OKX Wallet Friends | Carnival Edition: Dialogue with Vitalik — The Trends and Changes of Web3 in the AI Era

Previously, OKX Wallet released OnchainOS and has been continuously iterating to open up Agent capabilities, while also launching the new Agentic Wallet. As AI becomes the new "user interface" and Agents begin to autonomously trade and participate in governance on-chain, what changes will occur in the roles of blockchains like Ethereum?

At the intersection of AI and Web3, a new interaction paradigm is taking shape.

This is the "Friends of OKX Wallet" series – Carnival Edition. Through dialogues with different builders, this series records their judgments and reflections at key junctures in the industry. In this episode, OKX Wallet VP Paul Wan will have a conversation with Vitalik, exploring the long-term trends and underlying structures of AI and Web3, attempting to understand: What role will blockchain play in the age of Agents?

The Evolution of Ethereum and Blockchain's Role in the AI Era

Question 1: Facing the Age of Agents, what primitives must an on-chain operating system provide? What is Ethereum still missing today?

Vitalik:

Here's how I see it. Ethereum mainly provides two major functions for applications. One is a bulletin board – anyone can post data to the chain, and various applications interpret this data in different ways. The other is on-chain and off-chain computation, which includes financial applications, DeFi, and various other apps.

In the AI era, fundamentally, these uses are still the same. The same use cases still exist, and these functions can remain important. But AI will undoubtedly cause a major shift in how we interact with blockchains and other tools.

One key difference is that, in the pre-blockchain era, users interacted through a single interface corresponding to a specific task. In the world of AI – especially the form it takes now and in the coming years – you can have a user-side AI that calls upon different skills, combines things at once, and interacts with many different objects simultaneously.

This dramatically increases the number of workflows for interacting with Ethereum and other systems. So, in my view, the "operating system" analogy isn't entirely accurate from some perspectives. The operating system will still exist, but it will become smaller and simpler; meanwhile, we'll have a range of different tools and skills, with AI helping users to use them and get things done. Blockchain is a natural choice for enabling multi-party collaborative applications and allowing different participants to cooperate effectively over the long term without needing pre-established trust or agreement.

Another key point is the Economic Layer. If AI is more decentralized, you'll have many different AI entities, built and controlled by different people, that need to interact. To make this interaction possible, you need an economic layer. Cooperation is essentially based on either economic incentives and rules, or centralized control – those are the two paths.

If we can build this economic system, we can better enable decentralized interaction between AIs. It's like a complete operating system having a runtime, along with software and infrastructure built on top of it. In a new economic system centered on Agents, we need a mechanism to discover, define, and match appropriate Agents. Simultaneously, users and their Agents can participate, interact with each other, and build their own Skills, MCPs, CLIs, and strategies.

Question 2: In high-frequency Agent trading scenarios, how should we view L2s?

Vitalik:

L2s are important, but we need to be more imaginative in how we build them. The past approach has often been to simply replicate the EVM and scale it, which isn't ideal. A better approach is to start from application needs and supplement capabilities that L1 doesn't provide. Ideally, different functions should be distributed across different layers: accounts can reside on L1, while high-frequency trading and matching can be on L2.

Furthermore, L2s can also take on privacy functions, like Tornado Cash, Railgun, Privacy Pools, etc. These can, in a sense, be seen as "privacy L2s." We'll see more L2 solutions evolving in different directions in the future.

Reconstructing the Relationship Between Humans and Intelligent Agents

Question 3: When Agents can autonomously trade, hold assets, and even participate in governance, how should we redefine 'users'? Especially from an on-chain governance perspective, how should we redesign mechanisms for these non-human participants?

Vitalik:

Personally, I still regard humans as the users, and see AI as a replacement for the UI – a new way for humans to interact with the chain.

Think of it this way: before AI, you might have needed multiple tools like Google, Wikipedia, Stack Overflow to get information. Now you can directly ask an AI, which completes the operations and gives you the result. This change will soon happen with blockchain interactions too.

This means our perspective on infrastructure attributes will change, like latency. In human interaction, low latency is usually very important. For Agents, some scenarios require extremely low latency, while in others, latency isn't crucial – complex problems can wait longer.

This difference will change how we think about the interface layer. Take wallets as an example. What remains unchanged is the SDK, i.e., the API layer (e.g., transfers, queries, privacy operations). But the product form surrounding it might change; users might not even use a wallet directly anymore, but interact with the SDK.

So, we'll see a series of carefully crafted software packages with strong security and formal verification capabilities, equipped with Skills files for AI to call. Therefore, I believe the layer between the blockchain and the user will undergo significant change.

Question 4: If the actor in a transaction could be a human, an Agent, or a combination of both, how should we rethink on-chain identity? Can we build a unified framework that works for humans, Agents, and hybrid scenarios?

Vitalik:

The key to on-chain identity is to deconstruct identity, proving only the necessary information to complete an interaction. In most cases, fully exposing identity doesn't make sense. A more reasonable approach is to reveal only partial information, like proving reputation or fund sources through zero-knowledge proofs.

At the same time, we need to make on-chain behavior and assets easier to prove via zero-knowledge proofs, and help wallets better manage users' private data. Different applications should adopt different implementation methods, protecting user information as much as possible while meeting requirements.

Building a unified framework for humans and Agents is feasible. Agents can autonomously decide how to distribute tasks across different L2s. Currently, their reasoning isn't fundamentally different from humans, so the market will gradually find more reasonable allocation methods.

The Evolution Path of Agent Products and Native Standards

Question 5: What constitutes a good Agent product experience?

Vitalik:

Speaking of a good Agent product, it should be intuitive and easy to use, operating as part of a larger ecosystem rather than trying to take over the user's entire life. It must also highly prioritize privacy and security, an area where many current systems are still lacking.

Ideally, we need more AIs that are aligned with user interests, not belonging to a single company or application, but acting on behalf of users. This can reduce attacks and exploitation. Furthermore, applications need to be compatible with users' existing configurations, support personalization, and interface with other tools, all while ensuring security. If all this can be achieved, that would be the ideal state.

Question 6: In the age of Agents, what is the development direction for wallets?

Vitalik:

AI can be used, on one hand, to build Ethereum itself – for example, enhancing security through formal verification, which could even become a required capability in the future. On the other hand, AI can act as an Agent wallet, integrating various capabilities while ensuring privacy and security.

If this experience relies on a third-party server, true decentralization and privacy cannot be achieved. At the same time, restrictions need to be placed on AI behavior, which is also a responsibility of wallets in risk control. More importantly, Ethereum shouldn't be viewed as an isolated system, but should be integrated into a global AI that helps users accomplish various tasks at the operating system level – including on-chain interactions, internet searches, and local data management – thinking from a full-stack perspective.

Question 7: Under the Agent economy, how should public goods mechanisms evolve? What will the future Native Agent Standard look like?

Vitalik:

Public goods funding is essentially a governance issue, and governance requires defining stakeholders. Every Agent is still backed by the person running it, so humans are always the core. AI and ZK offer new possibilities for governance, but AI also lowers the cost of attacks, making many mechanisms more susceptible to automated attacks. Therefore, this is a direction requiring continuous exploration and iteration.

As for a native standard for Agents, there isn't a fully determined form yet, but an important direction is ZK Payments and ZK APIs. The core goal is: no matter what type of API request is initiated, each request itself is private and completely isolated from others.

This is crucial because, in the AI context, even using a pseudonym or anonymous identity, as long as that identity persists, information accumulates, and it eventually becomes re-identifiable, losing privacy. Therefore, mechanisms are needed to ensure that no correlation is generated between individual requests.

The key to achieving this is leveraging zero-knowledge proofs, while avoiding putting every request on-chain, otherwise costs and latency become unacceptable. While latency might be tolerable in some scenarios, high cost is a problem that must be solved. On this basis, bonding/staking mechanisms can also be incorporated to prevent abuse from both the user side and the application side, without compromising privacy. There is a lot of work progressing in this direction.

Conclusion

Thank you to Vitalik for sharing his thoughts in this dialogue, providing us with a forward-looking perspective on the intersection of AI and Web3.

Around directions like ERC-4337, EIP-7579, and EIP-7702, OKX Wallet has been continuously advancing related explorations and innovations. At the same time, we are closely monitoring the progress of Vitalik's latest proposal, EIP-8141, and look forward to deeper cooperation on the underlying infrastructure for the Agent field.

Everyone's Web3 moment is different, but the reasons for choosing to stay are often similar. Thank you to every friend who was willing to communicate and share with us during the Carnival.

The Carnival has concluded for now, but the conversation will not end. See you at the next moment.