Interesting to see ppl caring about marshalling overhead when working with LLMs
Some of us still prize compute efficiency, especially those who have been using Python for a long time and are contemplating the new kinds of code patterns that have emerged from data science...