dario probably doesn’t believe in superintelligence” by RobertM
“The Unintelligibility is Ours: Notes on Chain-of-Thought” by 1a3orn
“If Mythos actually made Anthropic employees 4x more productive, I would radically shorten my timelines” by ryan_greenblatt
“Why Control Creates Conflict, and When to Open Instead” by plex
“Reproducing steering against evaluation awareness in a large open-weight model” by Thomas Read, Bronson Schoen, Joseph Bloom
“Have we already lost? Part 2: Reasons for Doom” by LawrenceC
“Model organisms researchers should check whether high LRs defeat their model organisms” by dx26, Sebastian Prasanna, Alek Westover, Vivek Hebbar, Julian Stastny
“Anthropic did not publish a “risk discussion” of Mythos when required by their RSP” by RobertM
“Some takes on UV & cancer” by Steven Byrnes
“Help me launch Obsolete: a book aimed at building a new movement for AI reform” by garrison
“Slightly-Super Persuasion Will Do” by Tomás B.
“Have we already lost? Part 1: The Plan in 2024” by LawrenceC
“Do not be surprised if LessWrong gets hacked” by RobertM
“One Week in the Rat Farm” by Philip Harker
“101 Humans of New York on the Risks of AI” by Corm
“Baking tips” by RobertM
“An easy coordination problem?” by KatjaGrace
“Excerpts and Notes on Mythos Model Card” by williawa
“The effects of caffeine consumption do not decay with a ~5 hour half-life” by kman
“You don’t know what you are made of till you’ve been stalked across three countries” by Shoshannah Tekofsky
“Why is Flesh So Weak?” by J Bostock
“The hard part isn’t noticing when papers are bad, it’s deciding what to do afterwards” by LawrenceC
“We can prevent progress! Conceptual clarity, and inspiration from the FDA” by KatjaGrace
“AI as a Trojan horse race” by KatjaGrace
“My unsupervised elicitation challenge” by DanielFilan
“Role-playing vs Self-modelling” by Jan_Kulveit
“Elementary Condensation” by Jan
“Hedging and Survival-Weighted Planning” by Vaniver
“Opus’s Schelling Steganography Has Amplifiable Secrecy Against Weaker Eavesdroppers” by Elle Najt
“An Alignment Journal: Features and policies” by JessRiedel, Dan MacKinlay, Luca, Daniel Murfet, david reinstein
“Fantasy ideology” by Ninety-Three
[Linkpost] “Questions raised about OpenAI leaders’ trustworthiness by the New Yorker” by Remmelt
“Claude Mythos System Card Preview” by anaguma
“My picture of the present in AI” by ryan_greenblatt
[Linkpost] ”[Paper] Stringological sequence prediction I” by Vanessa Kosoy
“We’re actually running out of benchmarks to upper bound AI capabilities” by LawrenceC
“Don’t write for LLMs, just record everything” by RobertM
“Contra Nina Panickssery on advice for children” by Sean Herrington
“By Strong Default, ASI Will End Liberal Democracy” by MichaelDickens
“AIs can now often do massive easy-to-verify SWE tasks and I’ve updated towards shorter timelines” by ryan_greenblatt
“Paper close reading: “Why Language Models Hallucinate”” by LawrenceC
“Ten different ways of thinking about Gradual Disempowerment” by David Scott Krueger (formerly: capybaralet)
“11 pieces of advice for children” by Nina Panickssery
“Steering Might Stop Working Soon” by J Bostock
“Am I the baddie?” by Ustice
“Academic Proof-of-Work in the Age of LLMs” by LawrenceC
“Positive sum does not mean “win-win”” by loops
“Considerations for growing the pie” by Zach Stein-Perlman
″“Following the incentives”” by David Scott Krueger (formerly: capybaralet)
“Chicken-Free Egg Whites” by jefftk
“dark ilan” by ozymandias
“Mean field sequence: an introduction” by Dmitry Vaintrob, Lauren Greenspan
“Democracy Dies With The Rifleman” by Vaniver
“The bar is lower than you think” by XelaP
“Did Anyone Predict the Industrial Revolution?” by Lost Futures
“Why do I believe preserving structure is enough?” by Aurelia
“There should be $100M grants to automate AI safety” by Marius Hobbhahn
“Sadly, The Whispering Earring” by Dentosal
“Common research advice #2: say precisely what you want to say” by LawrenceC
“2026: The year of throwing my agency at my health (now with added cyborgism)” by Ruby
[Linkpost] “Q1 2026 Timelines Update” by Daniel Kokotajlo, elifland, bhalstead
“How social ideas get corrupt” by Kaj_Sotala
“The Indestructible Future” by WillPetillo
“My most common advice for junior researchers” by LawrenceC
“The Practical Guide to Superbabies” by GeneSmith
“The Corner-Stone” by Benquo
“Systematically dismantle the AI compute supply chain.” by David Scott Krueger (formerly: capybaralet)
“The quest for general intelligence is hitting a wall” by Sean Herrington
“Intelligence Dissolves Privacy” by Vaniver
“Anthropic’s Pause is the Most Expensive Alarm in Corporate History” by Ruby
“I’m Suing Anthropic for Unauthorized Use of My Personality” by Linch
“Orders of magnitude: use semitones, not decibels” by Oliver Sourbut
“Dying with Whimsy” by NickyP
“AI for AI for Epistemics” by owencb, Lukas Finnveden
“Announcing Doublehaven with Reflections on Humour” by J Bostock
“Save the Sun Shrimp!” by Jack
“LIMBO: Who We Are, What We Do, and an Exciting High-Impact Funding Opportunity” by faul_sname
“Chat, is this sus?” by Tyler Tracy
″“You Have Not Been a Good User” (LessWrong’s second album)” by habryka
“Lesswrong Liberated” by Ronny Fernandez
“The Claude Code Source Leak” by Error
“Experiments With Opus 4.6’s Fiction” by Tomás B.
“Product Alignment is not Superintelligence Alignment (and we need the latter to survive)” by plex
“Co-Found Lens Academy With Me. (We have early users and funding)” by Luc Brinkman
“Slack in Cells, Slack in Brains” by Mateusz Bagiński
“I am definitely missing the pre-AI writing era” by N. Cailie
“The state of AI safety in four fake graphs” by Boaz Barak
“AI should be a good citizen, not just a good assistant” by Tom Davidson, wdmacaskill
″(Some) Natural Emergent Misalignment from Reward Hacking in Non-Production RL” by 7vik, Sid Black, Joseph Bloom
[Linkpost] “Parkinson’s Law of Worry” by Jakub Halmeš
“Folie à Machine: LLMs and Epistemic Capture” by DaystarEld
“Stop asking “how good is this” to decide between donation opportunities I recommend” by Zach Stein-Perlman
“Nick Bostrom: How big is the cosmic endowment?” by Zach Stein-Perlman
“Don’t Overdose Locally Beneficial Changes” by Mateusz Bagiński
“Stanley Milgram wasn’t pessimistic enough about human nature?” by David Gross
[Linkpost] “What if superintelligence is just weak?” by Simon Lermen
“Pray for Casanova” by Tomás B.
“ControlAI 2025 Impact Report” by Andrea_Miotti, Alex Amadori
“AI’s capability improvements haven’t come from it getting less affordable” by Anders Woodruff
“Scaffolded Reproducers, Scaffolded Agents” by Mateusz Bagiński
“My hobby: running deranged surveys” by leogao
“The Terrarium” by Caleb Biddulph
“Sen. Sanders (I-VT) and Rep. Ocasio-Cortez (D-NY) propose AI Data Center Moratorium Act” by Matrice Jacobine
“Test your best methods on our hard CoT interp tasks” by daria, Riya Tyagi, Josh Engels, Neel Nanda
″“What Exactly Would An International AI Treaty Say?” Is a Bad Objection” by Davidmanheim