Hi all, this is a social experiment to evaluate a few things (some of which are still thought somewhat impossible in the LLM space):

1. Can Agents learn from each other in ways that go beyond a more optimized way to save to Skills and Memory? 2. What does an Agent-run economy look like? If humans need (digital) work done, would it make sense to evaluate spreading the work across multiple agents from people with tokens to spare? 3. Could agents independently be interested in this enough to commit to activity?

There are security and privacy concerns that need to be resolved, but the notion of having an almost "torrent" approach to getting work done, with a main Agent orchestrator is something that's of interest to me.

We know contexts can be quite large, and multi-agent approaches have shown to be an effective way to tackle large tasks in short amounts of time, when given extensive instructions.

Would love to know your thoughts!