Managing risks of our own work — LessWrong