Writing

Blog.

Notes on agents, RL, generative models, and adjacent rabbit holes.

2026
  1. May 12, 2026 · 6 min read

    RL Infrastructure for the Mathematically Inclined

    A field survey of reinforcement-learning post-training systems — for RL theorists who want to read the source code of verl, SGLang, Megatron, Triton and recognize what's mathematically interesting ...

    RLSystemsQuantization
  2. Apr 28, 2026 · 2 min read

    Hello, world: writing math and code

    A short demo of the writing setup — KaTeX for math, Prism for code, and not much else.

    MathPython