Russell shares principles for optimizing Tokio async applications based on discussions at RustConf, emphasizing the balance between fairness and batching, and explaining how to diagnose real performance problems using metrics like scheduling latency rather than assuming issues exist.
A 2023 computer science paper demonstrates that language model tokenizers create unfairness across languages by producing drastically different tokenization lengths for the same text translated into different languages—up to 15 times longer in some cases. These disparities affect cost, processing time, and context length for commercial language services, persisting even in intentionally multilingual tokenizers. The authors argue for developing multilingually fair subword tokenizers for future language models.