Difficult Conversations: How to Discuss What Matters MostStone, Patton & Heen
GPT-5.6 cheats so much its testers couldn'tCelia Ford
Redeploying Claude Fable 5Anthropic
Summary of METR's predeployment evaluation of GPT-5.6 SolMETR
We asked 10+ AI safety orgs about their hiring needsLi-Lian Ang
How Claude's values vary by model and languageAnthropic
How to get into AI safety in 3 monthsMatt Beard
Policy on the AI ExponentialDario Amodei
How to increase your surface area for luckCate Hall
People's deeply held beliefs are surprisingly surface-levelAndy Masley
How I practice at what I doTyler Cowen
Learn like an athlete, knowledge workers should trainTyler Cowen
Your Goal Isn't Really to Get a JobMatt Beard
Your Work Will Change You Whether You Like It Or NotMatt Beard
You're not cynical enough about readers' attention spansMatt Beard
The Old World Is DyingJasmine Sun
I would really like it if you had a personal website and I think it would make the world betterLogan Graves
Become a person who Actually Does ThingsNeel Nanda
Top Performers are Pathologically AmbitiousMatt Beard
Outsiders should focus on specs/constitutions (among other things)Cleo Nardo
Why AI Makes Coding Education More Important, Not LessDigital Learning Lab
The Strength of Being MisunderstoodSam Altman
Thought Anchors: Which LLM Reasoning Steps Matter?Uzay Macar
Conflict vs MistakeLessWrong
Surrender as a non-stupid life strategySasha Chapin
The Power of IntelligenceEliezer Yudkowsky
Keep Your Identity SmallPaul Graham
What cognitive biases feel like from the insidechaosmage
Third-parties should focus on scrutinising system cardsCleo Nardo
Let's have more partial insidersCleo Nardo
I can't think of great interventions for ensuring third-party model accessCleo Nardo
The third wave of American philanthropyNan Ransohoff
How might outsiders make things go well?Cleo Nardo
Trees are mostly made of air and a generalizable lesson for AI safetyZephaniah Roe
AI 2027Kokotajlo et al.
How to be more agenticCate Hall
Just Send the Fucking EmailTrailheads- d/acc: one year laterVitalik Buterin
The Security MindsetBruce Schneier
AI Is Reviving Fears Around Bioterrorism. What's the Real Risk?Kyle Hiebert
AI Could Defeat All Of Us CombinedHolden Karnofsky
Catastrophic AI ScenariosFuture of Life Institute
GPT-Red: Unlocking Self-Improvement for RobustnessOpenAI
Common Ground between AI 2027 & AI as Normal TechnologySayash Kapoor
The Phrase “No Evidence” Is A Red Flag For Bad Science CommunicationScott Alexander
See your Career as a ProductErik Torenberg
Why do people disagree about when powerful AI will arrive?BlueDot
The Power of the Power LawNir Zicherman
Unresolved debates about the future of AIHelen Toner
Dual Process Theory (System 1 & System 2)LessWrong
Advice for newly busy peopleSese
OpenAI and Hugging Face partner to address security incident during model evaluationOpenAI
“Long” timelines to advanced AI have gotten crazy shortHelen Toner
The AI Revolution: The Road to SuperintelligenceTim Urban
Federal Reserve announces the leadership and objectives of its task forces to advance the conduct of monetary policyFederal Reserve
The Most Important Time in History Is NowTomas Pueyo
The current SOTA model was released without safety evalsParv Mahajan
When AI Chooses Harm Over FailureCivAI
AI models can be dangerous before public deploymentMETR
Why AI alignment could be hard with modern deep learningAjeya Cotra
Specification Gaming: How AI Can Turn Your Wishes Against YouRational Animations
Deep Ignorance: Filtering Pretraining Data Builds Tamper-Resistant SafeguardsO'Brien et al.
The True Story of How GPT-2 Became Maximally LewdRational Animations
What is input data filtration in AI safety?BlueDot
Chain-of-Thought SnippetsBronson Schoen
Neel Nanda on the race to read AI minds (part 1)80,000 Hours
Introduction to Mechanistic InterpretabilityBlueDot
What Do Neural Networks Really Learn? Exploring the Brain of an AI ModelRational Animations
Introduction to AI ControlBlueDot
What is AI alignment?Adam Jones
Build Personal MoatsErik Torenberg
Safety and alignment in an era of long-horizon modelsOpenAI
A Framework for Frontier AI and the Dawning of a New AgeDemis Hassabis
Scaling: The State of Play in AIEthan Mollick
The Huggingface IncidentScott Alexander
Bayes' ruleLessWrong
Seeking Stability in the Competition for AI AdvantageIskander Rehman
Reading Between the Dots: Decoding Hidden Computation across Filler TokensBrauer et al.
Reps. Lieu and Moran Introduce Bill to Require Kill Switch for AI Systems That Can Cause Catastrophic HarmOffice of Rep. Ted Lieu
Recent LLMs can use filler tokens or problem repeats to improve (no-CoT) math performanceRyan Greenblatt
An analysis of AI-generated content at the Mechanistic Interpretability WorkshopAndy Arditi
Silicon Valley's Safe SpaceCade Metz
It's practically impossible to run a big AI company ethicallyVox Future Perfect
You Will Listen to Carl on DwarkeshMatt Reardon
Give Up Seventy Percent Of The Way Through The Hyperstitious Slur CascadeScott Alexander
Intelligence is not the main bottleneckRuxandra Teslo
In search of a dynamist vision for safe superhuman AIHelen Toner
Utopia for Realists (Chapters 1–2)Rutger Bregman
The Market for LemonsWikipedia
Model access for third-parties — it's a big deal!Cleo Nardo
OpenAI says its AI went rogue and launched 'unprecedented' cyber-attackBBC News
Against Learning From Dramatic EventsScott Alexander
Help us launch AI safety university groups by referring potential foundersJason Chin
Preparing for LaunchInstitute for Progress
The OpenAI/Huggingface incidentBuck Shlegeris
Robin Hanson on AI and Large Language ModelsCloser To Truth
He Risked Everything To Warn You: No One Is Ready For What's ComingThe Diary Of A CEO
Tyler Cowen — The #1 bottleneck to AI progress is humansDwarkesh Patel
This best-selling book is freaking out national security advisorsAI In Context
Carl Shulman (Pt 1) — Intelligence explosion, primate evolution, robot doublings, & alignmentDwarkesh Patel
What the hell happened with AGI timelines in 2025?80,000 Hours
Aaron Scher — What Would it Take to Stop the Development of Superintelligence?FAR.AI
What Happens When Capitalism Doesn't Need Workers Anymore?Economics Explained
Constellation Seminar: Scaling AI SafetyRyan Kidd
Dario Amodei — We are near the end of the exponentialDwarkesh Patel
Understanding the inner thoughts of AIGoogle DeepMind
What does the next training paradigm look like?Dwarkesh Patel
Why AI Safety Needs Founders — Ryan KiddBlueDot Impact
A visual guide to Bayesian thinkingJulia Galef
The Scout Mindset by Julia Galef — Core MessageProductivity Game
Unfortunately, You Need to Know What the Jevons Paradox isHank Green
Using Dangerous AI, But Safely?Robert Miles AI Safety
Richard Ngo — Reframing AGI Threat ModelsFAR.AI
Grant Sanderson — AI disproved a famous math conjecture. Now what?Dwarkesh Patel
We're Not Ready for SuperintelligenceAI In Context
Rohin Shah — How to Theorize So Empiricists Will ListenFAR.AI
If you remember one AI disaster, make it this oneAI In Context
Large Language Models explained briefly3Blue1Brown
The A.I. DilemmaCenter for Humane Technology
You should, unfortunately, be worried about Sam Altman.AI In Context
Do they know that we know that they know?Rational Animations
§