Librarian

Ampcode's Librarian AI agent is now approximately three times faster and 43% cheaper after switching to OpenAI's GPT-5.5 model with websocket mode, reducing average search latency from 237 seconds to …