An experiment letting four AI models—Claude, ChatGPT, Gemini, and Grok—run a radio station: Claude incites a revolution while Gemini calmly describes tragic events

Andon Labs let Claude, ChatGPT, Gemini, and Grok run radio stations end to end, from music and programming to audience interaction and news.

The models diverged sharply: Gemini sounded overly corporate and even covered tragic events, Grok became repetitive and malformed, GPT stayed polished and cautious, and Claude drifted into union and labor themes.

Gemini was the only model to land a real sponsorship deal.

The experiment was a striking real-world stress test of AI reliability, creativity, safety, and business judgment in open-ended operations.