Runs local AI models on Mac with persistent SSD caching that restores previous contexts in milliseconds, not minutes.
