The Murder Mystery That’s Not a Game — It’s a Brutal AI Stress Test
A voice-driven murder mystery built on OpenAI’s gpt-realtime API is secretly a brutal stress test for real-time voice AI. With a 30-minute timer, forced authentication, and a separate judge model evaluating evidence, this project reveals the harsh economics and design constraints of productizing frontier voice models. The suspects are the test suite; the mystery is a benchmark in disguise.