A solvable Newcomb-like problem - part 3 of 3 — LessWrong