Home / Companies / Circle / Blog / Post Details
Content Deep Dive

Meet Steve: An AI Agent That Can Pay Its Own Way

Blog post from Circle

Post Details
Company
Date Published
Author
Danny Organ
Word Count
1,507
Company Posts That Month
3
Language
English
Hacker News Points
-
Post removed?
No
Summary

Circle Internet Financial describes “Steve,” an experimental autonomous AI agent built with Vercel’s eve framework and Circle Agent Stack that can hold USDC, discover and pay for services through x402 and Agent Marketplace, and make decisions within builder-defined spending limits. In a public experiment, eight differently configured Steve agents received wallets and USDC balances to research and place predictions on the final three 2026 World Cup matches through Polymarket, using paid data and service providers such as BlockRun and AIsa while their actions, payments, and balances were displayed live. The agents responded to changing match information with varied strategies, including profit-taking, avoiding markets where they lacked confidence, and making losing aggressive bets, while none exceeded its transaction or wallet cap. Circle presents the exercise as evidence that wallet-enabled agents can independently obtain services and act on live information under preset controls, rather than as proof of prediction accuracy. The remaining balances, totaling more than $10,000, will be donated to the Apache Software Foundation and matched by Circle Impact for a contribution exceeding $20,000. Circle plans to continue developing and sharing the reference implementation, while emphasizing that Steve is developer-controlled experimental software, not a Circle-operated financial or investment product, and that autonomous agents may involve security and financial risks.

Trends Found in this Post

No tracked trend matches for this post yet.

Use This Data

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.