A Very Simple Model of AI Dealmaking — LessWrong