The Two-Update Problem, Part 2 — LessWrong