In college (around 2012ish) I made some poems by transcribing a recording of myself mumbling gibberish into Dragon Naturally Speaking, which partially relied on

In college (around 2012ish) I made some poems by transcribing a recording of myself mumbling gibberish into Dragon Naturally Speaking, which partially relied on Markov chains for its ASR. It would output some string of text, and then I'd delete all the words it output from its dictionary; repeat. It worked very well for creating a list of (slant) rhymes, I have not found a better way to do this. As a literary tool, I've found LLMs to be seriously lacking in the ability to fuck around like this.