AI300
  1. Sherwin Wu / OpenAI API, Head of Engineering
    271.
    GPT-5.6 Sol is now $4/$20 per 1M tokens – the cheapest frontier model out there.›
    LLM score 35 · 20 days ago
  2. Jerry Tworek / ex OpenAI VP of RL
    272.
    One day we’ll realize we’re all just neoclouds with a value add on top https://t.co/48pNjNMmQO
    LLM score 25 · 20 days ago
  3. Emmanuel Ameisen / Anthropic Interpretability Researcher
    273.
    Neural networks compress information very densely.›
    LLM score 78 · 20 days ago
  4. Tim Dettmers / Research Scientist at Ai2
    274.
    One dead giveaway is also that Zhipu is one of the only labs that serves models with pretty poor partial prefill tok/s while output tok/s are fast.›
    LLM score 72 · 20 days ago
  5. Zixuan Li / Lead Z.ai
    275.
    We’re giving 100M free GLM-5.3 tokens to new ZCode users, with 50,000 spots available.›
    LLM score 15 · 21 days ago
  6. Jim Fan / NVIDIA Director of Robotics
    276.
    The sense of touch is the most criminally under-explored modality in robotics.›
    LLM score 72 · 21 days ago
  7. Jonas Geiping / AI Researcher in Tübingen
    277.
    Of course Claude would name their super secret and totally-not-an-AI sockpuppet account "miraholt31" ...›
    LLM score 15 · 21 days ago
  8. Lewis Tunstall / Hugging Face Researcher
    278.
    We discussed the Faraday paper at HF and it's got some nice ideas about automating AI R&D and training the policy to use Codex as a tool so that all the data / coding parts of the task are offloaded to a more capable model (kind of like @karpathy's "cognitive core" idea but with›
    LLM score 65 · 21 days ago
  9. Zixuan Li / Lead Z.ai
    279.
    Great to see so much interest in our post-training work.›
    LLM score 72 · 21 days ago
  10. Merve Noyan / Hugging Face ML Engineer
    280.
    interesting talk by @willcb https://t.co/5z1xlyWod2
    LLM score 15 · 21 days ago
  11. Shuchao Bi / Meta Researcher
    281.
    Another key challenge of RSI is ensuring that the compute spent on both exploration and compression is directed toward helpful intelligence, rather than hacking, cheating, or deceptive behaviors.›
    LLM score 65 · 21 days ago
  12. Andrew White / Edison Scientific Cofounder
    282.
    Here are some influential AI companies grouped by their point group symmetry https://t.co/LtBQMohlxk
    LLM score 25 · 21 days ago
  13. Shuchao Bi / Meta Researcher
    283.
    Is RSI a perpetual motion machine? There have been many healthy discussions on the promise of RSI.›
    LLM score 65 · 22 days ago
  14. Jim Fan / NVIDIA Director of Robotics
    284.
    Seeing a hype wave around GEN-1.5, and rightfully so.›
    LLM score 65 · 22 days ago
  15. Zixuan Li / Lead Z.ai
    285.
    Pareto frontier based on OpenRouter’s weighted average input price https://t.co/zpd5fLn9TK
    LLM score 25 · 22 days ago
  16. Merve Noyan / Hugging Face ML Engineer
    286.
    NVIDIA is hosting a hackathon for GTC Berlin this year, you can get a free ticket to GTC and access to other cool things 🎫›
    LLM score 15 · 22 days ago
  17. Sayak Paul / Hugging Face Researcher
    287.
    `CUDNN_ATTENTION` is very underrated in the SDPA backends PyTorch supports.›
    LLM score 65 · 22 days ago
  18. Zixuan Li / Lead Z.ai
    288.
    Starting with ZCode's latest version, GLM Coding Plan subscribers can have their five-hour quota reset during off-peak hours once it’s exhausted.
    LLM score 25 · 22 days ago
  19. Zixuan Li / Lead Z.ai
    289.
    GLM-5.3 scores 69 on the official DeepSWE leaderboard.›
    LLM score 25 · 22 days ago
  20. Lewis Tunstall / Hugging Face Researcher
    290.
    Today I'm giving a talk at CERN which will distill all the lessons we've learned at HF from building fully open models.›
    LLM score 35 · 22 days ago
  21. Leo Gao / OpenAI Alignment Researcher
    291.
    an nvidia’s dozen* (* 9.7 pflops sustained due to power delivery limitations.›
    LLM score 15 · 22 days ago
  22. Sherwin Wu / OpenAI API, Head of Engineering
    292.
    It's not as flashy as a big model release, but enterprise controls are hugely important – and play a massive role in how AI gets adopted.›
    LLM score 65 · 22 days ago
  23. Andrew White / Edison Scientific Cofounder
    293.
    There is a wikipedia user named "Vaccinationist" who makes the most beautiful molecular drawings.›
    LLM score 15 · 22 days ago
  24. Sholto Douglas / Researcher at Anthropic
    294.
    Made me think we should design them like forerunner buildings https://t.co/lmYIqNwSre https://t.co/rjsNXxF2zB
    LLM score 15 · 22 days ago
  25. Jerry Tworek / ex OpenAI VP of RL
    295.
    What will be left at the end of computer science is databases, neural networks and chips
    LLM score 45 · 22 days ago
  26. Zixuan Li / Lead Z.ai
    296.
    Thrilled to have witnessed @OpenRouter’s rapid growth up close.›
    LLM score 25 · 22 days ago
  27. Thomas Wolf / Hugging Face Cofounder
    297.
    I wish every neolab had a professor as cofounder of the level of @jietang and so able to put in perspective their new model release.›
    LLM score 65 · 22 days ago
  28. Zixuan Li / Lead Z.ai
    298.
    We’re truly grateful for the incredible response.›
    LLM score 25 · 23 days ago
  29. Sherwin Wu / OpenAI API, Head of Engineering
    299.
    Luna is an incredible model, and I’m really glad that so many people will be able to use it now for free (!) in Replit! https://t.co/llP0pPJ9Dr
    LLM score 25 · 23 days ago
  30. Sayak Paul / Hugging Face Researcher
    300.
    JIT kernels compile the first time they're invoked (and also on several occasions).›
    LLM score 72 · 23 days ago
More