October 2024 Progress in Guaranteed Safe AI — LessWrong