On: Ask HN: How are Markov chains so different from tiny LLMs?

[Isamu]

Right, you can generate long sentences from a first-order markov model, and all of the transitions from one word to the next be in the training set but the full generated sentence may not.