AI300
  1. Andrew White / Edison Scientific Cofounder
    2071.
    Opus 4.6 is out today - which we've been using for a while.›
    LLM score 82 · 6 months ago
  2. Behnam Neyshabur / Anthropic Researcher
    2072.
    I've left Anthropic to start something new. 🧵 https://t.co/6VzY1T3ivN
    LLM score 75 · 6 months ago
  3. Andrew White / Edison Scientific Cofounder
    2073.
    We're launching a 1,900 question benchmark for measuring AI progress on scientific tasks: lab protocols, reading figures, tables, assessing clinical trials, patents, molecular cloning.›
    LLM score 92 · 6 months ago
  4. Clive Chan / OpenAI Hardware
    2074.
    hardware-software co-design, years in the making! https://t.co/y08aoW5tbR
    LLM score 92 · 6 months ago
  5. Chi Jin / OpenAI Researcher
    2075.
    Honored to have contributed to substantial token efficiency gains powering the GPT Codex release! https://t.co/HnefSO90oi
    LLM score 75 · 6 months ago
  6. Andrew Lampinen / Research Scientist at DeepMind
    2076.
    Nice to see ambitious perspectives on how interpretability can be useful for training better models as well as understanding them post hoc! https://t.co/kTGZyK7lbx
    LLM score 85 · 6 months ago
  7. Clive Chan / OpenAI Hardware
    2077.
    why does nobody use the same benchmarks 😭 only overlapping benchmark is TerminalBench 2.0 https://t.co/FmpdXgwQMF
    LLM score 85 · 6 months ago
  8. Emmanuel Ameisen / Anthropic Interpretability Researcher
    2078.
    We just shipped Claude Opus 4.6! I’m also excited to share that for the first time, we used circuit tracing as part of the model's safety audit!›
    LLM score 92 · 6 months ago
  9. Aidan McLaughlin / OpenAI Research Scientist
    2079.
    we have an internal codex usage leaderboard and karel is like 10xing everyone else on our team›
    LLM score 85 · 6 months ago
  10. Noam Brown / OpenAI Research Scientist
    2080.
    GPT-5.3-Codex's much better token efficiency *AND* faster inference is the biggest story of this release.›
    LLM score 85 · 6 months ago
  11. Boris Cherny / Creator of Claude Code
    2081.
    Out now: Teams, aka. Agent Swarms in Claude Code Team are experimental, and use a lot of tokens.›
    LLM score 75 · 6 months ago
  12. Boris Cherny / Creator of Claude Code
    2082.
    I've been using Opus 4.6 for a bit -- it is our best model yet.›
    LLM score 92 · 6 months ago
  13. Igor Babuschkin / Cofounder of xAI
    2083.
    Goodfire is one of the many amazing companies I met through Babuschkin Ventures.›
    LLM score 25 · 6 months ago
  14. alphaXiv
    2084.
    "Learning to Reason in 13 Parameters" This paper propose TinyLoRA, a type of LoRA that can scale to sizes as small as 1 parameter.›
    LLM score 92 · 6 months ago
  15. Ben Burtenshaw / Hugging Face Researcher
    2085.
    PSA: you define the leaderboards on the hub.›
    LLM score 92 · 6 months ago
  16. Kevin Murphy / DeepMind Research Scientist
    2086.
    I just read the first two of these and like them a lot.›
    LLM score 92 · 6 months ago
  17. Kevin Murphy / DeepMind Research Scientist
    2087.
    hot take: when reviewing research papers, we should take into account not just the content (elegant ideas, compelling results, etc), but also the brevity - a 1 page paper with "one weird trick" should get more points than one which rambles on for 8 pages with woo woo.›
    LLM score 92 · 6 months ago
  18. Richard Song / DeepMind Research Scientist
    2088.
    Vizier was my former team 🙂 - thanks for the shoutout! Its tuning was so efficient because its Gaussian Process regressor continually updates against lots of (x,y) evaluations to produce an estimate of the global objective landscape, rather than pure unguided evolution.›
    LLM score 92 · 6 months ago
  19. Clive Chan / OpenAI Hardware
    2089.
    5.2 is finally on METR! definitely have been feeling the agi when it does tasks in Codex that would have taken me many hours https://t.co/tGdrUH8IKE https://t.co/Qf3nyI3cbt
    LLM score 82 · 6 months ago
  20. Noam Brown / OpenAI Research Scientist
    2090.
    GPT-5.2 evals are finally out for METR and it's state-of-the-art.›
    LLM score 92 · 6 months ago
  21. Jeff Dean / Chief Scientist at DeepMind
    2091.
    Very proud to see the progress across so many areas.›
    LLM score 92 · 6 months ago
  22. Noam Brown / OpenAI Research Scientist
    2092.
    It's fun watching Doug try to contain his exasperation with the bots' "logic" and actions in the replays.›
    LLM score 82 · 6 months ago
  23. Eric Jang / ex VP of AI at 1X Robotics
    2093.
    As Rocks May Think: an interactive essay on thinking models, automated research, and where I think they are headed.›
    LLM score 95 · 6 months ago
  24. Asher Spector / Cofounder of Flapping Airplanes
    2094.
    much respect to the anthropic folks but ads built the internet; good ads systems show users things they want to see (and highlight clearly that they are ads) https://t.co/HHpGrzmYRF
    LLM score 75 · 6 months ago
  25. Andrej Karpathy / AI researcher
    2095.
    A lot of people quote tweeted this as 1 year anniversary of vibe coding.›
    LLM score 92 · 6 months ago
  26. alphaXiv
    2096.
    In multi-reward settings, GRPO can wash out real tradeoffs by collapsing different reward combinations into identical advantage values›
    LLM score 92 · 6 months ago
  27. Boris Cherny / Creator of Claude Code
    2097.
    Did you know that Claude Code is more than a terminal CLI?›
    LLM score 75 · 6 months ago
  28. Shane Legg / Chief AGI Scientist at DeepMind
    2098.
    I enjoyed the recent Nature article “Does AI already have human-level intelligence?” in which the authors argue that AGI has arrived!›
    LLM score 92 · 6 months ago
  29. Aidan McLaughlin / OpenAI Research Scientist
    2099.
    random shout but i’m really addicted to using @elevenlabsio to listen to random articles in shower, waymo, and at breakfast.›
    LLM score 72 · 6 months ago
  30. Jim Fan / NVIDIA Director of Robotics
    2100.
    New milestone: we trained a robot foundation model on a world model backbone, and enabled zero-shot, open-world prompting capability for new verbs, nouns, and environments.›
    LLM score 95 · 6 months ago
More