Safety cases for Pessimism — LessWrong