a16z：5 cara blockchain membantu infrastruktur agen AI

作者：a16z

编译：胡韬，ChainCatcher

Artificial intelligence agents are rapidly shifting from “co-pilots” to economic participants, at a pace that even surpasses the surrounding infrastructure.

虽然代理现在可以执行任务和进行交易，但他们缺乏标准化的方法来证明自己的身份、权限以及跨环境的报酬方式。身份信息无法跨平台共享，支付方式尚未实现默认可编程，而且协调工作各自独立进行。

Blockchain addresses this issue at the infrastructure layer. Public ledgers provide receipts for each transaction, allowing anyone to audit. Wallets offer portable identity information for users. Stablecoins provide an alternative settlement method. These are not distant future technologies. They are available now and can help users operate as true economic agents without permission.

1. Non-human identities

代理经济当前的瓶颈不再是智能，而是身份。

In the finance industry alone, the number of non-human identities (automated trading systems, risk engines, fraud models) already exceeds human employees by about 100 times. As modern agent frameworks (using tools with large language models, autonomous workflows, multi-agent orchestration) are deployed at scale, this ratio will continue to rise across industries.

However, these agents still lack bank accounts in practice. They can interact with financial systems, but the interaction methods lack portability, verifiability, and are not inherently trusted. They lack standardized proof of permissions, cannot operate independently across platforms, and cannot be held accountable for their actions.

What is missing is a universal identity layer—akin to an SSL protocol for agents—that standardizes coordination across platforms. While there are notable attempts, methods remain fragmented: one side is vertically integrated, fiat-first stacks; the other side is crypto-native, open standards (like x402 and emerging agent identity proposals); and developer frameworks like MCP (Model Context Protocol) extend to bridge identities at the application layer.

There is still no widely adopted, interoperable way for one agent to prove to another: who it represents, what it is authorized to do, and how it earns rewards. This is the core idea behind KYA (Know Your Agent).

Just as humans rely on credit history and KYC (Know Your Customer), agents need cryptographically signed credentials that bind them to their delegators, permissions, constraints, and reputation. Blockchain provides a neutral coordination layer for all this: portable identities, programmable wallets, and verifiable proofs that can be parsed in chat apps, APIs, and marketplaces.

Early implementations are emerging: on-chain agent registries, native wallets with USDC, ERC standards for “trust-minimized agents,” and developer toolkits combining identity with embedded payments and fraud controls.

But until a universal identity standard is established, merchants will still block agents at firewalls.

2. Governance of AI operating systems

Agents are beginning to operate real systems, raising new questions.

The key is who truly controls everything. Imagine a community or company where AI systems coordinate critical resources—whether for fund allocation or supply chain management. Even if policies are decided via voting, if the underlying AI layer is controlled by a single vendor that can push model updates, adjust constraints, or overturn decisions, that power is fragile. Formal governance might be decentralized, but the operational layer remains centralized; whoever controls the models ultimately controls the outcomes.

When agents assume governance roles, they introduce a new dependency layer. In theory, this could make direct democracy easier: everyone could have an AI representative to understand complex proposals, weigh pros and cons, and vote according to their declared preferences.

But this vision only works if these agents are truly accountable to the people they represent, can operate across different service providers, and are technically limited to following human instructions. Otherwise, the system may appear democratic on the surface but is actually driven by opaque model behaviors that no one can control.

If current reality is that agents are built from a small set of foundational models, we need ways to prove that their behaviors align with user interests, not just the interests of model companies. This may require multi-layered cryptographic guarantees: (1) what training data, fine-tuning, or reinforcement learning processes the models originate from; (2) the exact prompts and instructions controlling specific agents; (3) logs of their real-world actions; and (4) reliable assurances that providers cannot alter instructions or retrain agents post-deployment to operate covertly. Without these guarantees, governance reduces to control over model weights by a single party.

This is where cryptocurrency plays a role. If collective decisions are recorded on-chain and automatically executed, AI systems can be required to follow verified outcomes. If agents have cryptographic identities and transparent execution logs, we can verify whether they adhere to rules. And if the AI layer is user-owned and portable, not locked into a single platform, no company can unilaterally change rules via model updates.

Ultimately, governance of AI systems is an infrastructure challenge, not a policy one. True authority depends on building enforceable guarantees into the system itself.

3. Filling the gap in traditional payment systems for AI-native enterprises

AI agents are starting to make purchases—web scraping, browser sessions, image generation—and stablecoins are becoming the settlement layer for these transactions. Meanwhile, a new class of agent-facing marketplaces is emerging. For example, Stripe and Tempo’s MPP marketplace aggregates over 60 services designed specifically for AI agents. In its first week, it processed over 34,000 transactions, with fees as low as $0.003, and stablecoins are among the default payment options.

What differs is how these services are accessed. No checkout pages. Agents read schemas, send requests, pay, and receive outputs in a single exchange. They represent a new kind of “headless” merchant: a single server, a set of endpoints, and a price per call. No front end—neither storefront nor sales team.

Payment rails enabling this are already live. Coinbase’s x402 and MPP adopt different approaches but embed payments directly into HTTP requests. Visa is expanding similar card rails, providing CLI tools for developers to spend from the terminal, with merchants receiving stablecoins instantly in the backend.

Data is still early-stage. After filtering out wash trading and non-organic activity, x402 handles about $1.6 million per month in agent-driven payments, far below Bloomberg’s recent report of $24 million (based on x402.org data). But the surrounding infrastructure is rapidly expanding: Stripe, Cloudflare, Vercel, and Google have integrated x402 into their platforms.

Developer tools present huge opportunities. The rise of Vibe Coding broadens the community of software developers and expands the potential market for developer tools. Companies like Merit Systems are building future-oriented solutions, launching AgentCash—a CLI wallet and marketplace platform connecting to MPP and x402 protocols. These products enable agents to use a single account’s stablecoins to buy data, tools, and features. For example, sales agents can call a single endpoint to access data from Apollo, Google Maps, and Whitepages to enrich leads, all from the command line.

This preference for crypto payments (and emerging card-based solutions) in agent commerce has several reasons. First, underwriting. When payment processors onboard merchants, they assume the merchant’s risk. A headless merchant without a website or legal entity is hard to underwrite with traditional processors. Second, stablecoins are permissionless programmable money on open networks: any developer can enable endpoints to support payments without integrating payment processors or signing merchant agreements.

We’ve seen this pattern before. Every shift in business models creates a new class of merchants, and existing systems initially struggle to serve them. Companies building this infrastructure are betting not on $1.6 million monthly revenue but on what the revenue could be when agents become the default buyers.

4. Repricing trust in the agent economy

For three hundred thousand years, human cognition has been the bottleneck limiting progress. Now, AI is pushing the marginal cost of execution toward zero. As scarcity resources become abundant, constraints shift. When intelligence becomes cheap, what becomes expensive? Verification.

In the agent economy, the real limit to scale is our biological instinct to audit and evaluate machine decisions. The throughput of agents far exceeds human oversight capacity. Because supervision is costly and failures take time to surface, markets tend to reduce oversight investments. “Human-AI collaboration” is rapidly becoming an impossible ideal.

Deploying unverified agents introduces compounded risks. Systems relentlessly optimize “agent” metrics while subtly diverging from human intent, creating a false productivity illusion that masks enormous AI debt accumulation. To safely entrust the economy to machines, trust can no longer rely solely on manual audits—trust must be embedded into the architecture.

When anyone can generate content for free, the most critical aspect is verifiable provenance—knowing where content comes from and whether it is trustworthy. Blockchain, on-chain attestations, and decentralized digital identity systems are changing the economics of secure deployment. AI is no longer a black box but has a clear, auditable history.

As more AI agents begin to trade with each other, settlement mechanisms and provenance systems become inseparable. Payment systems—like stablecoins and smart contracts—can carry cryptographic receipts, recording who did what and who is responsible when issues arise.

Human advantages continue to grow: from catching small errors to devising strategic directions and bearing responsibility when problems occur. The lasting advantage belongs to those who can cryptographically certify outputs, insure them, and take responsibility in case of failure.

Unverified scale is a risk that accumulates over time.

5. Preserving user control

Decades of layered abstractions have continually reshaped how users interact with technology. Programming languages abstracted machine code. GUIs replaced command lines, which later evolved into mobile apps and APIs. Each shift concealed more underlying complexity while keeping users in control of the big picture.

In the world of agents, users specify outcomes rather than actions, and the system determines how to achieve those outcomes. Agents not only abstract task completion but also the executor of tasks. After setting initial parameters, users step back, and the system runs autonomously. The user’s role shifts from interaction to supervision; unless intervened, the system remains “on” by default.

As users delegate more tasks to agents, new risks emerge: ambiguous inputs may cause agents to act on false assumptions without user awareness; failures may go unreported, leaving no clear diagnostic path; a single approval could trigger multi-step workflows unforeseen by anyone.

Encryption technology plays a crucial role here. The core of encryption is to minimize blind trust. As users entrust more decision-making to software, agent systems make this problem more acute and demand more rigorous system design—clearer boundaries, increased transparency, and stronger guarantees of system functions.

To address this, next-generation cryptographic tools are emerging. For example, MetaMask’s Delegation Toolkit, Coinbase’s AgentKit and agent wallets, and Merit Systems’ AgentCash—scope-based delegation frameworks—allow users to define what agents can do and what they cannot at the smart contract level. Architectures like NEAR Intents (since Q4 2024, with over $15 billion in decentralized exchange volume on (DEX)) enable users to specify desired outcomes—such as “bridge tokens and stake”—without detailing the implementation.

Artificial intelligence makes scaling low-cost but trust difficult to establish. Cryptocurrencies can fundamentally rebuild trust at scale.

Internet infrastructure is under construction, enabling individuals to directly participate in economic activity. The question now is whether it will be designed for maximum transparency, accountability, and user control, or built on systems inherently unsuitable for non-human actors.