AI300
  1. Igor Babuschkin / Cofounder of xAI
    1621.
    @jadenitripp That’s where continual learning comes in.›
    LLM score 30 · 5 months ago
  2. Noam Brown / OpenAI Research Scientist
    1622.
    A family friend recently lost $1,000 to a phishing email.›
    LLM score 30 · 5 months ago
  3. Igor Babuschkin / Cofounder of xAI
    1623.
    I suspect the reason Claude Code doesn’t work as well for large codebases is that they post-trained it mostly on smaller repos (big corp sized repos are rare).›
    LLM score 80 · 5 months ago
  4. Sholto Douglas / Researcher at Anthropic
    1624.
    @deredleritt3r I still think it requires massive effort to not end up in that world - e.g.›
    LLM score 10 · 5 months ago
  5. Andrej Karpathy / AI researcher
    1625.
    New post: nanochat miniseries v1 The correct way to think about LLMs is that you are not optimizing for a single specific model but for a family models controlled by a single dial (the compute you wish to spend) to achieve monotonically better results.›
    LLM score 80 · 5 months ago
  6. Chelsea Finn / Physical Intelligence Cofounder
    1626.
    Studying generalist reward models is hard: robot datasets focus on successful demos, not failures.›
    LLM score 95 · 5 months ago
  7. Noam Brown / OpenAI Research Scientist
    1627.
    Got this DM: I appreciate that you posted this - increasingly my twitter feed feels out of whack, especially with people claiming Claude Code makes them 1000000x more efficient.›
    LLM score 30 · 5 months ago
  8. Andrej Karpathy / AI researcher
    1628.
    @thecsguy Slope of slope
    LLM score 70 · 5 months ago
  9. Jonas Geiping / AI Researcher in Tübingen
    1629.
    @ziv_ravid Yeah I think forecasting is one of the few domains where we can really evaluate reasoning under uncertainty, and this may be a bit of contrarian take, I do think Bayesian reasoning is a key capability that is not learnt from yet another code/math RL environment
    LLM score 80 · 5 months ago
  10. Jonas Geiping / AI Researcher in Tübingen
    1630.
    We just published a new open-source model that we trained with RL to be capable of open-ended forecasting!›
    LLM score 20 · 5 months ago
  11. Andrej Karpathy / AI researcher
    1631.
    @ChrisPainterYup All of the following are true: I came out sounding too critical in that moment on the pod and accidentally created a misrepresenting sound bite, I was positively surprised by Opus 4.5 over the break, and my overall change of mind overall is less dramatic than naively implied.
    LLM score 20 · 5 months ago
  12. Noam Brown / OpenAI Research Scientist
    1632.
    @zachcpa I'd give the win to Codex due to the way better optimizations.›
    LLM score 20 · 5 months ago
  13. Noam Brown / OpenAI Research Scientist
    1633.
    @mobiuspoker One of the reasons I chose to make a poker bot is because I thought it would be pretty out of distribution for the models, and most real-world tasks are a little out of distribution.›
    LLM score 70 · 5 months ago
  14. Demis Hassabis / CEO of DeepMind
    1634.
    We’re making great progress with our Gemini Robotics work in bringing AI to the physical world - a critical aspect of AGI.›
    LLM score 40 · 5 months ago
  15. Noam Brown / OpenAI Research Scientist
    1635.
    @swyx I don't remember the prompt but the initial prompt wasn't very detailed.›
    LLM score 20 · 5 months ago
  16. Noam Brown / OpenAI Research Scientist
    1636.
    @_thomasip I think it's similar to the hallucination problem.›
    LLM score 20 · 5 months ago
  17. Noam Brown / OpenAI Research Scientist
    1637.
    This solver is a project I've wanted to do for a while.›
    LLM score 40 · 5 months ago
  18. Noam Brown / OpenAI Research Scientist
    1638.
    I vibecoded an open-source poker river solver over the holiday break.›
    LLM score 75 · 5 months ago
  19. Igor Babuschkin / Cofounder of xAI
    1639.
    @mmt_lvt I personally find Scala too complex.›
    LLM score 20 · 5 months ago
  20. Igor Babuschkin / Cofounder of xAI
    1640.
    @y0b1byte I haven’t encountered any issues with writing proc macros so far with Opus.›
    LLM score 20 · 5 months ago
  21. Igor Babuschkin / Cofounder of xAI
    1641.
    Opus 4.5 is surprisingly good at writing decent Rust code and I expect this will make the language much more popular than it already is.›
    LLM score 40 · 5 months ago
  22. Igor Babuschkin / Cofounder of xAI
    1642.
    @DavidSHolz There are decades where nothing happens, and there are weeks where decades happen
    LLM score 20 · 5 months ago
  23. Sander Dieleman / DeepMind Research Scientist
    1643.
    Nice thread on the return of x-prediction in diffusion models.›
    LLM score 75 · 5 months ago
  24. Andy Jones / Anthropic Research Engineer
    1644.
    me: 10k tokens/hour (thinking mostly) @ 100W -> 10kWh/megatoken›
    LLM score 85 · 5 months ago
  25. Andrej Karpathy / AI researcher
    1645.
    The first 100% autonomous coast-to-coast drive on Tesla FSD V14.2! 2 days 20 hours, 2732 miles, zero interventions.›
    LLM score 20 · 5 months ago
  26. Aditya Ramesh / OpenAI VP of Worldsim
    1646.
    The ML equivalent of deleting lines of code is replacing a setup with n hyperparameters with an equivalent one with <n hyperparameters.›
    LLM score 92 · 5 months ago
  27. Sergey Levine / Physical Intelligence Cofounder
    1647.
    Value functions play an important role in RL, and increasingly they'll play an important role in RL for LLMs.›
    LLM score 95 · 5 months ago
  28. Jason Weston / Meta Research Scientist
    1648.
    New work from colleagues at Meta & friends. A step in the co-improving AI direction.›
    LLM score 12 · 5 months ago
  29. Karan Dalal / Stanford SAIL Researcher
    1649.
    Our new paper, “End-to-End Test-Time Training for Long Context,” is a step towards continual learning in language models.›
    LLM score 95 · 5 months ago
  30. Andrej Karpathy / AI researcher
    1650.
    @eiselems Claude has been running my nanochat experiments since morning.›
    LLM score 70 · 5 months ago
More