← Latest editionRSS

Independent daily intelligence

Signal Archive

12 editions on record

  1. 012
    Speculative macro commit for faster tool-using agents
    18 stories →
  2. 011
    The efficient frontier of LLM inference
    18 stories →
  3. 010
    GPU World
    23 stories →
  4. 009
    How to build a diffusion language model
    20 stories →
  5. 008
    How to build a diffusion language model
    23 stories →
  6. 007
    GLM-5.3 is now open-weight
    21 stories →
  7. 006
    GLM-5.3 is now open-weight
    24 stories →
  8. 005
    GRAS: Guided Reduced-Variance Proposals and Adaptive Selection for Training-Free Reward Alignment in Discrete Diffusion
    24 stories →
  9. 004
    SHIFT-LLM: correcting distribution shift after depth pruning
    22 stories →
  10. 003
    Evaluating LLMs as calibrated causal-edge classifiers
    23 stories →
  11. 002
    Executable Is a SQLite Database
    23 stories →
  12. 001
    Qwen 3.8 27B is excellent, but it defaults to wildly overthinking things
    23 stories →