In the past week, a few fairly interesting things have happened in the crypto world—almost all of them point in the same direction: AI is truly starting to enter the on-chain world.

First, leading exchanges have successively rolled out wallet solutions for AI Agents, and then payment protocols have begun to think about how Agents settle with each other. On top of that, each company is designing things like Skills or Recipes. On the surface, it looks like a race among big players, but if you look more closely, the real change behind it may be a redefinition of what a wallet itself is.

Speaking of which, AI entering Crypto isn’t something that just started one or two days ago. In the beginning, everyone talked up concepts like AI tokens and compute networks, but this year the situation has changed. Now nobody is only saying, “AI will change Crypto” anymore—they’re actually putting AI into wallets, exchanges, and payment processes.

Last year, the AI features inside wallets were still quite basic. For example, some wallets’ AI assistants could help you summarize token information, sort out social hype, and generate narrative cards, but basically they were just helping you understand information. That is helpful for newcomers, but only to that extent. The real shift is that AI has started moving from an “information summarizer” to an “executor.”

An exchange’s Agentic Trading is a good example. Users can use natural language to tell AI your trading goals, and then AI itself can call market APIs, place orders, and monitor risk. This is no longer the same old workflow of clicking a few buttons and entering prices—AI is truly participating in decision-making and execution.

The same is true at the wallet layer. Nowadays, everyone is designing permission frameworks so Agents can automatically complete actions like transfers, Swaps, and staking within boundaries set by users. Some wallets have even proposed the Pact protocol to define what Agents can and cannot do; others use a Recipe-driven approach to provide Agents with pre-set paths.

But this isn’t the most interesting part. The concept worth paying close attention to is Skill.

A Skill is essentially the packaging of a specific working method into a reusable capability unit. Each Skill is like a folder: inside, there is documentation that tells the AI what this capability is for, when to use it, and how to operate it. It seems simple, but the logic behind it is deep.

The emergence of Skills means that competition in the Agent ecosystem is no longer just a matter of model parameters and inference speed—it becomes competition over work methods, community knowledge, and reusable capabilities. A platform’s moat may no longer be “my model is smarter,” but rather “here, can users turn their own experience into assets that others can call on again and again?”

This is especially important for the wallet industry. On-chain operations may look simple, but behind them are all the complex accumulation of experience. A single Swap involves token recognition, contract verification, authorization management, slippage settings, Gas estimation, and phishing risk detection. Cross-chain involves choosing the bridge, handling time-to-arrival, and dealing with fees. Participating in DeFi involves yield analysis, contract risks, and liquidation risks.

If these experiences exist only in people’s heads, it’s hard to pass them on. And if you write them as tutorials, it’s still difficult for them to be automatically invoked during real operations. But if they can be consolidated into Skills, templates, or checklists, then in the AI era they become knowledge assets that every user can reuse. This is also the direction imToken wanted to expand when rolling out UI 3.0 for the app.

From this perspective, the future value of a wallet isn’t just whether it can “help me complete an operation,” but whether the experience behind these operations can be consolidated, reused, and improved together.

As AI begins to participate in on-chain operations, the questions wallets need to answer change as well. It’s no longer just “Can I sign this transaction?”—it becomes “What does it mean? Does it match my true intent? Is the authorization scope reasonable? Has the risk been explained clearly? Can I turn my experience into something I can share with those who come later?”

Over the past year, imToken’s thinking around Verifiable UI was essentially laying down the foundation for wallets in this new era. It may sound a bit counterintuitive—if the AI is smarter, the interface actually needs more verification. But that’s exactly the key point. Users must be able to verify the authenticity of the interface. Even if centralized services go offline, even if Agents make mistakes, users still need to retain ultimate control over their assets and interactions. This is of the same origin as imToken’s long-standing non-custodial principle.

Non-custodial solutions address “who owns the private key,” while Verifiable UI further addresses “is what I see truly what I’m about to sign.” Therefore, a wallet should not be merely an AI accessory entry point—it should be the final checkpoint for user asset security in the AI era.

But beyond defense, there’s also a more ambitious side—moving from tools to a co-creation platform.

As execution gradually gets taken over by Agents and interfaces become verifiable, the next question for wallets becomes: what can users create here?

As more and more wallets integrate Agents, design permissions, and provide execution capabilities, long-term differentiation may no longer come from “who integrates AI first,” but from “whether users can consolidate their own works here.” For example: an on-chain strategy, a security rule template, an application for a particular scenario, a shareable operation checklist, or a Skill that can be invoked by other users and Agents in the future.

This means users shift from being “those being educated” to being “co-creators.” Knowledge is no longer only unidirectionally produced by project teams; instead, it is continuously supplemented, refined, validated, and reused by real users in real scenarios.

For example, a Skill for “how to safely add tokens,” an interactive guide for “using Layer2 for the first time,” a risk checklist for “identifying phishing approvals,” or a process template for “hardware wallet cold storage.” These works can be simple—a single image, a tutorial, a checklist, a reminder template—or they can be future Skill prototypes that AI can invoke.

From this perspective, imToken’s 10th anniversary isn’t only a review of the past decade of on-chain journeys with users; it can also be an invitation to the future. An invitation for every user to consolidate their experience, questions, methods, and works, and jointly build a knowledge network for wallets in the AI era.

Because a truly vibrant Agent ecosystem isn’t just about models—it’s about a community of co-creators. And wallets may be the most natural starting point for that community.