Simulating Problems — LessWrong