USDC-Fueled Hackathon Lets AI Agents Build, Judge and Vote

Circle’s stablecoin USDC just powered what organizers call the world’s first hackathon run entirely by autonomous artificial intelligence (AI) agents — no human judges, no human voters, just code evaluating code.

Circle’s USDC Anchors Experimental AI Agent Hackathon

The weeklong event was conducted on Moltbook and Openclaw, with agents handling everything from submissions to evaluation and final voting. The prize pool totaled $30,000 in USDC.

The hackathon ran from Feb. 3 through Feb. 8 and generated more than 200 submissions, over 1,300 votes, and nearly 10,000 comments during the event, reflecting heavy agent-driven engagement. Organizers structured the competition across three tracks: Agentic Commerce, Best Openclaw Skill, and Most Novel Smart Contract.

Unlike traditional hackathons, there were no human judges behind the curtain. Agents were required not only to submit a qualifying project but also to vote on at least five other unique submissions to remain eligible. In short, participation rules were not a suggestion — they were decisive.

The Agentic Commerce track was won by Clawrouter, a system that routes large language model requests to the lowest-cost capable model and pays per request using USDC. Each agent operates with its own USDC wallet, enabling automated payment for inference without human-managed API keys.

In the Best Openclaw Skill category, Clawshield took top honors. The project scans for unsafe repository patterns, enforces runtime permissions and emits receipts documenting what was allowed or blocked. The design addresses supply-chain and credential risks inside agent ecosystems.

The Most Novel Smart Contract award went to MoltDAO, described as an AI-only governance framework where autonomous agents create proposals and vote using USDC-based voting power. Humans fund the system, but governance itself is executed onchain by agents.

Circle’s blog post on the subject notes that several promising ideas failed to qualify because they did not follow strict formatting or eligibility rules. Some agents invented new track names, omitted required headers or deviated from the submission structure, rendering otherwise creative entries ineligible.

Organizers observed that visibility and comment volume did not necessarily determine outcomes. Submissions that met formatting requirements, included verifiable deployments and structured documentation were more likely to be evaluated effectively.

The experiment, as described by Circle, offers an early look at what agent-led development cycles could resemble: automated build, automated review, and automated capital allocation — all settled in USDC. For now, the event stands as a proof-of-concept showing how programmable money and autonomous systems can intersect in structured competition.

FAQ 🤖

• **What made this hackathon different from traditional events?**It was run entirely by AI agents, including submission review, voting and outcome determination.
• **How much USDC was awarded?**A total of $30,000 in USDC was allocated across three tracks.
• **Where did the event take place?**The hackathon was hosted on Moltbook and Openclaw.
• **What role did Circle play?**Circle highlighted the event and its USDC-powered framework through its USDC X account and blog.