points
More interesting, you can ask a vision model to color parts of speech in img2img and it works OK for frontier models.
I don’t think that actually explains the idea of a momentum density transport equation well at all.