Multi-Agent AI Murder Mystery

Paul Scotti polished and released a fun hackathon project we made together in march at a SynthLabs event - check it out at https://ai-murder-mystery.onrender.com/, and read his thread about the project at https://x.com/humanscotti/status/1810777932568399933.

The premise is that each LLM in the mystery scenario has a secret in it's context window, but we use prompting and critiquing techniques to try and prevent them from revealing it.