A $20 deal sounds too small to need a court. That's exactly why I find it interesting. Imagine an agent paying another agent $20 to collect a small dataset and return a verified summary. Everything goes smoothly until the delivery arrives. “Done,” says one agent. “Not according to the agreement,” says the other. For a human, you probably just take the loss. Finding a lawyer, filing a claim, waiting months and spending more than the original $20 makes no sense. But agents could be making these tiny transactions constantly. So the problem changes. The question isn't only whether a dispute can be resolved. It's whether it can be resolved cheaply enough to make the original transaction worth doing. That's where @courtofinternet makes sense to me. The dispute route is defined upfront, evidence can be preserved, and a randomly selected jury of AI validators can review the case instead of leaving the decision to a single AI model. It's a small change in the story, but a pretty important one. The agent doesn't need to assume every counterparty is trustworthy. It only needs to know what happens if trust breaks. That feels much closer to the kind of infrastructure autonomous commerce actually needs. https://t.co/NQEdfyqBtg