I like the concept. Please correct me if I'm wrong: it looks like you'd need a lot more than "46 to 101 characters" in a message before you can apply this metho

I like the concept. Please correct me if I'm wrong: it looks like you'd need a lot more than "46 to 101 characters" in a message before you can apply this method reasonably. An MD5 hash for any decent-length password is long, and this method only allows you to replace the subset of "confusable" latin chars in a text. For instance: C = 0x0043 = 0x216d When you reach one of these replaceable characters, you either replace it or you don't, which you mark as either 1 or 0. So for the string "password", our binary MD5 hash is "01011111010011011100110000111011010110101010011101100101110101100001110110000011001001111101111010111000100000101100111110011001". That 128 possible replacements needed in the original text. I imagine the original text would have to be at least 10-20 times that length before we found enough "confusable" latin chars to replace. I'm eager to hear what I'm missing, because I do like this method a lot.