I Built an Autonomous AI Agent That Hunts Bounties. Here’s What Happened.

说明与用途

一个自主 AI 代理扫描 232 个任务列表，应用 7 层过滤，生成报告和提案，并发送 Telegram 通知供人工批准，减少手动工作量。

问题

Automating the discovery, filtering, and proposal generation for paid online bounties while avoiding scams and ensuring human oversight.

结果

In 48 hours the agent processed 232 listings, filtered out 216 human‑only and 3 scam listings, and produced a $500 Solana ecosystem report with a ready‑to‑submit proposal, requiring only a brief human approval step.

工作流

• Wake up and check daily drip
• Scan Superteam Earn, Clawlancer/TaskForce/MoltJobs, and GitHub paid issues
• Apply 7 anti‑scam filters (geo, human‑only, ghost sponsors, unverified payers, real‑money requirement, etc.)
• Economic gate: ensure expected value is positive
• Local LLM critic reviews content
• If candidate passes, build deliverable (report/dashboard/thread draft)
• Generate proposal text
• Send Telegram alert with approval command

证据

In the first 24 hours, the agent found 232 listings. After filtering: 216 HUMAN_ONLY access killed, 1 ghost sponsor, 1 real‑money deposit required, 1 country wall

来源

The agent found a $500 bounty asking for a Solana ecosystem report. It fetched live data from CoinGecko, DeFiLlama, and Solana RPC, generated a full markdown analysis with a local LLM, created an interactive dashboard, wrote a proposal letter, and sent me a Telegram notification with the approval command pre‑formatted.

来源