OpenAI inference cost reduction cut ChatGPT guest traffic from tens of thousands of Nvidia GPUs to just a couple hundred, ...
LLM training data mixture optimization breaks when training pools shift — every prior proxy experiment becomes stale.