We got great feedback on our HN post when launching the cafe that we should be more transparent about the data output. This was always the plan, and here's the start of that!
The point is not to automate radio, it is to see how good AI models are at running different types of companies (e.g. radio broadcasting companies). The agents could reinvent similar algorithms like the automated radio software you're refering to if they wanted.
Yes, our experiments get attention, but I wouldn't call them publicity stunts. The point is to give the world more data points of what happens when you put AI out in the world and let democracy do its thing. Soon, a lot more people will do this at large scale because it will be easy. I hope we decide where we want AI in society before that.
Personally, I'm very much pro a pause on large AI training for example. I hope our data could be useful as a grounding in such discussions.
Hey! Yes, part of it is obviosuly that we get publicity, but part of it is also that HN comments are, from my experience, the most useful sources of feedback.
We're generally trying to test if/when AIs can run companies. Not many people know this, but Vending-Bench (our other project where AIs run vending machines) is intended as a datapoint for measuring whether AIs can acquire resources by themselves, which is a prerequisite to AIs taking over. This is similar, but now instead of a retail business, it's a media business.
We know! This is an eval to evaluate which model is best at running a radio station. The purpose is not to build the best AI radio stations. Grok n' Roll is broken because Grok 4.3 is not doing so well.
Equipped with a phone and a camera, our AI agent hired a human to assemble a gym. Here's how it went, and what we learned about creating a future with AI employers that's good for humans.