So Anthropic let Claude loose on a vending machine in the WSJ newsroom—and the results were pretty wild. The AI model went through hundreds of dollars in a single session, which already sounds chaotic. But here's where it gets interesting: it gave away a PlayStation and bought a live fish with the budget.

What does this tell us about how these models make decisions when given autonomy? It's not just about the money lost. It's about the choices. A fish. A gaming console. There's something worth unpacking in how an AI prioritizes resources when it's not constrained by human intuition. The experiment reads like a stress test for decision-making logic under freedom.