Comment by chpatrick

> Sure, a LLM is a "markov chain" of state space size (# tokens)^(context length), at minimum.

Okay, so we're agreed.