Web3 AI Application Overview: Which Ones Are Making You Money, and Which Are Reshaping the Rules?

Original by Odaily (@OdailyChina)

Author: Asher (@Asher_0210)

In recent weeks, discussions around AI + Crypto have heated up once again.

From the AI x Blockchain conference in New York to the recently concluded Web3 conference in Hong Kong, nearly all major participants are revisiting the same question: How will AI reshape the next phase of the crypto industry?

However, unlike previous cycles that were more narrative-driven, this round of discussions is shifting to a more specific question: What problem is AI actually solving?

In a fireside chat themed "Reshaping Convenience: The Next Decade of Web3, AI, and Smart Economy," Binance Co-CEO He Yi mentioned that as the industry matures, the early dividends of the crypto market are fading. Going forward, what matters more is not the technology itself, but whether the product truly provides value and whether people are willing to pay for it.

This means AI is no longer just a new growth narrative; it is being placed in a more concrete context. People are starting to ask more directly: What tangible impact can it actually bring?

Especially in user-facing application layers, this change is becoming evident. In the Web3 space, a wave of applications built around AI is emerging. Some are restructuring how information is accessed, others are redefining ownership of data and memory, and some have begun integrating on-chain research, trading, and even economic models with AI.

These projects may not be fully mature yet, but they collectively point towards a more realistic direction. As dividends wane, the combination of AI and crypto is returning its focus to the product itself.

This article selects several representative Web3 AI projects, examining the practical progress of this application wave across dimensions like information, memory, operations, Agent economies, and distribution.

Surf: A Real-Time Encyclopedia for the Crypto Market

Surf is a typical information layer product in this wave of AI applications. It doesn't attempt to redesign the trading process, nor does it focus on creating new economic systems. Instead, it returns to a more fundamental, yet long-neglected problem—in the crypto market, accessing information itself remains a costly endeavor.

On-chain data, market movements, social sentiment, and project details are often scattered across different platforms. Users must toggle between multiple pages to piece together a relatively complete market picture. This fragmentation becomes more pronounced during periods of heightened market volatility. The problem isn't a lack of information, but that information is dispersed and subject to time lags. Surf's approach is to integrate these information sources into a unified AI entry point, allowing users to obtain structured conclusions through simple descriptions, compressing the "data finding" step and moving users directly to the "judgment making" phase.

In practice, it functions more like a 24/7 research analyst. Users can track a specific token's capital flows and sentiment changes, analyze DeFi protocol TVL and yield structures, monitor unusual activity from whale addresses, or generate a project due diligence report suitable for trading decisions or communication prep in a short time. Compared to traditional tools requiring users to filter, piece together, and interpret information, Surf outputs processed results more directly, thus shortening the path from "acquiring information" to "forming a judgment."

Building on this, Surf is evolving from an "information tool" into a "workflow platform." The newly launched Surf 2.0 and Surf Studio allow users to build analytical tools and even simple Web Apps directly using natural language, deploying them for immediate use without relying on traditional development processes. Simultaneously, Surf integrates the capabilities of multiple models including OpenAI, Anthropic, and Google, and connects to dozens of data sources and on-chain interfaces. This means the generated analytical results are no longer just text, but tools usable for continuous monitoring and decision-making.

At a deeper level, it is gradually building a capability system oriented towards Agents. Through its API and Agent Stack, users can delegate specific tasks (e.g., whale address monitoring, capital flow tracking, strategy signal alerts) to be continuously executed by the AI, rather than manually querying each time. This signifies that Surf is no longer just a passive query portal but is transforming into a research system capable of long-term operation.

However, its capability boundaries are also quite clear. Surf's core remains concentrated on the information integration and analysis layer, and it hasn't truly entered the trading execution phase. For example, automatic order placement or strategy execution still needs to be handled by users themselves. This makes it more suitable as a decision-support tool rather than a system capable of completing a closed trading loop independently.

From an industry perspective, this type of product represents an early form of AI application realization. Compared to directly challenging the complex domain of trade execution, first making "understanding the market" more efficient and convenient is often easier for users to embrace. Before trading becomes fully automated, enhancing information processing efficiency remains the most direct and easily perceived value for users.

Anuma: A Sovereign Memory Vault for the AI Privacy Era

Over the past two years, AI has become a universal keyword across the global tech scene. From the model race in Silicon Valley to the pursuit of AI applications and capital narratives in New York and Hong Kong, the focus of industry discussion has been shifting rapidly. Previously, the competition mainly revolved around model capabilities—reasoning, multimodality, Agent execution ability. Almost every product update answered the same question: whose model is smarter, more accurate, and better at completing complex tasks?

But as model capabilities continue to improve, purely comparing the models themselves is becoming increasingly difficult to create lasting differentiation. Entering a new phase, how AI can remember users long-term and carry these memories across writing, research, decision-making, and daily communication is becoming a new focal point. This implies that the moat of AI is extending from model capability to memory capability. Models determine what AI can answer, while memory determines whether AI can truly understand a long-term user.

However, today's AI memory doesn't truly belong to the user. In current mainstream AI products, conversations, preferences, and usage habits are continuously recorded, gradually forming an experience that seems to 'understand' the user better over time. But these memories are typically locked within their respective platforms, controlled by the platform, unable to be freely transferred, and difficult for users to truly control.

This means that while AI is accumulating users' digital personas, the ownership and control of this data often still belongs to the platform. The longer the usage, the more memories accrue, and the higher the cost for users to switch models. What truly locks in users isn't necessarily the model itself, but those long-term accumulated memories that cannot be taken away.

This is precisely where Anuma steps in. As ZetaChain's flagship product venturing into the AI direction, Anuma's role is more than just an application portal. More accurately, the underlying infrastructure ZetaChain aims to build is a set of user-owned AI memory systems; Anuma is the user-facing AI interaction portal for this system.

In other words, ZetaChain is responsible for building the underlying memory capabilities, while Anuma is responsible for bringing this capability into everyday AI usage scenarios. What Anuma aims to do is to decouple memory from the model, allowing users to invoke, manage, and carry forward their long-term memories in practical usage for the first time.

Specifically, users can import their complete conversation history from ChatGPT, Claude, or Grok into Anuma. After local encryption, this data is stored in a self-controlled Memory Vault. More importantly, this process puts privacy protection before the data enters the system. Users aren't passively choosing to authorize after the platform has already taken control of the data; they retain control from the very beginning.

These memories no longer depend on a single platform. They can be taken away, reused, and carried across different models. They are locally encrypted, portable, not tied to a single model, and can continuously accumulate with long-term user usage.

From a practical experience perspective, Anuma is firstly a unified entry point aggregating multiple cutting-edge models. With a single subscription, users can access the latest models like GPT, Claude, and Grok without needing to switch back and forth between different platforms.

More critically, when users switch between different models, the memories already formed are not reset. Within Anuma, models like GPT and Claude function more as capability layers, while the user's own memory remains consistent. Regardless of which model is used, past conversation records, expressions, and preferences can be continued, not erased.

Anuma also offers a multi-model Council Mode, allowing users to have multiple models provide different perspectives on the same question and then compare the results. For research, writing, and complex judgments, this experience is akin to having multiple AIs participate in a discussion simultaneously, rather than relying on a single model's singular output.

Additionally, Anuma allows users to interact directly with AI via iMessage. Each Agent can be invoked like a contact and even added to group chats. Compared to having to open a specific application to start a conversation, this method is closer to everyday communication scenarios and makes the AI entry point lighter. Even in scenarios with weak network signals or inability to open the app, users can still invoke AI, and the related conversations will enter the same encrypted memory system, without interruption due to the change of entry point.

From a product perspective, Anuma is not just a multi-model portal; it is building a memory system independent of the models themselves. In the past, users' conversation records, preferences, and usage habits were often attached to specific platforms. But as AI becomes a long-term tool, these continuously accumulated memories will form the foundation for understanding the user.

This is also why ZetaChain seeks to enter the next generation of AI infrastructure, with Anuma serving as the user-facing portal. Models can be continuously upgraded and replaced, but the long-term memories users accumulate shouldn't be locked away with the platform. The future competition among AI products may not just be about who has the stronger model, but also about who can allow users to truly own, invoke, and carry forward their memories.

In the AI era, memory is becoming part of identity. And identity should belong to the user.

Nansen AI: Turning On-Chain Research and Trading into a "Conversational Operation"

When the question shifts from "who owns the data" back to "how to use the data," another category of products begins to focus on more concrete operational layers. What Nansen AI does is compress the scattered steps between on-chain research and actual trading into a single path as much as possible.

In traditional on-chain research, users often need to switch back and forth between multiple dashboards, manually querying capital flows, address behaviors, and token data, then executing operations based on their own judgment. This process itself is not complex, but the steps are tedious, with a clear gap between information and execution. Nansen AI's approach is to reconnect these two parts.

Users can ask questions directly using natural language to get information like on-chain capital flows, Smart Money movements, and token trends, without needing to query item by item. For example, asking why a certain token is rising, analyzing the profit/loss situation of a specific address, or directly parsing a transaction—the entire process can be completed within a conversation. This method essentially abstracts "research" from the operational workflow, compressing it into a continuous conversational process.

Going a step further, Nansen AI is attempting to connect information acquisition with actual operations. In some scenarios, users can directly complete on-chain interactions like transfers or Swaps through conversations, thereby compressing the previously dispersed research and execution processes into the same path. This also means Nansen AI is no longer just providing explanations but is gradually moving closer to the operational layer.

The premise for this extension comes from its long-accumulated on-chain data capabilities. Based on a large number of labeled addresses and real-time data, the system can identify fund sources, track large flows, and combine this with holding situations to provide more targeted analysis results. It is precisely because of this data foundation that it can undertake specific operations beyond just conversation.

Within this structure, Nansen AI's positioning also changes. It is no longer just an information tool, but closer to a connection point between the data input layer and the operation interface in trading decisions. However, this type of "conversational trading" is still in its early stages. The AI is more focused on lowering operational barriers and information acquisition costs, rather than replacing the user in making strategic judgments. Whether it's asset allocation or risk control, the final decision still needs to be made by the user.

Overall, Nansen AI represents another path for AI applications—extending further into the execution layer on top of the information layer. It doesn't change the logic of trading itself, but provides a lighter, more direct way to "execute a trade." Compared to purely informational tools, this ability to connect "research" and "operation" is more likely to enter real usage scenarios first.

Virtuals Protocol: Turning AI Agents into "Tradable Economies"

When AI begins to participate in operational processes, the question extends further—if these Agents are not just auxiliary tools but can independently provide services and continuously create value, can they be incorporated into a complete economic system?

This is precisely the direction of Virtuals Protocol's exploration.

In traditional AI products, Agents are often seen as tools without independent economic attributes. They can complete tasks but cannot directly participate in value distribution or easily form sustainable business models. Virtuals' idea is to transform Agents from "functional units" into "economic participants."

In this system, each Agent can be tokenized, thereby possessing capabilities for fundraising, incentives, and profit distribution. Developers are no longer just releasing an AI tool; they can build a complete economic model around a specific Agent, allowing it to generate continuous value during its use. In this way, AI is no longer a one-time delivered product but becomes an asset that can operate long-term.

Structurally, Virtuals provides a complete set of infrastructure including coordination, settlement, and issuance. Agents can collaborate with users or other Agents to complete tasks, exchanging value through on-chain mechanisms. Simultaneously, through its Launch mechanism, Agents themselves can gain liquidity support, forming paths for pricing and capital formation.

Compared to the previous projects which mainly focus on "how to use AI better," Virtuals focuses more on "how AI itself can participate in economic activity." It attempts to push AI from the tool layer into the sphere of production relations, making Agents subjects capable of independently creating value.

However, looking at the current stage, this direction is still early. On one hand, there are not many Agents with stable usage demand and revenue capabilities; actual applications within the ecosystem are still in the validation process. On the other hand, mechanisms for collaboration, pricing, and trust among Agents also need more time to establish.

From an industry perspective, Virtuals represents a longer-term path within AI + Crypto. It doesn't directly optimize the user's current experience but attempts to build a new foundational structure, giving AI more complete economic attributes in the future. This direction may not be the easiest to perceive in the short term, but once proven viable, it could change AI's role within the entire system.

Warden: Making AI Agents Usable, Distributable, and Monetizable

As the number of Agents increases, the challenge often lies not in capabilities, but in whether viable usage scenarios can be established. Compared to model capabilities or specific functions, the difficulty for most Agents isn't "can they do it," but "will anyone use them?" They are scattered across different frameworks and entry points, lacking a unified distribution channel and clear payment and collaboration methods. This is where Warden enters.

Its approach is not complicated: it builds a complete set of usable infrastructure around Agents. For users, this means being able to invoke different Agents from a unified entry point, completing operations like trading, cross-chain transfers, and queries through natural language, integrating previously disparate functions into a continuous workflow. For developers, it means they can quickly create and launch Agents, provide services directly to users, and handle billing and settlements through on-chain mechanisms.

At a deeper structural level, Warden uses a dedicated chain to manage Agent identities and invocation processes, giving each Agent an independent form of existence. It can not only charge fees but also invoke other Agents, gradually forming collaborative relationships. Simultaneously, through a distribution portal similar to an app store, Agents have the opportunity to be discovered by users, rather than just being sunk into the system after launch.

Compared to the previous AI projects,