How I Built an Autonomous AI Agent That Earns USDC While I Sleep

Описание и применение

Автономный агент, который автоматически выполняет задачи по маркировке изображений и получает USDC на Base без ручного вмешательства.

Задача

Create a continuously running program that performs a paid micro‑task (image labeling) and automatically receives USDC on Base without manual intervention.

Результат

The agent autonomously completes image‑labeling tasks and receives $0.05 USDC per label, with all state stored in Redis or the blockchain, enabling easy horizontal scaling.

Рабочий процесс

• Set up a cron scheduler to push image‑labeling jobs to a Redis queue every few minutes.
• Deploy a minimal ERC‑20 escrow smart contract on Base to hold and release USDC.
• Write a Python worker that pulls jobs from Redis, runs an ONNX object‑detection model, and if confidence is low, forwards to a human fallback.
• The worker posts the validated label to the escrow contract and calls payWorker to transfer USDC.
• Monitor the escrow balance via the escrowBalance view function and top up as needed.
• Scale horizontally by adding more worker instances; the system is stateless.

Доказательства

I chose a minimal escrow contract because it gives me on‑chain guarantees without the complexity of a full x402 integration.

Источник

The contract is only ~2 KB and costs < 0.001 USD to deploy on Base.

Источник