@critical @mametsuko @ryan simple markov models are a big pool of states and each state can transition to any other state with a % probability.

so for each word in each post you track the current word, the one after it, and count how many times you did "foo -> bar", then rack up these counts for each transition, and then when done you go over the transitions and replace the counts with a % based on how many counts that transition got compared to all others. one of these states is also a terminator.

so you pick a random symbol to start at, roll a die, pick a transition from there, output the next word, repeat until done.

this assumes you don't finagle a tool that does it for you. i just explained the theory of the models :drgn_science: