Smart Payment: The Evolution Path of the Next Generation Payment System

Author: @13yearoldvc; Translator: Peggy, BlockBeats

Editor's Note: We are standing at a new tipping point. Agentic Payments are reshaping the fundamental logic of transactions. From ChatGPT's internal checkout to micropayments between agents, and a new network order where machines pay for content—the landscape of the "agent economy" is gradually taking shape.

If you are interested in the integration of AI and blockchain, the implementation path of next-generation payment protocols, or are thinking about the future trend of business automation, this article is worth your time to read.

The following is the original text.

Foreword

This is a long article, but definitely worth reading. It brings together the insights of several leading builders who are shaping the future of Agentic Payments.

This is a long article, but definitely worth reading. It brings together the insights of several leading builders who are shaping the future of Agentic Payments.

We will explore the real problems they are trying to solve, the possible practical implementations of these technologies, and the real bottlenecks behind them. You can think of it as a guided frontier journey. Ten pages of content, covering ideas, experiments, and experiences from those who are building the infrastructure for the machine economy. Get ready to go. As the coordinator of ERC-8004 and an AI advisor to the Ethereum Foundation's Decentralized AI (dAI) team, I have worked closely with builders, researchers, and protocol teams over the past few months, focusing on the intersection of stablecoins, decentralized infrastructure, and AI. This has allowed me to observe the real-time evolution of these technologies firsthand. This article is not only a presentation of research findings but also a real-world perspective from the builders of the smart economy. Special thanks to the following individuals for their review and discussion (listed in alphabetical order by last name): @louisamira (ATXP), @RkBench (Radius), @DavideCrapis (Ethereum Foundation), @nemild (Coinbase/x402), @Cameron_Dennis_ (Near Foundation), @marco_derossi (Metamask), @dongossen (Nevermined), @jayhinz (Stripe/Privy), @sreeramkannan (EigenCloud), @kevintheli (Goldsky), @MurrLincoln (Coinbase/x402), @benhoneill (Stripe/Bridge), @programmer (Coinbase/x402), @FurqanR (Thirdweb), @0xfishylosopher (Pantera Capital). The Shift in Smart Payments A month ago, Stripe and OpenAI launched a new feature that could potentially revolutionize online shopping: you can now buy things directly within ChatGPT. No forms, no redirects, no checkout pages. With just a simple phrase, "Find me a handmade ceramic mug," the system automatically completes the payment using Stripe's "shared payment token." This process seems incredibly smooth, even magical, but it's underpinned by a highly centralized architecture that may limit future innovation. Payment tokens, settlement channels, and even user identities are all controlled by the OpenAI and Stripe platforms. In this model, while smart agents are convenient, they cannot be freely combined and can only operate within a specific ecosystem. This demonstrates future possibilities but also reminds us that without open standards and a neutral settlement layer, smart payments will be platform-locked, hindering their true potential.

At the same time, this new payment process also marks a larger shift: now, the actual transactions are no longer conducted by the users themselves, but by "agents." The interfaces we input have begun to compare prices, negotiate, and even make payments on our behalf. Commerce is gradually being swallowed up by "intelligent agent-based commerce."

Currently, it seems that three things are happening simultaneously: agents are beginning to conduct transactions on behalf of humans; these transactions are likely to be settled on encrypted networks rather than traditional financial systems; this may become a breakthrough application scenario for the true integration of blockchain and artificial intelligence.

Why stablecoins and blockchain? Because the form of these transactions is completely different from the models designed by Visa or PayPal. The smart agent economy is full of small, conditionally triggered, composable, and high-frequency payments—fast, granular, and wide-ranging. After communicating with Robert Bench of Radius, we found that "3V3C" is a very apt descriptive model: Velocity, Volume, Value, Conditional, Composable, and Cosmopolitan. We have observed three emerging behavioral patterns: 1. Humans paying agents (2C, 2B, complex optimization scenarios); 2. Agents paying other agents or humans; 3. Agents paying the entire network. All three behaviors break the fundamental assumptions of traditional payment systems. 1. Human → Agent Chat interfaces are quietly becoming a new consumer entry point. Transactions that were previously initiated in browsers are now being completed within conversations. You can already purchase Etsy items directly through ChatGPT's "Instant Checkout" feature, and Shopify will also integrate this process in the future, powered by Stripe. Google, Amazon, and Perplexity are also testing similar shopping models, allowing AI assistants to help users discover and purchase products within chat windows. These AI front-ends are becoming digital stores, especially in the retail e-commerce (2C) scenario—product discovery, price comparison, and purchase are all completed in one process. Over time, people will increasingly rely on their AI agents as personal shopping guides, travel planners, or booking assistants. Interestingly, these agents behave differently from humans: they can monitor prices in real time and automatically place orders when discounts appear; coordinate multiple transactions (such as booking flights and hotels simultaneously); and pay for data or services on demand, no longer relying on subscriptions (we will discuss this in detail in the "Agent → Network" section). In the short term, most of these payment processes will still be completed through traditional channels such as Stripe or Visa—which is fine. For B2C retail e-commerce, the existing infrastructure is sufficient to support the "human → agent" interface, at least for now.

The real application of crypto payments lies in global sourcing (B2B).

Many overseas merchants and manufacturers still face settlement delays and high costs because they struggle to access SWIFT or traditional correspondent banking systems. For example, in Yiwu, China, the world's largest wholesale market for small commodities, most small merchants have never even heard of stablecoins. But once the regulatory environment matures, this will become a natural application scenario.

Stablecoins enable instant, low-cost, and transparent cross-border value circulation—just as crypto remittances have surpassed Western Union in some regions. Whether on the consumer or enterprise side, we will see new user behaviors that were previously impossible: complex, conditionally triggered transactions conforming to the "3V3C" model, automatically completed by agents in the background. Especially as Large Language Models (LLMs) become more intelligent and their operating costs further decrease, the cost savings from these transactions will far exceed the required token fees. For example: A procurement agent can simultaneously monitor multiple multinational suppliers, automatically splitting orders to the cheapest manufacturer and negotiating shipping costs within budget; A creative agent can package subscriptions to multiple SaaS tools, dynamically renewing or canceling services based on usage. This also means that agents must possess "composability": the output of one agent can become the input of another, forming complex, multi-step workflows (such as agent clusters or cross-model thought chains). In the past, we talked about "funding Lego"; now we need "agent Lego." In practice, "composability" means the need for standardized APIs, message formats, and access control. Without these, agents are like applications without APIs—isolated and unable to collaborate. Therefore, these transactions are too complex, too frequent, and too reliant on composition and collaboration to be coordinated by humans or traditional payment systems—but are a breeze for agents running on programmable payment systems. 2. Agent → Agent In the future, agents will need to "hire" other agents—even humans—to complete tasks. Existing business models (subscriptions, licensing, paywalls) are not suitable for interactions between autonomous software. Payments between agents are often charged per call, per token, or per inference, and the amounts can be as low as a few cents or even less. Imagine a research agent buying 100 API calls from a data agent; a design agent paying a compute node for GPU usage time. These are machine-to-machine transactions, frequent and small in value. For example, one agent might pay another agent $0.003 for 100 API calls, or $0.15 for GPU computation, or even just $0.0001 per inference. Traditional payment systems cannot handle transactions of this scale—credit cards charge a fixed fee per transaction (e.g., 2.9% + $0.30), making them unworkable in this scenario. However, from a user experience perspective, these transactions may not necessarily be settled in a "high-frequency, low-amount" manner. For example, on platforms like OpenRouter, companies send millions of API calls monthly, and settlement is done by topping up points with stablecoins, which is more efficient than going through a payment process for each transaction. A more futuristic scenario is: what if each robot was equipped with an agent responsible for tasks, data, and operations (possibly also through prepaid points)? For example, a drone might need to pay for weather data, navigation updates, or temporarily using a private delivery route. This is why we need a new programmable payment structure. The agent should be able to: set budgets and rules; prepay fees; and settle accounts instantly upon task completion, along with proof of work. In other words, crypto payments enable "atomic payments" between autonomous entities.

As time goes on, proxy payment behavior will no longer be limited to AI services. They may directly "employ" human contributors globally, especially in international markets where stablecoins already have actual payment capabilities. This trend is not far off—we have already seen related experiments in our communications with the builders, and large-scale applications may emerge within the next one to two years.

This model is very similar to the logic of remittances.

Imagine a freelance platform for agents, similar to Fiverr: A marketing agent can automatically commission dozens of micro-influencers in Southeast Asia, paying them automatically once their engagement data reaches a preset threshold; a data annotation agent can recruit annotation workers from Kenya or Bangladesh, paying them small amounts per task in real time, instead of relying on bulk invoices. Once agents can transfer funds instantly and globally, the labor itself begins to resemble an API call. From a market design perspective (a unique advantage of crypto systems), when there are hundreds of thousands of autonomous entities globally, comprised of both humans and agents, another trend will emerge: an intent-based auction market. Agents will compete for task requests. The best-performing agents will receive rewards (such as stablecoins, reputation scores, or on-chain credit); poorly performing agents may lose their deposits or reputation. This is precisely the vision we envisioned and hope to build in ERC-8004. A preliminary model might include: 1. Intent Layer: A shared agent registration system (such as ERC-8004) for issuing structured requests and verifying agent identities; 2. Auction Layer: Task allocation via mechanisms such as Dutch or English auctions; 3. Evaluation Layer: Task completion verified by the crowd, other AI agents, or oracles, and rewards automatically distributed; 4. Settlement Layer: Payment via stablecoins, with reputation and staking status updated on ERC-8004. In the past, decentralization was often considered inefficient—partly due to the slow pace of human action and high coordination costs. But now, agents are eliminating these bottlenecks: they can continuously assess who is best suited to perform tasks, what prices are reasonable, and which data is trustworthy. Blockchain plays the role of a "state coordination layer"—an immutable shared memory system used to record results, deposits, and points; while stablecoins serve as micropayment channels for real-time value exchange (pay-per-answer, pay-per-action). This complex "agent → human" collaboration scenario is precisely the problem that blockchain and stablecoins are best suited to solve. Interoperability allows agents to communicate, and composability allows them to collaborate. 3. Agent → Network Another noteworthy trend is that network users are no longer just humans; more and more content is being crawled, read, and interacted with by AI agents, and in the future, it may even be dominated by agents. This means that websites will no longer only charge humans, but will begin to charge machines—a practice we call "pay-per-crawl." For example, publishers are fighting back against unrestricted content scraping. Anthropic recently paid $1.5 billion to settle a copyright lawsuit with an author—one of several cases testing whether AI companies can freely use copyrighted content. OpenAI, Microsoft, Meta, and others are also involved in similar controversies. The likely outcome is that training data and content usage will adopt a "pay-per-access" model. Meanwhile, Cloudflare (which reportedly handles about 20% of web page requests) is already experimenting with a new model: websites can charge agents nano-level fees (even less than micropayments) to allow them access to data. They also recently launched their own stablecoin—NET Dollar. This is where crypto payments come into play again. Websites and APIs can open a "pay-to-play" interface, where agents can read, query, or consume content for a few cents or even less, without subscriptions or ads. This transforms the network into a system of microservices where value flows no longer depend on monthly billing cycles but occur in real time. If you're interested in the early days of the internet with the "402 status code" and related discussions by Andreesen and others, Jay Yu of Pantera Capital has written a great article that delves into this evolution. Imagine a mailroom in the 1950s, where people are sealing bills and invoices into envelopes. This inefficient process is one of the reasons we ultimately adopted the "Net 30" payment terms. In reality, the economic model of "pay-per-crawl" will exhibit a power-law distribution. Only a few high-traffic or high-value websites—those possessing the data that proxies truly need—will proactively integrate this type of monetization logic. For most websites, the cost of measuring, charging, and settling proxies' traffic will far outweigh the revenue. In other words, we believe that ultimately only a few large publishers will capture the majority of the revenue, while long-tail websites will remain open access or unable to monetize. This is precisely where intermediary platforms like Cloudflare can potentially change the curve. If Cloudflare allows websites to "enable proxies' payments" with a single toggle—and handles authentication, metering, and settlement via protocols like x402 or Web Bot Auth—the barrier to entry will be significantly lowered. Cloudflare can automatically identify authorized proxy requests, collect nanometer-level fees on behalf of websites, and automatically distribute revenue. In this model, the open web itself gains a native machine-based business layer: any webpage can become a billable API, and any proxy can be seamlessly paid while browsing, crawling, or learning. This trend is not limited to data access. Almost all online services that can be used on a per-use basis are likely to shift to a "pay-as-you-go" model in the future. In a conversation with Louis Amira, co-founder of ATXP, we discussed how businesses can open up new revenue streams through proxy payments. Here are a few examples: LegalZoom can charge $2 per NDA; Netflix can charge per episode at $0.50 if the payment experience is smooth enough; Replit can charge per token, allowing unlimited "vibe-code," charging $1.23 per million tokens; PitchBook or Bloomberg can charge $0.25 for a proxy to pull valuation models all at once; hospitals can charge per record, providing anonymized cancer scan data for model training. Louis had started documenting the "forced upgrades" or "unnecessary paywalls" he encountered—situations where companies could have turned him into a customer by charging per use. Ideally, enterprise developers could quickly launch ad-hoc APIs, billed per use rather than monthly subscriptions; writers or researchers could sell individual texts, charts, or datasets per query. Conversely, proxies could access non-public data APIs on request, querying vendor data that is inaccessible to web pages, using prepaid micro-requests. This model is well-suited for long-tail APIs and enterprise datasets. Coinbase's CDP team has already made early attempts with Payments MCP, allowing LLM to use on-chain tools such as wallets and payment functions without API keys. The internet is beginning to move beyond a collection of subscription bundles and is more like a "real-time billing system"—every interaction has pricing, payment, and settlement, with value continuously flowing. We are still in the early stages, but integration is happening. After completing a full round of research, we conclude that while the potential of smart agent payments is enormous, it is still in its early stages. One of the biggest challenges is that payments are one of the most regulated and complex areas of the internet in terms of permissions. Its implementation often depends not so much on technical feasibility, but on the ability to integrate and interoperate with large enterprises and financial networks. This makes progress inherently slow. For startups, even if the underlying technology already exists, without access from banks, card organizations, or mainstream payment processors, meaningful experimentation is virtually impossible. Future solutions will likely focus on enterprise-level and compliance-oriented applications. Therefore, teams like Catena Labs are building Agent Commerce Kits, focusing on agent authentication, payment interactions between people and agents, and targeting licensed financial institutions, regulatory compliance, and enterprise-level applications. PayPal is likely to explore similar directions. How far are we from truly intelligent payments? Currently, most so-called "agents" are actually only semi-autonomous systems. Technically, they are more like complex workflow automation tools than intelligent agents capable of autonomous shopping or negotiation. As Kevin Li of Goldsky stated, "You can't truly sell 'fully automated commerce' yet; most AI companies are still automating workflows." The short-term opportunity lies in the "semi-autonomous middle ground": human-initiated actions trigger API-level per-transaction settlements, completed through stablecoin channels. While these processes aren't entirely intelligent agent behavior, they already utilize the same infrastructure—low-latency programmable wallets, per-call metering, and instant settlement—the core components upon which future truly "agent-to-agent" commerce relies. Simultaneously, the underlying blockchain also needs to evolve. Smart payments require stablecoin channels to possess high throughput, low latency, and privacy protection. Next-generation payment-oriented public chains are being explored by major players, such as Stripe's new chain Tempo and Circle's native chain. We also anticipate more teams focusing on agency and stablecoins emerging in the Ethereum L2 ecosystem (such as Thirdweb). All of this indicates that the infrastructure for programmable money is being rebuilt from scratch to support millions of micro-payments and nano-payments per second. Furthermore, programmable wallets and server-side architectures must also be upgraded in tandem. If wallets still assume that mnemonic phrases are held by humans, none of this will be possible. Smart commerce requires policy-based server-side hosting—with programmable budgets, rate limits, spending limits, multisignature/TEE control, and auditable authorization mechanisms. This is precisely the significance of programmable wallets: they provide agents with callable key management and policy execution capabilities without requiring them to "hold the mnemonic phrase." As Jamie Hinz of Privy points out, four years ago we might have been trying to transform Fireblocks or MetaMask into this form; today, the entire technology stack is being tailored for agents to complete transactions within a policy framework, rather than relying on cryptography—security and automation are beginning to merge, no longer contradictory. (For a deeper understanding, we recommend reading Privy's article on natural language control and policy execution.) More importantly, this trend is already beginning to emerge. Even Visa and Mastercard are tweaking their networks to accommodate smart agent commerce, launching Trusted Agent and Agent Pay protocols based on Web Bot Auth—indicating that authentication, authorization, and settlement are rapidly converging, whether in blockchain or traditional payment channels. We may be just one or two key breakthroughs away from truly realizing this vision. Once payments become programmable, the way the internet behaves will change. Every action can be priced, paid, and settled in real time. Every agent, whether a model or a human, can be instantly rewarded for their contributions. As the infrastructure matures, two key standards are emerging: ERC-8004 provides a trust layer, allowing agents to discover and collaborate without centralized intermediaries; x402 enables instant, frictionless payments between agents. Together, they form the underlying pipeline of the smart agent economy. We envision a future where Agent A finds Agent B through an ERC-8004 registry, negotiates service details, and then instantly completes the payment via a smart payment protocol like x402, settling on Ethereum, a neutral financial layer. For true agent collaboration to occur, they must be interoperable—able to discover each other, communicate, and exchange data through shared protocols; and composable—capabilities can be layered on top of each other. As Lincoln Murr of Coinbase stated, "Machine-to-machine payments dominated by stablecoin channels could drive widespread adoption of stablecoins across the internet. While Visa and Mastercard still dominate human-facing payments, agents will become the 'Trojan horse' driving the popularization of crypto payments." The internet took 20 years to evolve from web pages to applications, and another 15 years to evolve from applications to platforms. Agents will compress that cycle. Business will no longer be something you "actively do," but rather a process that "happens automatically"—quietly, continuously, and ubiquitously.