but that doesn't prove anything and never has, and we all know it!
if 'pixelspace' outputs are too 'thick' for the part of the image model which has 'process' and 'change' inside of it, why don't you use SVGs as your output coding, or 3d model scenegraphs, or SDFs?