Join the interpretability research hackathon — LessWrong