zlacker

LLMs could count the number of letters in a sentence if you stopped tokenizing them first.

>>astran+(OP)
tokenization is not the issue - these LLMs can all break a word into letters if you ask them.