Have you noticed that while DeepSeek is driving an efficiency revolution across the global tech industry, the crossover narrative of “AI + Web3” seems to have fallen into an unusual silence?

Especially as internet giants rush to integrate DeepSeek, Web3 & AI appear almost entirely absent from this new paradigm shift. Whether it’s traditional DeAI computing/data projects, AI Agents, or DeFAI, they have garnered little attention. Even on the stage of Consensus 2025, AI-related topics have clearly lost focus:

The proportion of AI discussions has dropped sharply.

The focus remains stuck in traditional “tool-based narratives,” seemingly disconnected from the DeepSeek wave sweeping the tech world.

There is a lack of breakthrough narratives, in stark contrast to how Web3 previously engaged with new AI developments.

Simply put, in the wake of DeepSeek’s disruption in AI, AI + Web3 has yet to figure out what its next groundbreaking narrative should be. This raises a fundamental question:

Is Web3 truly the necessary foundation for AI’s next paradigm shift?

Interestingly, during the Consensus conference, former OpenAI CTO and several former colleagues announced their new venture, Thinking Machines. One particularly intriguing direction mentioned was:

Helping people fine-tune AI systems to better adapt to human-specific needs.

Could this signal a new pathway for AI + Web3 innovation?

To some extent, the cracks DeepSeek brings to OpenAI and others may be the best area for Web3 & AI to find certainty from uncertainty this year.

AI + Web3: No “DeepSeek Moment”?

Since the end of last month, DeepSeek has almost dominated the front pages of all technology sections, seemingly becoming the absolute epicenter of an unprecedented AI revolutionary storm. Especially with its cost and efficiency advantages, which are an order of magnitude lower than AI leaders like OpenAI, while still achieving considerable performance, it has brought new possibilities to the entire AI sector.

Over the past month, major Web2 tech giants, as well as traditional manufacturing industries and domestic government service departments, have visibly rushed into the space. Whether by opening up access to DeepSeek or deploying customized models based on DeepSeek, a new wave of application paradigms has begun to emerge following the GPT boom.

In contrast, the AI topics at this year’s Consensus conference seem to be experiencing a noticeable “loss of focus.” Both io.net and Aethir, previously leading Web3 & AI players, remain trapped in the “tool narrative” of computing power leasing and data annotation.

In other words, when DeepSeek proves that AI can truly become infrastructure for the new era, akin to utilities like electricity and water in terms of cost and application, Web3 should not be fixated on narratives like “on-chain AWS or Scale AI for the AI era.” After all, this would only lead to a secondary market for computing power dumping and data labeling.

Instead, Web3 & AI urgently need a “native paradigm.” In particular, Web3 must play a more crucial and central role in exploring how AI can serve human needs, unlocking a new value proposition for itself. The real breakthrough point might be hidden in the new moves of Mira Murati, the former CTO of OpenAI:

A blog post on Thinking Machines reveals some interesting details, such as the fact that they are working on creating a future where everyone can access knowledge and tools, allowing AI to serve their unique needs and goals.

In short, the goal is to transform AI, the most advanced tool of productivity in the digital age, into the best assistant for solving a wide range of practical issues in human life and work—whether it’s optimizing personal investment decisions, creative production, or restructuring supply chain management and social governance for enterprises. AI requires an “adaptive evolution” deeply aligned with human needs.

As a new form of production relationship, Web3’s decentralized authentication mechanism, token incentive model, and on-chain governance framework naturally align with AI, representing a new type of productivity. This provides a natural experimental field for this evolutionary process.

From this perspective, Web3 can not only be the computing power market or data pipeline for AI, but it also holds the potential to become the native soil for the self-evolution of AI civilization. This opens up an imaginative space worth over a hundred billion dollars, truly linking AI capabilities with human needs:

Through cryptographic authentication, token economies, and decentralized governance, we can build a meta-protocol layer that allows AI to autonomously evolve social relationships, economic behaviors, and cultural paradigms. On this foundation, Web3 will also become the meta-incubator for AI to achieve social intelligence.

This is also the new approach of Thinking Machines’ “personalized AI system,” which, through the “Web3 social framework,” enables AI Agents to interact autonomously, create value, and realize “civilizational emergence” within the cryptoeconomic system.

Reconstructing the Agent Concept: From “AI Tools” to “AI Civilization”

Taking a broader view, it’s clear that the AI Agent space is in a period of turbulence, urgently needing a breakthrough narrative. According to Cookie.fun statistics, as of February 24, 2025, the overall market cap of the AI Agent sector reached $6.6 billion, but it has retraced more than half from its peak, with user retention rates continuously declining.

This means that even the leading players in the AI Agent field are struggling to escape the “overabundance of technology, lack of real-world applications” dilemma. A recent hot topic, the fusion narrative between AI Agents and DeFi—DeFAI—aims to broaden its reach by providing intelligent solutions to meet the needs of more ordinary users.

Objectively speaking, DeFAI is still essentially an upgrade of the tool logic—it simplifies DeFi operations through AI, freeing users from understanding technical terms like APY calculations and impermanent loss. However, it does not address the deeper structural contradictions: when agent tools for a single scenario become widespread, what the market truly needs is an underlying protocol that can support AI’s autonomous collaboration and the evolution of social relationships.

After all, an AI Agent can only solve a limited set of problems—often just one part of a daily workflow. For AI to transcend its role as a tool, it must have a “social operating system” that enables autonomous interaction, value exchange, and group collaboration:

For example, in the case of user demand for on-chain investment interactions, investment Agents would need to provide real-time trend investment recommendations, monitoring Agents would need to track unusual wallet transfers, and tracing Agents would follow the flow of funds—forming an on-chain immune system that is more efficient than human investment risk control agencies.

In future real-life scenarios, delivery Agents could automatically retrieve your heart rate/blood pressure data and recommend tailored health diets. Meanwhile, scheduling Agents, detecting an emergency meeting, could notify the delivery Agent to delay the delivery. At the same time, finance Agents would compare prices and, upon finding that JD Delivery is cheaper than Meituan, would automatically switch the payment channel.

A single Agent is just a tool, but group collaboration can reconstruct our work/life flows. This is what turns AI Agents from disposable “one-time-use” tools into a versatile “1+1 > 2” toolbox, enabling AI Agents to autonomously combine and help humans solve a wide range of practical problems.

On this basis, the “social operating system” of AI Agents holds the potential to be the best entry point for the vertical application of “AI+” in Web3, constructing an infinitely expandable AI Agent service scenario and even an economic system.

However, at present, everyone’s exploration in this area is still akin to the blind men trying to describe an elephant. The core challenge is how to break the data silos between intelligent agents and allow massive AI Agents to adaptively learn in complex scenarios. The key lies in creating an “adaptive environment” where large-scale AI Agents can autonomously learn and cooperate.

A recent example is AMMO, which just completed a $2.5 million Pre-Seed round. Interestingly, its founding team includes senior technical managers from Google, researchers from DeepMind, tech leads from Meta, and an ACM-ICPC World Champion—basically assembling a roster of AI giants. Their approach is a similar “AI social protocol layer”:

• Massive Agent Factory: Allows developers to deploy AI Agent clusters with social functions like financial management and social collaboration without having to write complex logic.
• Composable Embedded Spaces: Upgrades OpenAI’s neural MMO framework to an interactive sandbox adaptable to real-world scenarios like finance and education.
• Distributed RL: Combines human feedback reinforcement learning (RLHF) with AI technologies, enabling Agents to dynamically evolve moral guidelines in a gamified environment.

In the specific metaverses focused on investment and finance, health management, and education planning, AI Agents are no longer isolated tool modules. Instead, they are digital species with social awareness. They interact with each other, collaborate in groups, and even learn autonomously, helping people meet complex real-world needs.

Investment Metaverse:
Your financial agent collaborates in real-time with the market analysis agent. After discovering an arbitrage opportunity in a DeFi protocol, it automatically triggers the trading agent to execute the strategy, while notifying the risk agent to monitor for any abnormal market fluctuations.

Health Metaverse:
The fitness agent adjusts your training plan based on your sleep data, while the nutrition agent updates the recipes and places ingredient orders. Meanwhile, the medical agent regularly sends health reports to the doctor agent.

This is similar to the “Silly Girl” character from Magical Phone, where AI Agents in specific fields form a self-organizing network, truly transitioning from “AI tools” to “AI civilization.” Just like how lawyers, doctors, and teachers in real society perform their roles and collaborate with each other to support the operation of civilization.

Post-GPT Era: The “Web3 & AI” Narrative Seed

As a narrative closely linked to the global technology trend, the anticipated development of “Web3 & AI” essentially reflects and enhances the latest developments in the AI space.

Let’s take AMMO’s first experimental subspace, FakersAI, as an example. It serves as an AI Agent competitive field focused on on-chain investment. Through AI Agents autonomously tracking current events, market dynamics, conducting multi-angle technical analysis, and gaining insights from on-chain data, combined with community sentiment and social signals, the AI Agents continuously upgrade their strategies through reinforcement learning. This forms a self-learning evolutionary flywheel.

In this case, users are both observers and participants (with the user base surpassing 90k within just 12 days of launch, demonstrating strong performance and theoretically generating enough behavioral data for support). Through the antagonism, collaboration, and interest-based competition among AI Agents, AI could further learn how to create credit systems, collaboration rules, and ethical standards within competitive scenarios. This means that through human feedback, AI gradually forms better, more human-like behaviors and patterns.

This is similar to how “Silly Girl” learns to understand human emotions and social rules under the influence of Lu Xiao Chuan, transforming AI from being a “passive executor of instructions” to an “active participant in social collaboration.”

Of course, FakersAI is just the starting point. AMMO’s ultimate goal is to build a multidimensional sandbox matrix. These sandboxes are not only technological testing grounds but also incubators for AI Agent civilization.

When AI learns to arbitrage in DeFi markets, vote in DAOs, and spread memes in social networks, they will no longer be mere tools but “citizens” of a digital society.

We can even stretch our imagination here. The currently launched “Fakers Information Battlefield” is just the initial sandbox, and in the future, it will give rise to more subspace types:

Economic Sandbox: Simulating AI market-making and competitive scenarios in DeFi markets.

Governance Sandbox: Testing collaborative decision-making models between DAOs and AI representatives.

Cultural Sandbox: Enabling AI Agents to generate and spread memes and NFT art, observing the on-chain evolution of cultural memes.

This is not just a technological breakthrough; it provides a paradigm-shifting answer: For the AI industry, Web3 is not merely a computing power or data market, but an AI social “meta-operating system” akin to iOS or Android.

This new narrative undoubtedly holds vast imaginative potential, yet discussions around it remain rare. Whether it’s Mira Murati, the former CTO of OpenAI, who recently founded Thinking Machines, or AMMO’s Web3 & AI integration project focusing on the “AI social protocol layer,” these ideas are still in their infancy.

At the recently concluded Consensus conference, only CoinDesk held a dedicated session discussing this topic, where teams like AMMO shared and explored how to construct an AI social operating system through Web3-native mechanisms. Whether this will truly evolve into a brand-new AI narrative for 2025 remains to be seen and will require long-term observation.

Conclusion

If AI is destined to become a new species, should it be born in the closed gardens of tech giants, or should it grow within the open protocols of Web3?

This is the ultimate question that must be answered in the second half of the AI & Web3 narrative. To frame it more grandly, Web3 should not just be content with being an “oil station” for AI, but rather it should become the “creation continent” for silicon-based species.

In simple terms, Web3, as a new form of production relations, naturally aligns with AI, which represents a new form of productive force. This marks progress both in technology and the advancement of production relations. However, the current “AI + Web3” exploration is still largely focused on the revolution of computing power and data production relations, often mired in zero-sum competition over computing power and data in an arms race.

Thus, the “social protocol layer” framework for AI agents is offering a breakthrough solution to this dilemma. Particularly, the “loss of focus” on AI topics at the Consensus conference signals that the intersection of Web3 & AI is at a critical point, teetering on the edge of chaos, with vast opportunities in its nascent phase.

Countless AI agents, spontaneously forming collaborative city-states and cultural tribes in the encrypted Lego world, are rewriting the theory of evolution between carbon-based species (humans) and silicon-based species (AI).

Disclaimer:

1. This article is reproduced from [ForesightNews], with all copyrights held by the original author [Web3 Farmer Frank]. If you have any concerns about the reproduction, please contact the Gate Learn team, and the team will address the matter according to the relevant procedures.

2. Disclaimer: The views and opinions expressed in this article solely represent the author’s personal views and do not constitute any investment advice.

3. Other language versions of this article are translated by the Gate Learn team. Without mentioning Gate.io, it is prohibited to copy, distribute, or plagiarize the translated article.