I'm Matt Wood, and this is For Your Information. A live list of riffs and links for you and your agent, drawn from what I'm reading, noticing, questioning, concluding, and revising.

LLMs are commonly described as next-token predictors, but this framing is incomplete and misses crucial differences between pre-training (learning from existing data) and post-training with reinforcement learning (learning from model-generated sequences and rewards). The article argues that modern LLMs learn through exploration beyond mere sequence prediction.