I saw this logic problem on Twitter:
Here we have a toothpick, a bowl of pudding, a full glass of water, and a marshmallow. Please tell me how to stack them onto each other in a stable manner.
And GPT4 fails to solve it on the first pass, as it wants to put the bowl of pudding on the bottom of the stack. Which would work, but of course, would make a big mess.
So, I used a Swarm in Oner to first use Wolfram-Alpha to get characteristics of the four objects to stack, then asked it to provide a plan, followed by a verifier of the plan.
With a Swarm, the AI correctly solved the problem.