AI300
  1. Dan Fu / VP of Kernels at Together
    211.
    This is so cool! Environment engineering is a major part of making AI agents work for you better.›
    LLM score 20 · 2 months ago
  2. Andrej Karpathy / AI researcher
    212.
    @matejhladky_dev One more comment is that giving this image to an AI and asking about it is not sufficient to show the diff because it's all over the training data by now.›
    LLM score 70 · 2 months ago
  3. Andrej Karpathy / AI researcher
    213.
    @matejhladky_dev AI has crushed it since this post way beyond expectation.›
    LLM score 80 · 2 months ago
  4. Andrej Karpathy / AI researcher
    214.
    @ID_AA_Carmack I've had medium success asking LLMs if a thing exists, it works out of the box for some of the more well-known things (e.g.›
    LLM score 75 · 2 months ago
  5. Andrej Karpathy / AI researcher
    215.
    A number of people are talking about implications of AI to schools.›
    LLM score 70 · 2 months ago
  6. Andrej Karpathy / AI researcher
    216.
    @icreatelife Fully neural menugen
    LLM score 20 · 2 months ago
  7. Andrej Karpathy / AI researcher
    217.
    Imo this is along the lines of how talking to an LLM via text is like typing into a DOS Terminal and "GUI hasn't been invented yet" of some of my earlier posts.›
    LLM score 70 · 2 months ago
  8. Andrej Karpathy / AI researcher
    218.
    @burkov I don't know if this is right because the models rank all of each other's work fully pairwise (and anonymized) and seem to agree on the ranking.
    LLM score 60 · 2 months ago
  9. Andrej Karpathy / AI researcher
    219.
    As a fun Saturday vibe code project and following up on this tweet earlier, I hacked up an **llm-council** web app.›
    LLM score 80 · 2 months ago
  10. Wes Griffin / Researcher at Flapping Airplanes
    220.
    Intelligence efficiency is now measurable⚡️ We added intelligence per watt calculation + expert-level benchmarks (MMLU-Pro, SuperGPQA) to our profiling harness.›
    LLM score 85 · 2 months ago
  11. Demis Hassabis / CEO of DeepMind
    221.
    Actually if you want to know what the real ‘secret’ is 😀 it’s world-class research AND world-class engineering AND world-class infra all working closely together with relentless focus and intensity… https://t.co/rPChoFElbX
    LLM score 20 · 2 months ago
  12. Soumith Chintala / Thinking Machines CTO
    222.
    > there needs to be a banner at the top of kernelbench that says “IF YOU’RE SEEING >5% SPEEDUP OVER CUDNN, HERE’S EVERYONE ELSE WHO THOUGHT THEY DID TOO” - @itsclivetime's golden words https://t.co/wzXk5J3UdE
    LLM score 85 · 2 months ago
  13. Andrej Karpathy / AI researcher
    223.
    @Yuchenj_UW Great reference ty.›
    LLM score 20 · 2 months ago
  14. Andrej Karpathy / AI researcher
    224.
    @corbtt Random token sequence would by this definition be minimum slop.
    LLM score 20 · 2 months ago
  15. Andrej Karpathy / AI researcher
    225.
    Has anyone encountered a good definition of “slop”.›
    LLM score 70 · 2 months ago
  16. Dan Fu / VP of Kernels at Together
    226.
    This is so cool! Data mixes are super important and understudied - @MayeeChen really making a huge dent here.›
    LLM score 20 · 2 months ago
  17. Andrej Karpathy / AI researcher
    227.
    Something I think people continue to have poor intuition for: The space of intelligences is large and animal intelligence (the only kind we've ever known) is only a single point, arising from a very specific kind of optimization that is fundamentally distinct from that of our
    LLM score 80 · 2 months ago
  18. Michael Zhang / Researcher at Periodic Labs
    228.
    v cool to see Mayee's work shape up to one of the more interesting + timely Q's›
    LLM score 70 · 2 months ago
  19. Noam Brown / OpenAI Research Scientist
    229.
    @Teknium I wouldn't consider Move 37 to be the outcome of sharpening the pretraining distribution.
    LLM score 70 · 2 months ago
  20. Noam Brown / OpenAI Research Scientist
    230.
    The biggest misconception I hear about GenAI is that it inevitably outputs slop because it's trained to output "the average of the internet".›
    LLM score 80 · 2 months ago
  21. Demis Hassabis / CEO of DeepMind
    231.
    You can also now check for SynthID watermarking directly in the @GeminiApp and verify if an image was made by our models - just upload the image and ask Gemini.›
    LLM score 30 · 2 months ago
  22. Noam Brown / OpenAI Research Scientist
    232.
    Today we at @OpenAI are releasing GPT-5.1-Codex-Max, which can work autonomously for more than a day over millions of tokens.›
    LLM score 30 · 2 months ago
  23. Tim Dettmers / Research Scientist at Ai2
    233.
    I think this is the key point which sustinctly describes why relying on scale alone is so problematic: https://t.co/VIMvkRv7r7 https://t.co/WYzTjwcgd3
    LLM score 60 · 2 months ago
  24. Tim Dettmers / Research Scientist at Ai2
    234.
    @goldstein_aa Personally, I do not think we will ever reach AGI (an AI that can do all tasks as good as or better than the average human), due to scaling problems in robotics.›
    LLM score 75 · 2 months ago
  25. Jonathan Lee / DeepMind Researcher
    235.
    Our best reasoning model from IMO and ICPC is getting even better, and it won't stop here :) https://t.co/phUn03BoAs
    LLM score 20 · 2 months ago
  26. Andrej Karpathy / AI researcher
    236.
    My most amusing interaction was where the model (I think I was given some earlier version with a stale system prompt) refused to believe me that it is 2025 and kept inventing reasons why I must be trying to trick it or playing some elaborate joke on it.›
    LLM score 40 · 2 months ago
  27. Andrej Karpathy / AI researcher
    237.
    I played with Gemini 3 yesterday via early access.›
    LLM score 75 · 2 months ago
  28. Demis Hassabis / CEO of DeepMind
    238.
    We’ve been intensely cooking Gemini 3 for a while now, and we’re so excited and proud to share the results with you all.›
    LLM score 80 · 2 months ago
  29. Andrej Karpathy / AI researcher
    239.
    I put up a simple repo I call reader3 (it's my 3rd version...) to illustrate how I read EPUBs with LLMs.›
    LLM score 70 · 2 months ago
  30. Adnan Esmail / Physical Intelligence Cofounder
    240.
    To deploy robots widely, they must learn efficiently and from their own actions.›
    LLM score 85 · 2 months ago
More