That's an oversimplification of what they are doing. If it was direct vertex manipulation for animating the mesh without any intervening rig structure (like how alembics work) your description would be be more on the money.