AI300
  1. Benjamin Spector / Cofounder of Flapping Airplanes
    91.
    All reliable clusters are alike; each unreliable cluster is unreliable in its own way.
    LLM score 15 · 6 days ago
  2. Vahid Kazemi / ex MTS at xAI
    92.
    Was experimenting with the new Gemini Flash today and went on a rabbit hole of exploring autostereograms.›
    LLM score 35 · 6 days ago
  3. Keller Jordan / OpenAI Researcher
    93.
    Modded-NanoGPT optimization result #18: @zhanpeng_zhou has achieved a step count of 3225 with a preconditioned Muon variant called PMuon.›
    LLM score 62 · 6 days ago
  4. Eric Jang / ex VP of AI at 1X Robotics
    94.
    .@tinkerapi by TML is a highly underrated product.›
    LLM score 65 · 6 days ago
  5. Sander Dieleman / DeepMind Research Scientist
    95.
    💯AI research often isn't so much research, as it is "throwing stuff at the wall and seeing what sticks".›
    LLM score 65 · 6 days ago
  6. Cameron Wolfe / Researcher at Netflix
    96.
    Do you need to learn how to properly evaluate your agent? Here’s a step-by-step guide for how to do this, informed by best practices in recent research…›
    LLM score 82 · 6 days ago
  7. John Carmack
    97.
    I have been very impressed by @SemiAnalysis_ .›
    LLM score 65 · 6 days ago
  8. Ronak Malde / DeepMind Researcher
    98.
    Gabe and team is killing it like always! It's going to be more important than ever to push the frontier on long-horizon tasks in frontier domains, and this is a huge step in that direction https://t.co/sg8yBc0eMK
    LLM score 25 · 7 days ago
  9. Keller Jordan / OpenAI Researcher
    99.
    Modded-NanoGPT optimization result #15: A Newton-Muon based setup has achieved a step count which is below Muon and slightly above NorMuon.›
    LLM score 65 · 7 days ago
  10. John Carmack
    100.
    It is easy enough to make your own, but I think standard relu should have been defined as passing the value at zero, so gradients flow backward through it, allowing some things to be zero weight initialized when symmetry breaking isn’t an issue.
    LLM score 72 · 7 days ago
  11. Tim Dettmers / Research Scientist at Ai2
    101.
    Not to degrade from this work, but TurboQuant is not a competitive method nor a good benchmark.›
    LLM score 72 · 7 days ago
  12. Sebastien Bubeck / OpenAI MTS
    102.
    We now know that with an appropriate harness both Mythos and GPT-5.5 can reproduce what our internal model did in one-shot for the unit distance problem.›
    LLM score 72 · 7 days ago
  13. Michael Elabd / DeepMind Researcher
    103.
    I am leaving the Foundational Research team at DeepMind!›
    LLM score 15 · 7 days ago
  14. Lewis Tunstall / Hugging Face Researcher
    104.
    The Carbon tech report is now on bioRxiv.›
    LLM score 65 · 7 days ago
  15. Sholto Douglas / Researcher at Anthropic
    105.
    Huge credit to the OAI team for solving the unit distance problem with 5.5 - it is now my go to example that models can in fact pull together disparate ideas into new discoveries.›
    LLM score 72 · 7 days ago
  16. Yoshua Bengio
    106.
    Looking forward to speaking at BloombergTech next week with @shiringhaffary to discuss AI safety and our research progress at @LawZero_! https://t.co/pLvgDDGd80
    LLM score 15 · 7 days ago
  17. Andrew Lampinen / Research Scientist at DeepMind
    107.
    We've updated the preprint of our Naturalistic Computational Cognitive Science paper — we've clarified and streamlined the arguments, and expanded examples where we see increasing naturalism already yielding new theoretical insights, from RL to perceptual neuroscience.›
    LLM score 72 · 7 days ago
  18. Sayak Paul / Hugging Face Researcher
    108.
    I am bullish and biased, but the best way way use flash attention 3 or 4 is via 🤗 kernels:›
    LLM score 65 · 7 days ago
  19. Boris Cherny / Creator of Claude Code
    109.
    > … [W]e keep finding things that are mysterious, even unsettling.›
    LLM score 65 · 7 days ago
  20. Andrew Ma / ex MTS at xAI
    110.
    Actually this is a great point (from the encyclical)›
    LLM score 65 · 7 days ago
  21. Jason Phang / OpenAI Researcher
    111.
    When did AI psychosis change from meaning “psychosis induced by AI interactions” to “being deluded about the state and capabilities of AI”?
    LLM score 35 · 7 days ago
  22. Behnam Neyshabur / Anthropic Researcher
    112.
    It doesn’t need to be this way.›
    LLM score 65 · 8 days ago
  23. Yoshua Bengio
    113.
    “Like nuclear energy, AI must be at the service of all and of the common good.›
    LLM score 35 · 8 days ago
  24. Sayak Paul / Hugging Face Researcher
    114.
    Consider tests as MVP as those features.›
    LLM score 45 · 8 days ago
  25. Keller Jordan / OpenAI Researcher
    115.
    Computer, analyze what the world today would be like if humans had cured death from old age in the year 1000AD›
    LLM score 25 · 8 days ago
  26. Niklas Muennighoff / AI Researcher at Stanford
    116.
    Great resource on how Cursor Composer works (also useful is https://t.co/TmWQjyv0Lt); much of it still relevant for C2.5! https://t.co/2M28qAADwD
    LLM score 65 · 8 days ago
  27. Damek Davis / Assoc. Professor Wharton Stats
    117.
    In the next year, say a graduate student resolves three problems on par with the Erdos unit distance conjecture, but their proofs arose from simple prompting of ChatGPT.›
    LLM score 45 · 8 days ago
  28. Cameron Wolfe / Researcher at Netflix
    118.
    A lot of research has dismissed the benefits of process rewards over the last few years, but the way that we test if process rewards are helpful is oftentimes flawed IMO.›
    LLM score 72 · 8 days ago
  29. Keller Jordan / OpenAI Researcher
    119.
    Fascinated by the idea that 50,000 years from now will be more like 49,000 years from now than 2026 is like 2006›
    LLM score 25 · 8 days ago
  30. Ronak Malde / DeepMind Researcher
    120.
    I’ve left Google DeepMind. The last two years have been an incredible whirlwind.›
    LLM score 35 · 9 days ago
More