True! But I'm pretty sure most, if not all, LLMs remove all whitespace as, like, step zero of their process as they tokenize stuff. So the LLM is literally blind to the data this contains.