Steven Byrnes comments on The nature of LLM algorithmic progress (v2)