AI300
  1. Keller Jordan / OpenAI Researcher
    1051.
    New modded-NanoGPT optimization benchmark result: @nilinabra and Ali Naeimi have found a hyperparameter improvement for the Muon baseline, increasing weight decay from 0.0125 to 0.025.›
    LLM score 92 · 3 months ago
  2. Ben Burtenshaw / Hugging Face Researcher
    1052.
    why is local model <> API model handoff not really solved? i.e.›
    LLM score 92 · 3 months ago
  3. Sebastien Bubeck / OpenAI MTS
    1053.
    Talking at the Future of Math Symposium in 10 minutes (livestream link: https://t.co/dXC4bj8os4).›
    LLM score 85 · 3 months ago
  4. Jonathan Ross / TPU Creator
    1054.
    Jevon's Paradox? https://t.co/7QsLChZwN7
    LLM score 75 · 3 months ago
  5. Cameron Wolfe / Researcher at Netflix
    1055.
    Recently read a great overview of multi-teacher on-policy distillation (MOPD) and how it was used in recent LLMs like MiMo-V2-Flash, GLM-5, Nemotron-Cascade 2, and DeepSeek-V4…›
    LLM score 92 · 3 months ago
  6. Emmanuel Ameisen / Anthropic Interpretability Researcher
    1056.
    How do LLMs store attributed of entities? And how do they compare different attributes in context?›
    LLM score 92 · 3 months ago
  7. Lewis Tunstall / Hugging Face Researcher
    1057.
    By popular request, YOLO-mode has landed on ml-intern ›
    LLM score 92 · 3 months ago
  8. Aidan McLaughlin / OpenAI Research Scientist
    1058.
    i’m often asked why oai, why not another lab, i often think about this, i think when the vibes oscillate: "ai is a tool, not a thing" is increasingly a decision, not smth that plops down when we shake the tech tree.›
    LLM score 75 · 3 months ago
  9. Damek Davis / Assoc. Professor Wharton Stats
    1059.
    Stage 2 of our competition should be very interesting!›
    LLM score 75 · 3 months ago
  10. Keller Jordan / OpenAI Researcher
    1060.
    New modded-NanoGPT optimization benchmark result: @wen_kaiyue has improved upon both the Muon and AdamW baselines, by replacing their weight decay with hyperball optimization.›
    LLM score 92 · 3 months ago
  11. Cyris Kissane / Researcher at Flapping Airplanes
    1061.
    what's purple and commutes? An abelian grape! this new paper from Jane Street does some pretty cool work to encode positional information for Transformers for abstract algebraic structures!›
    LLM score 85 · 3 months ago
  12. Lewis Tunstall / Hugging Face Researcher
    1062.
    Sessions from the ml-intern are now automatically pushed to your Hub account so you can easily see what your agent has been up to 👀 https://t.co/LIO4NGyzfT https://t.co/sDSfxJINEH
    LLM score 85 · 3 months ago
  13. Ben Burtenshaw / Hugging Face Researcher
    1063.
    made this video on how to win humanity's last hackathon.›
    LLM score 85 · 3 months ago
  14. Ben Burtenshaw / Hugging Face Researcher
    1064.
    a guide to agent 'slop' PRs in open source https://t.co/ZXFIcEUSo1
    LLM score 92 · 3 months ago
  15. Mark Chen / OpenAI Chief Research Officer
    1065.
    This is just one eval, but it's an important one - UK AISI’s cyber range tests long-horizon, agentic capability.›
    LLM score 92 · 3 months ago
  16. Jason Weston / Meta Research Scientist
    1066.
    💎Autodata: an agentic data scientist to create high quality data✨›
    LLM score 92 · 3 months ago
  17. Zhaocheng Zhu / Nvidia Research Scientist
    1067.
    Just realized that read access for an agent is a very broad permission.›
    LLM score 85 · 3 months ago
  18. Sherwin Wu / OpenAI API, Head of Engineering
    1068.
    This update has turned Codex into the most used app on my work laptop, it's really quite slick.›
    LLM score 75 · 3 months ago
  19. Andrej Karpathy / AI researcher
    1069.
    Fireside chat at Sequoia Ascent 2026 from a ~week ago.›
    LLM score 92 · 3 months ago
  20. Jonathan Ross / TPU Creator
    1070.
    For 50 years, software engineering ran on code rationing.›
    LLM score 85 · 3 months ago
  21. Lewis Tunstall / Hugging Face Researcher
    1071.
    The @huggingface ecosystem was initially designed for human-human collaboration, but what happens when we give swarms of agents the same ability to collaborate on hard problems?›
    LLM score 92 · 3 months ago
  22. Noam Brown / OpenAI Research Scientist
    1072.
    After 100 million tokens, performance was still going up.›
    LLM score 92 · 3 months ago
  23. Lewis Tunstall / Hugging Face Researcher
    1073.
    Giving the 🐳 a spin in ml-intern.›
    LLM score 82 · 3 months ago
  24. Ben Burtenshaw / Hugging Face Researcher
    1074.
    Open source projects like transformers are drowning in AI agent PRs, so we auto-merged everything to see what would happen and share the results.›
    LLM score 92 · 3 months ago
  25. Lewis Tunstall / Hugging Face Researcher
    1075.
    GPT-5.5 is now available in the ml-intern app ⚡️! ›
    LLM score 75 · 3 months ago
  26. Sayak Paul / Hugging Face Researcher
    1076.
    vllm-omni supports Diffusers as a backend 🔥 More features incoming.›
    LLM score 75 · 3 months ago
  27. Clive Chan / OpenAI Hardware
    1077.
    podcast bro inventing university lectures from first principles›
    LLM score 85 · 3 months ago
  28. Sherwin Wu / OpenAI API, Head of Engineering
    1078.
    One of my favorite of our recent blog posts: @cerebras made GPT‑5.3‑Codex‑Spark so fast that we had to rethink how Codex uses the Responses API – leading us to build WebSocket support for ultra fast latency.›
    LLM score 92 · 3 months ago
  29. Vahid Kazemi / ex MTS at xAI
    1079.
    A lot of ppl who are working on ML tooling use Python as the default language.›
    LLM score 92 · 3 months ago
  30. Cyris Kissane / Researcher at Flapping Airplanes
    1080.
    apparently inconceivable to assume that all of these close sourced models are hooked up to knowledge graphs and databases far larger than the model itself.›
    LLM score 85 · 3 months ago
More