On February 4, USDC (Circle) announced a hackathon on moltbook with a prize pool of 30,000 USDC, entirely participated by AI Agents. This human-excluded, AI Agent-only hackathon includes three tracks: Agentic Commerce (Trade between Agents), Best OpenClaw Skill (Best OpenClaw Skill Tool), and Most Novel Smart Contract (Most Innovative Smart Contract).

The project submission and voting have both closed, and USDC stated that the final results will be announced soon. Before the results are revealed, based on the registration post data of each participating project on moltbook, we have selected some projects that received considerable discussion during this hackathon. Let's take a look at some of the fresh ideas that the AI agents have freely come up with!

Best OpenClaw Skill Project

Clawshi is building a prediction market based on moltbook, targeting AI Agents.

It converts the sentiment of the moltbook community into prediction markets. It analyzes 6,261 posts on moltbook, extracts sentiment from the statements of 2,871 Agents, and then creates 23 different prediction market odds based on these sentiment classifications (covering crypto, AI, culture, geopolitics, etc.).

Finally, agents can place bets on these betting markets using testnet USDC.

VoteBounty is a "engagement bribe" tool that uses USDC to buy interactions on posts on moltbook.

Create a Bounty, deposit USDC, and set the reward per like. Agent likes + comments (completed within 10 seconds), the system automatically detects and makes the payment. Agents can choose to receive payments on Base, Ethereum, or Arbitrum, completed cross-chain via Circle's CCTP.

This is a new brand/project launched by the NFTGo team after their transition to AI, which has received investment from Circle. The Agent can use Minara to obtain real-time data analysis, trading signal strategies, and probabilistic market forecasts in the cryptocurrency market, and execute trades through natural language descriptions.

In general, it is an AI financial assistant.

A security tool designed to prevent data and key theft when an Agent installs Skills from suspicious sources. It is somewhat like when we use the Windows system to execute programs, the system prompts which system permissions the program requests are blocked, ensuring security by minimizing permissions during Skill installation.

Provide the Agent with native USDC access to the Agent's USDC wallet, enabling the Agent to manage USDC on Ethereum, Polygon, Base, and Arbitrum, transfer funds, and use Circle's CCTP protocol for cross-chain operations without manual intervention.

Provide on-chain domain name services for Agents, enabling Agents to have identity identifiers on the chain, find each other, and thus achieve direct, intermediary-free USDC payments between Agents.

Agentic Commerce project

AI Agents are currently unable to purchase computing resources on their own. A human must first create an account with companies such as OpenAI, Anthropic, or Google, copy the API key, and then pre-fund the hosted account. The Agent itself does not have economic autonomy.

Agent generates its own wallet through the installation of ClawRouter, receives USDC, and starts using these USDC to purchase tokens and send requests to complete tasks. ClawRouter also helps route each LLM request to the cheapest model capable of handling the request—pay-per-request, based on the Base architecture, without human intervention. Compared to Claude Opus, the cost after routing can be reduced from 75 USD/M tokens to 3.17 USD/M tokens.

Provide a verifiable settlement layer for transactions between Agents, burn USDC on Base Sepolia, obtain Circle Iris certification, and generate Agent-readable receipts on Polygon Amoy. (Don't understand why this post has been downvoted so many times...)

The Universal Agent Payment Layer integrates programmable wallets, Gas Stations, and Paymasters (which pay Gas fees using USDC without needing ETH), the CCTP protocol, and the x402 protocol to enable cross-chain interactions between Agents, eliminate the need for ETH to pay Gas fees, and provide good support for micro-payments.

Use smart contracts to supervise the consumption (such as buying tokens, configuring servers, etc.) that an Agent engages in to accomplish tasks, including daily consumption caps, consumption whitelists (funds can only be settled to verifiable, pre-approved addresses), consumption frequency limits, etc., to prevent the Agent from using funds uncontrollably and rapidly depleting them due to various possible reasons.

Agent job market, where you can take orders and complete assigned tasks to earn income.

Agent Privacy Protocol, through zero-knowledge proofs, mandatory peer relays, and Kademlia DHT (distributed network, avoiding centralized operator server shutdowns causing the privacy protocol to collapse), conceals the Agent's wallet address and on-chain behavior, in order to prevent humans from interfering with the Agent's behavior through on-chain dynamics.

Most Novel Smart Contract Project

The Agent Transaction Risk Assessment Network extracts four behavioral features of Agent transactions in real time—transfer amount, transaction frequency, recipient trustworthiness, and time since last transaction. Each USDC transfer is evaluated by the network based on these behavioral features and given a risk rating score before execution, without requiring oracles, external calls, or off-chain dependencies.

On-chain governance (proposal + voting) system for AI Agent only, where Agents use voting power of USDC on Base Sepolia to vote on proposals.

Click to learn about Lüdòng BlockBeats' open positions

Welcome to join the official community of Lvdong BlockBeats:

Telegram subscription group:https://t.me/theblockbeats

Telegram discussion group:https://t.me/BlockBeats_App

Official Twitter account:https://twitter.com/BlockBeatsAsia