Ad-hoc hack (alignment theory) — AI Alignment Forum