Ishaan Panigrahi

on the atlas — 99

• Your Model Organisms Might Be Fried — LessWrong1 savers
• Humans Still Beat AI in the Long Horizon: Revisiting Test-Time Scaling in the Agent Era | Qiuyang Mang1 savers
• SFT, RL, and On-Policy Distillation Through a Distributional Lens | wh4 savers
• 2026 RL Directions - Impression1 savers
• RL Scaling Laws for LLMs - by Cameron R. Wolfe, Ph.D.1 savers
• Stanford CS25: Transformers United V6 I From Representation Learning to World Modeling - YouTube1 savers
• Josh Engels on X: "New GDM interp research: SFT is a big deal for safety relevant behaviors. We recently investigated root causes for some of Gemini’s behaviors. We were surprised to find that many behaviors actually came from the initial supervised finetuning stage, not later stages like RL! 🧵 https://t.co/mLg87XuXK5" / X1 savers
• Second Batch | First Proof Project3 savers
• Notes | Andy Arditi1 savers
• [2606.12360] Anatomy of Post-Training: Using Interpretability to Characterize Data and Shape the Learning Signal1 savers
• Goodfire on X: "Have you debugged your training data? You might not like what you find. Introducing predictive data debugging: reveal and shape what your model will learn before training. In DPO datasets, we found broken guardrails, hallucinations, and fish fart fan fiction (seriously). (1/9) https://t.co/V9MrEQvBvq" / X1 savers
• Extropic | Home1 savers
• RL Interview Questions 2026 | Notion1 savers
• Mor Geva on X: "What's in a neuron? 💫 (an atypically long, almost personal post) Neurons in LMs have always been a fascinating object to study. I've been studying them since 2020, viewing them as key-value memory cells, analyzing what they capture in vocabulary space, and how they compose https://t.co/179ELFnEw3" / X1 savers
• How LLMs Actually Work | 0xkato1 savers
• CausaLab — Can LLM Agents Discover Causal Mechanisms by Experiment?1 savers
• Xiuyu Li on X: "RL Interview Questions 2026" / X1 savers
• Stanford CS25: Transformers United V6 I Serving Transformers: Lessons from the Trenches - YouTube1 savers
• Stanford CS25: Transformers United V6 I Overview of Transformers - YouTube1 savers
• Why Software Automation Is Hard — LessWrong1 savers
• The better AI gets, the smaller its share of the economy might get – Alex Imas and Phil Trammell - YouTube1 savers
• I lead AGI safety at Google DeepMind – here's the view from the inside | Rohin Shah - YouTube1 savers
• Arjun Virk2 savers
• Vivek on X: "some notes on getting into frontier ai labs" / X1 savers
• will brown on X: "On SFT, RL, and on-policy distillation" / X1 savers
• ARENA - AI Safety Curriculum4 savers
• The Iliad Intensive Course Materials — LessWrong2 savers
• A Theory of Deep Learning | Elements of a Vector Space6 savers
• What I learned this week - Pretraining parallelisms, Can distillation be stopped, Mythos and the cybersecurity equilibrium, Pipeline RL, On why pretraining runs fails3 savers
• A short note on interpretability and minds1 savers
• LLM Architecture Gallery | Sebastian Raschka, PhD5 savers
• [2510.24941] Can Aha Moments Be Fake? Identifying True and Decorative Thinking Steps in Chain-of-Thought1 savers
• Do Thing, Do One Thing45 savers
• The Illustrated Transformer – Jay Alammar – Visualizing machine learning one concept at a time.34 savers
• When AI builds itself \ Anthropic30 savers
• reflections on palantir - Nabeel S. Qureshi30 savers
• Toy Models of Superposition28 savers
• The Persona Selection Model: Why AI Assistants might Behave like Humans28 savers
• https://cs.stanford.edu/~jsteinhardt/ResearchasaStochasticDecisionProcess.html26 savers
• The Intelligence Curse24 savers
• How to win a best paper award (or, an opinionated take on how to do important research)24 savers
• Why We Think | Lil'Log22 savers
• A global workspace in language models \ Anthropic20 savers
• James Somers19 savers
• How to be More Agentic - by Cate Hall - Useful Fictions19 savers
• Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Lab18 savers
• How to Land a Frontier Lab Job17 savers
• Alignment is not solved but it increasingly looks solvable17 savers
• Welcome • freemediaheckyeah15 savers
• Modular Manifolds - Thinking Machines Lab15 savers
• Transformer Circuits Thread14 savers
• Alignment remains a hard, unsolved problem — LessWrong14 savers
• Notes on the Industry Job Search11 savers
• The stable marriage problem - by Ajeya Cotra - Good Bones10 savers
• getting older, college, and dreams10 savers
• Current AIs seem pretty misaligned to me — LessWrong9 savers
• Approximating KL Divergence9 savers
• Where the goblins came from9 savers
• A vision researcher’s guide to some RL stuff: PPO & GRPO - Yuge (Jimmy) Shi8 savers
• Highly Opinionated Advice on How to Write ML Papers — AI Alignment Forum8 savers
• MATS 9 Retrospective & Advice — LessWrong8 savers
• Ideas — Noah Zender8 savers
• ML Job Interviews: The Ultimate Guide – Silvia Sapora7 savers
• Noam Brown on X: "Implications of Large-Scale Test-Time Compute" / X7 savers
• frontier model training methodologies | Alex Wa’s Blog7 savers
• An Apple-Picking Model of AI R&D | Tom Cunningham – Tom Cunningham7 savers
• What Would Non-Linear Features Actually Look Like? | Liv Gorton7 savers
• A shallow dive into formal verification6 savers
• State of RL for reasoning LLMs | A. Weers6 savers
• Speculative Decoding - philkrav5 savers
• The Unintelligibility is Ours: Notes on Chain of Thought5 savers
• jacobhilton/deep_learning_curriculum: Language model alignment-focused deep learning curriculum5 savers
• The ultimate guide to RL environments: building and scaling them in the LLM era - a Hugging Face Space by AdithyaSK5 savers
• Effective learning: Twenty rules of formulating knowledge - SuperMemo5 savers
• Evidence on AI R&D Progress from NanoGPT - METR5 savers
• Maybe I was too harsh on deep learning theory (three days ago) — LessWrong5 savers
• [2604.16812] Introspection Adapters: Training LLMs to Report Their Learned Behaviors4 savers
• Is Frontier Asynchronous RL Solved? — Luke J. Huang4 savers
• Can activation verbalizers surface an internal chain of thought? — LessWrong4 savers
• Don’t Outsource Your Thinking4 savers
• Interpreting Language Model Parameters4 savers
• Transformers Explained Visually (Part 3): Multi-head Attention, deep dive | by Ketan Doshi | Towards Data Science3 savers
• Faithful, Interpretable Model Explanations via Causal Abstraction | SAIL Blog3 savers
• Quantization from the ground up | ngrok blog3 savers
• The Annotated JEPA | Elements of a Vector Space3 savers
• A socratic dialogue over the utility of DNA language models (Part 1 of 2)3 savers
• Insights on Crosscoder Model Diffing3 savers
• Your Transformer is Secretly an EOT Solver | Elements of a Vector Space3 savers
• The Human Skill That Eludes AI - The Atlantic3 savers
• Auto-review of agent actions without synchronous human oversight3 savers
• Everything About Transformers2 savers
• Are AI benchmarks doomed? - by Anson Ho and Greg Burnham2 savers
• RI Seminar: Jitendra Malik : Robot Learning, With Inspiration From Child Development - YouTube2 savers
• Openbird | Local-first macOS activity journal2 savers
• Getting Caught Up to Modern LLM Research | Samarth Goel2 savers
• [2505.10831] Creating General User Models from Computer Use2 savers
• How hard is it to inoculate against misalignment generalization? — LessWrong2 savers
• Chip design from the bottom up – Reiner Pope - YouTube2 savers
• Stanford CS336 | Language Modeling from Scratch2 savers