Tiling agents theory — AI Alignment Forum