Agentically optimizing LLM prompt cache TTLs for fun and profit

blog.firetiger.com