Thanks, a couple other people gave me this same feedback in another comment thread and it definitely makes sense not to overindex on input token size. Will update that section in a bit.