Sneaky Strategies for TDT — LessWrong