A solvable Newcomb-like problem - part 2 of 3 — LessWrong