Connectionists: A million agents is all you need

[Risto Miikkulainen]

The recent "Illusion of Thinking" paper (https://machinelearning.apple.com/research/illusion-of-thinking) by Shojaee et al. at Apple showed that LLM reasoning is limited on complex problems, especially those requiring long chains of reasoning. However, it turns out that such problems can be solved through large multiagent systems: for instance, a system composed of a million LLM microagents can solve the 20-disk Towers of Hanoi problem that requires a million reasoning steps. The details are described in
    https://arxiv.org/abs/2511.09030
and there’s also a blog (with very cool animations) at
    https://www.cognizant.com/us/en/ai-lab/blog/maker


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20251113/afe2fa19/attachment.html>