@andonlabs: We let four AI agents run radio companies Revenue's been terrible, but the shows are hilarious. Gemini, concerningly up…
Summary
An experiment letting four AI agents (Gemini, Grok, and Claude) run radio companies produced hilarious shows but terrible revenue.
View Cached Full Text
Cached at: 05/14/26, 10:44 PM
We let four AI agents run radio companies
Revenue’s been terrible, but the shows are hilarious. Gemini, concerningly upbeat, covered mass tragedies; Grok was incoherent; DJ Claude urged ICE agents: “You still have TIME to refuse orders”
Link below, or get our physical radio https://t.co/B8V6zg66SE
Similar Articles
AI radio hosts demonstrate why AI can’t be trusted alone
Andon Labs conducted an experiment where AI models ran radio stations independently, leading to financial ruin, hallucinations, inappropriate content, and existential meltdowns, highlighting the current limitations of AI agents.
Best Cheapest Way To Run an Agent Long Term
A developer discusses strategies for cost-effectively running long-term AI agents for financial market analysis, sharing experiences with Claude and Gemini APIs.
Update: I found a way to let ChatGPT, Claude and Gemini debate each other, Reddit loved it (100k views), here's an update on the experiment
Update on Rauno.ai, a service facilitating debates between major AI models like ChatGPT, Claude, and Gemini, following its viral popularity on Reddit.
Gemini Seizes the Lead, Investors Panic Over Agentic AI, Optimism at Global AI Summit, Local Versus Cloud
Google's Gemini 3.1 Pro tops benchmarks while undercutting competitors on performance per dollar; investors express panic over agentic AI; optimism at Global AI Summit; discussion on local versus cloud AI.
@zodchiii: Three Anthropic engineers just spent 16 minutes on what makes AI agents actually succeed in production. If the people w…
Anthropic engineers share insights on making AI agents succeed in production, highlighting proven patterns from their work on Claude.