I write about mechanistic interpretability, AI safety, LLM research, and ideas I find interesting. Posts support LaTeX math notation.