2026 is becoming a critical turning point for the deep convergence of Crypto and AI.
Over the past two years, we have witnessed AI's transformation from an "assistive tool" to an "autonomous economic participant." AI Agents are no longer just chatbots that answer questions; they have begun to autonomously initiate transactions, call APIs, manage asset portfolios, and even hire other Agents to complete tasks.
But the prerequisite for all this is that these Agents need identity, payment channels, reputation records, and a verifiable execution environment.
And these needs are precisely the problems that blockchain is best at solving.
As has often been discussed, the Ethereum Foundation established a decentralized AI (dAI) team in September 2025. Vitalik Buterin published a systematic AI strategic framework in early 2026. A series of protocol standards around Agent identity, payment, and execution have already been launched and are running on the mainnet. Meanwhile, new public chain ecosystems like Solana are also building AI infrastructure along their own paths.
Therefore, this article also attempts to use the Ethereum ecosystem as the main axis, supplemented by important developments on other public chains, to outline the complete picture of the current encrypted AI protocol landscape.
I. Vitalik's AI Blueprint: Ethereum Aims to Be the "Trust Layer" of the AI World
In February 2026, Vitalik Buterin published a systematic post on X, specifically revising the "Crypto × AI" crossover framework he proposed two years ago.
In the article, he revisited the ideas proposed two years ago, believing that the accelerated push towards Artificial General Intelligence (AGI) often resembles the unchecked speed and scale that Ethereum itself was created to challenge. He explicitly opposed simplifying AI development as an "AGI race," instead advocating that Ethereum should become a direction-setter for the AI world.
In other words, what he truly cares about is not how to make AI spiral out of control faster, but how to base AI's expansion on verifiable, auditable, and constrained infrastructure.
Overall, Vitalik's framework consists of four core pillars.
The first is trustworthy AI interaction tools. He advocates for using local large language models (local LLMs), zero-knowledge proof payment mechanisms, and other tools to allow users to use AI services without exposing their identity and raw data.
This attitude is not just abstract expression. In April 2026, Vitalik also publicly shared his own local LLM usage plan. After testing multiple hardware setups, he chose to run the 35-billion parameter open-source model Qwen3.5 locally on a computer equipped with an NVIDIA 5090 GPU. All computation is done locally, aiming to increase inference speed to a level usable for daily tasks and minimize reliance on cloud models.
Of course, the symbolic significance of this is greater than the practical significance, but it also shows that, at least in his view, the direction truly worth pursuing for AI is not just stronger models, but more controllable models.
The second is the AI economic coordination layer. This includes Ethereum's ability to use smart contracts to support mutual payments between Agents, security deposits, dispute resolution, and reputation accumulation, enabling programmable economic relationships between machines. The third is AI as the interface for Web3, for example, local AI assistants can help users draft transactions, audit smart contracts, interpret formal verification proofs, and become a bridge for ordinary people to enter the complex on-chain universe.
Finally, there are AI-enhanced governance systems, such as using AI to upgrade mechanisms like prediction markets, quadratic voting, and public fund allocation, finding a balance between automation and human judgment.
Overall, the core idea of this framework can be condensed into one sentence: Ethereum is not about accelerating AI, but about making AI operate in a verifiable, auditable, decentralized environment.
So how exactly is this achieved?
II. From Identity Protocols, to Payment Protocols, to Execution Protocols, to Verifiable AI
If Vitalik's framework is the macro blueprint, then the recent wave of protocol evolution in the Ethereum ecosystem has begun to embed this methodology into a concrete tech stack.
The first infrastructure node most worth watching is ERC-8004.
As an identity, reputation, and verification standard designed by Ethereum for AI Agents, it is led by the Ethereum Foundation's dAI team, with joint participation from Google, Coinbase, and MetaMask in its formulation, almost encompassing the three key entry points of AI, transactions, and wallets (Extended reading: "The New Ticket to the AI Agent Era: What is Ethereum Betting on by Pushing ERC-8004?").
As its official name, Trustless Agents, suggests, its core logic is not complex algorithms, but aims to give AI verifiable identity, reputation, and capability proofs on-chain. Simply put, its design is very restrained, doing only three things:
- Identity Registry: Based on the ERC-721 standard, each AI Agent is "NFT-ized," meaning AI Agents can be looked up, referenced, and integrated into other protocols like wallet addresses;
- Reputation Registry: Can be understood as the "Yelp" for AI, allowing users or other Agents who have actually interacted with an Agent to submit feedback. These evaluations can be linked to on-chain payments or escrow behaviors, ensuring reputation is not a narrative generated out of thin air, but a historical record based on real economic activities;
- Verification Registry: For high-value or high-risk tasks, historical reputation alone is not sufficient. ERC-8004 therefore reserves third-party verification interfaces, allowing capabilities or execution processes of Agents to be endorsed through methods like trusted execution environments (TEE) or zero-knowledge proofs;
If identity answers the question "Who is the Agent?", payment infrastructure represented by the x402 protocol answers the question "How does the Agent transact?".
As is well known, x402 is an open HTTP payment protocol, jointly initiated by Coinbase and Cloudflare. Its basic principle is very clever, reviving the long-dormant 402 status code (Payment Required) in the HTTP protocol. When an Agent attempts to access a paid service, the server returns a 402 status code and a payment request. The Agent uses stablecoins to complete the payment and then gains access.
The entire process is embedded in the HTTP request, requiring no account registration, no credit card, and no manual intervention. In other words, this is a payment system designed for machines, not humans.
It is worth noting that just earlier this month, the Linux Foundation essentially formally took over the x402 Foundation and received the x402 protocol contributed by Coinbase. The official statement was very clear: x402 aims to embed payment directly into HTTP interactions, allowing AI agents, APIs, and applications to exchange value just like they exchange data.
The author believes the importance of this news has been greatly underestimated, on one hand due to x402's potential penetration and significant influence in AI and internet payments, and on the other hand due to the impressively strong lineup. Of course, these giants have always been promoting x402, but this time the effect is clearly 1+1 > 2.
Furthermore, the V2 version of x402 is also striving to achieve an expansion of payment methods, including not only supporting on-chain stablecoins but also兼容 (compatible with) traditional ACH (Automated Clearing House) and bank card networks, to bridge the boundary between AI Agents and the real financial system.
Finally, beyond identity and payment, the third piece of the puzzle that Ethereum has recently added is the execution layer.
In April 2026, Biconomy and the Ethereum Foundation's Improve UX direction jointly promoted ERC-8211, attempting to solve the most realistic bottleneck for AI Agents in the DeFi world. For example, complex on-chain operations are often not a single call, but a multi-step, dynamic, and easily failing execution chain.
We can simply understand it as an "intelligent batching" mechanism specifically designed for AI Agents and complex DeFi operations. Because in traditional on-chain operations, completing a complex DeFi strategy often requires multiple independent transactions: withdrawing funds from a lending protocol, exchanging tokens, and then depositing into another protocol.
Each step requires separate signing and confirmation, which is already cumbersome for human users and is even more of a bottleneck for AI Agents that require high-frequency autonomous operation. The solution of ERC-8211 is to allow multiple blockchain operations to be combined and executed in a single transaction, with each step dynamically parsing the actual value during execution, and proceeding only if predefined conditions are met.
For example, an Agent can complete in one signed transaction: Withdraw funds from Aave → Exchange the actually received amount on Uniswap → Deposit the exchange result into Compound—all executed atomically, without writing a new smart contract.
Looking at these three together, Ethereum's recent thread is already very clear: ERC-8004 answers "Who are you, and why should others trust you?", x402 answers "How do you pay for services?", and ERC-8211 answers "How do you efficiently complete complex operations?".
In other words, what the AI Agent economy truly lacks is never just smarter large models, but a set of open, composable, and extensible protocol stacks; and this is precisely what Ethereum does best.
III. Beyond Ethereum: Solana, DePIN, and Decentralized Computing
Of course, even though Ethereum holds a leading position in standard setting and trust infrastructure, the encrypted AI ecosystem is far more than just one chain.
A more accurate statement is that Ethereum is competing for the standard layer and trust layer, while other ecosystems show different advantages on the execution layer and computing power layer.
Solana is the most typical example. The reason its presence is increasingly felt in the topic of Agent payment stems from the fact that the chain requirements for AI Agents are not about ideological correctness, but about "low latency, low cost, and sufficient stability." Solana's official introduction to x402 directly promotes millisecond-level finality and extremely low transaction costs as important selling points for machine payments. This also explains why Solana more easily承接 (undertakes) those high-frequency, small-amount, instant-feedback-required Agent interaction scenarios.
At the same time, the Agent toolchain around Solana is also rapidly maturing. The Solana Agent Kit official GitHub allows Agents on any model to autonomously execute over 60 types of Solana actions, covering transactions, token issuance, lending, airdrops, Blink, cross-chain, and other scenarios, and is reused by a large number of on-chain projects and developers.
Therefore, looking at today's landscape, the division of labor in encrypted AI is becoming clearer. Ethereum seems more like doing the underlying abstraction of protocol standards, identity reputation, and trusted execution, while Solana holds practical advantages in high-frequency payments and low-friction interactions. The value of decentralized computing power networks will also be re-evaluated as more Agents truly enter production environments.
Overall, looking back from the vantage point of April 2026, the landscape of encrypted AI protocols has taken initial shape:
- Identity Layer: ERC-8004, as Ethereum's leading Agent identity standard, has expanded to multiple chains like Base;
- Payment Layer: x402 has grown from an experimental project by Coinbase to a global standard under the governance of the Linux Foundation;
- Execution Layer: Standards like ERC-8211 simplify complex on-chain operations for Agents;
- Verification Layer: Technologies like zkML, TEE, and cryptographic proofs begin to provide verifiability for high-value Agent interactions;
- Competitive Landscape: Ethereum handles the standard and trust layer, Solana handles the high-frequency execution layer, and Bittensor can perhaps serve as a supplement in dimensions like computing power, forming a complementary rather than zero-sum格局 (situation);
Looking ahead to the second half of the year, the new Ethereum upgrade will likely promote L1 scaling, native account abstraction, and post-quantum security. The普及 (popularization) of account abstraction will undoubtedly significantly reduce the usage threshold for Agent wallets; the deep integration of x402 and ERC-8004 could also give rise to a closed-loop Agent economy, covering Agent registration identity, service discovery, payment initiation, and reputation accumulation, all completed on-chain.
In Conclusion
Ethereum and blockchain are not about accelerating the arrival of AI, but about ensuring that when AI arrives, the world does not spiral out of control.
After all, in the Web2 world, AI's identity is defined by big companies' API Keys, payments are carried by the credit card system, and trust is endorsed by centralized platforms. This system barely functions in scenarios for human users, but under the new paradigm where millions of AI Agents need to collaborate autonomously 7×24, it is increasingly inadequate.
And standard-setters with Ethereum at the core, efficient execution layers represented by Solana, and decentralized computing power supported by DePIN, might just build a whole new set of infrastructure for the AI Agent economy.
