Agents Will Pay Like Locals, Not TouristsAI Agent 会像本地人而非游客那样付款
AI agents won't swipe cards like tourists — they'll build vendor relationships, negotiate B2B credit terms, and settle on stablecoin rails. The window for entrepreneurs is now.AI Agent 不会像游客刷信用卡——它们会建立商家关系、谈判 B2B 信用条款，并通过稳定币轨道结算。留给创业者的窗口期就是现在。
The Setup
A tourist haggles with every vendor, pays with cash or card, trusts no one. A local walks straight to her regular butcher — on credit. When we imagine how AI agents will pay for things, we keep thinking like tourists. a16z crypto’s Sam Broner argues we should think like locals.
Agents behave more like businesses than consumers. They’re infinitely duplicable, but the best platforms are not — meaning a small number of dominant agents per vertical will emerge, each with pre-negotiated vendor relationships, volume pricing, and B2B credit lines. ChatGPT already has partnerships with Shopify, Amazon, and Expedia. Smaller agents are stuck paying retail.
Key Takeaways
- Two payment relationships: user → agent (subscription/credit), and agent → vendor (B2B terms, net-30 invoices). The second is where the real volume flows.
- Cards are a bad fit: Built for humans in the loop, cards can’t handle sub-cent micropayments (Visa minimum: $0.30), streaming payments, or first-time cross-border agent transactions.
- Stablecoins fill the gap: Programmable, global, no 30-cent floor, trivially integrated into APIs. An agent streaming $0.001/sec to a compute provider uses the same rail as a $50,000 B2B invoice.
- Payments are sticky: Agents built on stablecoin rails today will keep using stablecoin rails tomorrow. Early relationships compound.
Why It Matters
This essay reframes the entire “agents + payments” narrative. It’s not about making checkout smoother — it’s about who controls the financial infrastructure of the agentic economy. Cards have an innovator’s dilemma problem: they can’t upgrade fast enough, and the fee model punishes the exact use cases agents need most (micro and macro extremes).
The stablecoin opportunity isn’t theoretical. Agents are being built now. The infrastructure gaps — statements, arbitration, batch approvals, credit lines — are the actual product surface for the next wave of fintech.
What to watch:
- Stablecoin payment infrastructure startups targeting API-heavy agent workflows
- Whether ChatGPT/OpenAI builds native stablecoin settlement for its vendor partnerships
- Clarity Act passing — unlocks yield-bearing stablecoins, massively expands the use case
背景
游客在集市里和每个摊贩讨价还价，用现金或信用卡付款，对谁都不信任。本地人直接走向熟悉的屠夫——赊账结算。当我们想象 AI Agent 如何付款时，我们总是以游客视角思考。a16z crypto 的 Sam Broner 说：应该用本地人的视角。
Agent 的行为更像企业，而不是消费者。Agent 可以无限复制，但最好的平台不能——这意味着每个垂直领域都会出现少数几个主导 Agent，各自拥有预谈判的商家关系、批量定价和 B2B 信用额度。ChatGPT 已经与 Shopify、Amazon 和 Expedia 建立了合作。小 Agent 只能按零售价付款。
关键要点
- 两种支付关系：用户 → Agent（订阅/信用），Agent → 商家（B2B 条款、Net-30 发票）。真正的交易量在第二种。
- 信用卡不适合：为有人参与的场景设计，无法处理亚分级微支付（Visa 最低 $0.30）、流式支付，或首次跨境 Agent 交易。
- 稳定币填补空白：可编程、全球化、无 0.30 美元门槛、可轻松集成到 API。Agent 向计算提供商每秒流式支付 $0.001，和处理 $50,000 的 B2B 发票，用的是同一条轨道。
- 支付具有粘性：今天建立在稳定币轨道上的 Agent，明天仍会使用稳定币轨道。早期关系会复利增长。
为什么重要
这篇文章重新定义了整个”Agent + 支付”叙事。这不是关于让结账更顺畅——而是谁控制 Agent 经济的金融基础设施。信用卡面临创新者的困境：升级速度不够快，且费用模型恰好惩罚 Agent 最需要的使用场景（微支付和大额支付两个极端）。
稳定币机会不是纸上谈兵。Agent 现在就在被构建。报表、仲裁、批量审批、信用额度——这些基础设施缺口，才是下一波金融科技的真正产品机会。
值得关注：
- 针对 API 密集型 Agent 工作流的稳定币支付基础设施创业公司
- ChatGPT/OpenAI 是否会为其商家合作内置稳定币结算
- Clarity Act 通过——解锁生息稳定币，大幅扩展使用场景
Related Essays相关文章
Join Rex's Lab on Telegram 加入 Rex's Lab
Crypto · AI · Investing — raw thinking, before it becomes a tweet. 加密 · AI · 投资 — 推文之前，更原始的思考。