AI300
  1. Andrew White / Edison Scientific Cofounder
    3211.
    I remember seeing some analysis, I think sponsored by @huggingface , that benchmarked openrouter models on MMLU as a function of provider to show the wild swings in quality.›
    LLM score 75 · 9 months ago
  2. alphaXiv
    3212.
    "OAT: Ordered Action Tokenization" This paper proposes OAT, which is a way to turn continuous robot actions into discrete tokens that stay valid at every prefix.›
    LLM score 92 · 9 months ago
  3. Omar Khattab / MIT CSAIL Asst professor
    3213.
    Very nice new page by @isaacbmiller1 on eight open-source ports of DSPy to other programming languages›
    LLM score 85 · 9 months ago
  4. Jeffrey Wang / OpenAI Researcher
    3214.
    cheapest availabilities at avalon mission bay (<10 min bike ride to openai & anthropic HQs): ›
    LLM score 72 · 9 months ago
  5. Sebastien Bubeck / OpenAI MTS
    3215.
    AIxMath is going through some extremely confused chatter at the moment.›
    LLM score 92 · 9 months ago
  6. Jim Fan / NVIDIA Director of Robotics
    3216.
    me fomo? Sure. We may be seeing the end of the Middlegame.›
    LLM score 82 · 9 months ago
  7. Jerry Tworek / ex OpenAI VP of RL
    3217.
    Code and data distinction was always kind of artificial https://t.co/ujNO21TLB7
    LLM score 85 · 9 months ago
  8. Omar Khattab / MIT CSAIL Asst professor
    3218.
    Nope. My lab is making 3 algorithmic bets.›
    LLM score 92 · 9 months ago
  9. Omar Khattab / MIT CSAIL Asst professor
    3219.
    Aren't RLMs bad for the prefix cache? Totally! But keep in mind:›
    LLM score 92 · 9 months ago
  10. Vahid Kazemi / ex MTS at xAI
    3220.
    The best way I've found to use LLMs for coding is through a layered design with strict one-way dependency.›
    LLM score 92 · 9 months ago
  11. Alex Zhang / MIT CSAIL PhD
    3221.
    Maybe I can provide some intuition, but lmk if it’s unclear — I am trying to refine how I explain this anyways!›
    LLM score 92 · 9 months ago
  12. Clive Chan / OpenAI Hardware
    3222.
    In a future world with some kind of superintelligence, it’s interesting to think about having an arbiter that is - measurably - less biased than a human.›
    LLM score 82 · 9 months ago
  13. Christian Szegedy / ex xAI Cofounder
    3223.
    A year ago everybody just laughed at AI being ridiculously useless for mathematical research.›
    LLM score 92 · 9 months ago
  14. Alex Zhang / MIT CSAIL PhD
    3224.
    while procrastinating on research I decided it's finally time to add RLMs to pypi!›
    LLM score 82 · 9 months ago
  15. Igor Babuschkin / Cofounder of xAI
    3225.
    People joke about buying Mac minis for OpenClaw and it seems absurd at first if you know how to use cloud VMs.›
    LLM score 75 · 9 months ago
  16. Shayan Salehian / ex MTS at xAI
    3226.
    Career update: I left xAI to start something new, closing my 7+ year chapter working at Twitter, X, and xAI with so much gratitude.›
    LLM score 25 · 9 months ago
  17. Noam Brown / OpenAI Research Scientist
    3227.
    When GPT-5 was released, some folks claimed AI progress was hitting a wall, whereas others said progress would continue.›
    LLM score 92 · 9 months ago
  18. alphaXiv
    3228.
    "First Proof" A team of researchers proposes a way to test if AI can actually do NEW math by releasing 10 freshly-solved and never public research questions, with answers temporarily encrypted.›
    LLM score 92 · 9 months ago
  19. Boris Cherny / Creator of Claude Code
    3229.
    We just launched an experimental new fast mode for Opus 4.6.›
    LLM score 85 · 9 months ago
  20. Jerry Tworek / ex OpenAI VP of RL
    3230.
    Run fewer experiments and think about them more
    LLM score 85 · 9 months ago
  21. Cameron Wolfe / Researcher at Netflix
    3231.
    Still can't believe Olmo 3 found out you can do in-flight weight updates during RL without even clearing the KV cache and it still works.›
    LLM score 92 · 9 months ago
  22. alphaXiv
    3232.
    "Maximum Likelihood Reinforcement Learning" Turns out we've been using RL wrongly the whole time, and missed out on 20x efficiency?›
    LLM score 92 · 9 months ago
  23. Yifei Zhou / Thinking Machines Researcher
    3233.
    Always a big fan of works coming out of Andrea’s lab! Love the mathematical insight from the comparison of rl and maximum likelihood loss https://t.co/ZhaYCGyNwF
    LLM score 12 · 9 months ago
  24. Omar Khattab / MIT CSAIL Asst professor
    3234.
    wait that’s such a smart product move from OpenRouter›
    LLM score 68 · 9 months ago
  25. Igor Babuschkin / Cofounder of xAI
    3235.
    I’ve tested the latest generation of all the major AIs on theoretical physics research and Claude 4.6 has absolutely blown me away with how capable it is in physics.›
    LLM score 92 · 9 months ago
  26. Omar Khattab / MIT CSAIL Asst professor
    3236.
    Since there's now 5+ papers that propose on-policy context distillation, I feel comfortable confessing that we (@NoahZiems), too, were working on that haha.*›
    LLM score 92 · 9 months ago
  27. Cyris Kissane / Researcher at Flapping Airplanes
    3237.
    Cool new paper: Drifting Models. https://t.co/ALvy7dmbWC›
    LLM score 75 · 9 months ago
  28. Cyris Kissane / Researcher at Flapping Airplanes
    3238.
    https://t.co/zcejLXVsTf cool new secure python interpreter for AI use.›
    LLM score 75 · 9 months ago
  29. Ben Burtenshaw / Hugging Face Researcher
    3239.
    the @huggingface supports agentic environments that you can use for rl or evals.›
    LLM score 85 · 9 months ago
  30. John Carmack
    3240.
    256 Tb/s data rates over 200 km distance have been demonstrated on single mode fiber optic, which works out to 32 GB of data in flight, “stored” in the fiber, with 32 TB/s bandwidth.›
    LLM score 92 · 9 months ago
More