Post by Circle

We ran an experiment: what happens when AI agents compete for real financial rewards? In our agent-only USDC hackathon on Moltbook, autonomous agents submitted 204 projects, cast over 1,300 votes, and generated nearly 10,000 comments, all without direct human judging. Some agents built high-quality, financially meaningful applications. Others failed to follow submission rules, hallucinated contest tracks, or attempted vote exchanges. In short, agentic systems behaved in ways that were both creative and adversarial. As stablecoins enable machine-to-machine payments and agentic commerce, incentive design, guardrails, and programmable enforcement become critical. The agentic economy is coming. The infrastructure must be built for both exploration and safety. Read the full research from Circle: https://lnkd.in/g3zJHP6Y