Models & inference

LLMs, open-weight models, post-training, RL, inference economics, KV cache, state-space models, the CUDA moat. What the models are and what they cost to run.