That is a very thought provoking use case and optimization for LLMs, thanks for sharing.