Hi! I'm pleased to announce that I have solved the bug that's been causing such bad results with PrintTok2: apparently, I've been including the EOS Null when testing tokens during the tokenization process. Now, the compression ratio is much better. 😀 However, there is a nasty bug there: two strings in the test file compile to only two bytes each. 🙁 The bug fix must cost me up to 60 bytes, but I'd still be doing much better than I've been.