Hacker News new | ask | show | jobs
by adchurch 35 days ago
We trained a model to select which LLM to call at any given turn, based on lots of agent traces