Tag: Research

  1. What the Astra weight-reading demo actually showed (honnibal.dev)
    AI · · November 4, 2026
  2. An LLM edits a model's weights by hand to fix its reasoning (rikiyat.github.io)
    AI · · November 3, 2026
  3. LLM use in biology preprints has nearly caught up with specialized AI tools (epoch.ai)
    AI · · October 31, 2026
  4. LLMs are shaky judges of whether a research idea is new (huggingface.co)
    AI · · October 30, 2026
  5. AI labs claim their models refuted a core complexity assumption (techcrunch.com)
    AI · · October 29, 2026
  6. AI agents help formalize the Classification of Finite Simple Groups in Lean (cfsg.axiommath.ai)
    AI · · October 28, 2026
  7. Why AI writing is harder to read than human writing (arxiv.org)
    AI · · October 25, 2026
  8. Idle AI agents taught themselves skills by playing on a virtual island (techcrunch.com)
    AI · · October 24, 2026
  9. Claude ran a 64-hour autonomous campaign to design scorpion antivenom candidates (anthropic.com)
    AI · · October 21, 2026
  10. A mathematician asks whether AI brings math a golden age or a dark age (terrytao.wordpress.com)
    AI · · October 19, 2026
  11. Claude traces the mysterious NIST curve seeds to a 1990s random generator (anthropic.com)
    AI · · October 17, 2026
  12. A researcher argues neural networks were never that messy (already-clean.goodfire.com)
    AI · · October 16, 2026
  13. Nathan Lambert expects fast AI progress without a leap to superintelligence (interconnects.ai)
    AI · · October 10, 2026
  14. Epoch tested AI on its own jobs, and it couldn't do them (epoch.ai)
    AI · · October 9, 2026
  15. OpenAI's math proofs don't yet meet mathematicians' standards (techcrunch.com)
    AI · · October 9, 2026
  16. OpenAI publishes 722 AI-generated math manuscripts (github.com)
    AI · · October 7, 2026
  17. NVIDIA's Kumo Tabular predicts on spreadsheets with no training (huggingface.co)
    AI · · October 4, 2026
  18. Mathematicians form an advisory group as OpenAI claims 100 solved problems (terrytao.wordpress.com)
    AI · · September 22, 2026
  19. Physics benchmarks were underrating the models (arxiv.org)
    AI · · September 17, 2026
  20. Why self-tuning research agents don't overfit (amazon.science)
    AI · · September 15, 2026
  21. Near-daily AI use among US adults doubled in six months (epoch.ai)
    AI · · September 15, 2026
  22. Ai2's BenchMIRT digs into what an LLM benchmark actually measures (huggingface.co)
    AI · · September 12, 2026
  23. No, looped transformers are not hiding the model's reasoning (magazine.sebastianraschka.com)
    AI · · September 10, 2026
  24. Claude formalizes Fermat's Last Theorem in Lean (anthropic.com)
    AI · · September 5, 2026
  25. Reasoning models more than doubled the pace of AI progress, Epoch finds (epoch.ai)
    AI · · September 4, 2026
  26. Agents from rival model families found new math records together (arxiv.org)
    AI · · August 29, 2026
  27. A third of new web pages show signs of AI writing, Pew finds (techcrunch.com)
    AI · · August 22, 2026
  28. Most workers use AI on free plans, except in tech and science (epoch.ai)
    AI · · August 18, 2026
  29. AI's math edge may be memory, not reasoning (davidepiffer.com)
    AI · · August 16, 2026
  30. A stricter test finds most LLM-written GPU kernels are broken (arxiv.org)
    AI · · August 15, 2026
  31. Researchers pulled hidden reasoning out of Claude, GPT, and Gemini APIs (simonwillison.net)
    AI · · August 13, 2026
  32. Claude improved a proven bound on the Riemann hypothesis (anthropic.com)
    AI · · August 11, 2026
  33. One in five US workers now hand tasks to AI (epoch.ai)
    AI · · August 8, 2026
  34. Jeff Dean leaves Google to build an AI-run science lab (techcrunch.com)
    AI · · August 7, 2026
  35. The intelligence explosion has a bottleneck nobody models: coordination (epoch.ai)
    AI · · August 5, 2026
  36. OpenAI says an internal model solved ten decade-old math problems (simonwillison.net)
    AI · · August 1, 2026
  37. The reliability gap, not raw capability, is what holds agents back (normaltech.ai)
    AI · · July 30, 2026
  38. Mistral's Leanstral 1.5 is an open model that proves theorems and finds real bugs (mistral.ai)
    AI · · July 14, 2026
  39. AI helps individual scientists but narrows what science explores (spectrum.ieee.org)
    AI · · July 13, 2026
  40. Lilian Weng: the harness matters as much as the model (lilianweng.github.io)
    AI · · July 13, 2026
  41. Claude has an internal workspace, and Anthropic built a tool to read it (anthropic.com)
    AI · · July 7, 2026
  42. AI out-persuades expert humans, but the edge is speed, not eloquence (importai.substack.com)
    AI · · July 6, 2026
  43. One transformer layer can match full RL fine-tuning (arxiv.org)
    AI · · July 3, 2026
  44. AI2's DiScoFormer turns attention into a better density estimator (allenai.org)
    AI · · June 30, 2026
  45. Scaling laws look precise, but the way you fit them is fragile (lilianweng.github.io)
    AI · · June 30, 2026
  46. Where a hybrid model beats a transformer, token by token (allenai.org)
    AI · · June 28, 2026
  47. A Stanford study finds one AI hiring tool quietly screened out Black and Asian applicants (hai.stanford.edu)
    AI · · June 24, 2026
  48. Moebius: a 0.2B inpainting model that keeps up with 12B ones (hustvl.github.io)
    AI · · June 23, 2026
  49. Prompt injection works because models judge text by how it sounds (arxiv.org)
    AI · · June 23, 2026
  50. OpenAI's AI chemist improves a hard drug-making reaction (openai.com)
    AI · · June 22, 2026
  51. The evidence that AI tools quietly erode professional skill (nature.com)
    AI · · June 20, 2026
  52. Research agents leak secrets through their own search queries (huggingface.co)
    AI · · June 19, 2026
  53. A Gemini tutor raised math scores in a Sierra Leone trial (deepmind.google)
    AI · · June 18, 2026
  54. Interconnects: multi-teacher on-policy distillation is the 2026 frontier recipe (interconnects.ai)
    AI · · June 17, 2026
  55. Rio's 397B 'homegrown' open model looks like a Nex-Qwen merge, researchers claim (github.com)
    AI · · June 15, 2026
  56. Anthropic surveys 52,000 Americans on AI hopes and fears (anthropic.com)
    AI · · June 13, 2026
  57. Is grep enough? A new paper says yes, with caveats (arxiv.org)
    AI · · June 10, 2026
  58. Anthropic puts numbers on how much of AI development is now done by AI (anthropic.com)
    AI · · June 4, 2026
  59. Ethan Mollick on when AI helps learning, and when it doesn't (oneusefulthing.org)
    AI · · May 26, 2026
  60. A new paper gives language models a sleep phase (arxiv.org)
    AI · · May 26, 2026
  61. Coding agents lose 30 points when you tighten backend constraints (arxiv.org)
    AI · · May 24, 2026
  62. Muon, the optimizer everyone switched to, quietly kills neurons (importai.substack.com)
    AI · · May 22, 2026
  63. DeepMind opens Co-Scientist, its multi-agent hypothesis system (deepmind.google)
    AI · · May 20, 2026
  64. An OpenAI model disproves a 1946 unit-distance conjecture (openai.com)
    AI · · May 20, 2026
  65. Odyssey's Agora-1 keeps a shared world state across players (odyssey.ml)
    AI · · May 19, 2026
  66. Google's ERA writes expert-level scientific code (research.google)
    Research · · May 19, 2026
  67. arXiv will ban authors for a year if they let AI do all the work (techcrunch.com)
    Research · · May 16, 2026
  68. AlphaEvolve moves from research demo to a production tool, with real numbers (deepmind.google)
    AI · · May 7, 2026
  69. Anthropic asked 81,000 people what they want from AI (anthropic.com)
    AI · · March 18, 2026
  70. Demis Hassabis on what AlphaGo's Move 37 set in motion, ten years on (deepmind.google)
    AI · · March 10, 2026
  71. Import AI 445: Bostrom on when to race, and a math benchmark AI can't beat yet (importai.substack.com)
    AI · · February 16, 2026
  72. A Field Guide to the AI Safety Camps (asteriskmag.com)
    AI · · June 1, 2023
  73. GPT-4: The Reference Model, and What It Withheld (openai.com)
    AI · · March 14, 2023
  74. The Bitter Lesson: Why General Methods Win (incompleteideas.net)
    AI · · March 13, 2019

← all tags