Results from the language model hackathon — LessWrong