I want to briefly advocate for thinking about active inference type ideas in terms of bodily models rather than internal models.