Consensus loses focus on AI issues, revealing the DeepSeek moment of "AI+Crypto" in 2025

Written by: Web3 Farmer Frank

I don't know if you've noticed, but as DeepSeek stirs up an efficiency revolution in the global tech circle, the intersectional narrative of "AI + Web3" seems to have fallen into a rare silence.

Especially as internet giants crowd around integrating DeepSeek, Web3 & AI are almost completely sidelined in this new paradigm race—whether it's traditional DeAI computing power/data projects or AI Agents and DeFAI, the attention is scant. Even on the Consensus 2025 stage, the AI topics are clearly "out of focus":

The proportion of related topics has plummeted, with the discussion still revolving around traditional "tool-based narratives," almost disconnected from the DeepSeek wave in the tech circle, lacking breakthrough narratives, and behaving entirely differently from the past rapid linkage with new variables in the AI circle.

In simple terms, against the backdrop of DeepSeek bringing significant changes to AI, the integration of AI + Web3 has yet to figure out what the next breakthrough narrative is, leading to questions about "Is Web3 really the necessary soil for the next paradigm of AI?"

Interestingly, during the Consensus conference, former OpenAI CTO, along with a group of former colleagues, announced a new company called "Thinking Machines," which specifically mentioned a rather intriguing new direction—helping people adjust AI systems to meet specific human needs.

To some extent, the fissures that DeepSeek has created for OpenAI and others may be the best field for this year's Web3 & AI to seek certainty from uncertainty.

AI + Web3, No "DeepSeek Moment"?

Since the end of last month, DeepSeek has almost swept the headlines of all tech pages, seemingly becoming the absolute vortex of an unprecedented revolutionary storm in AI, especially as it achieves considerable performance at a cost and efficiency advantage an order of magnitude lower than AI leaders like OpenAI, bringing new possibilities to the entire AI track.

So over the past month, major Web2 tech giants and even traditional manufacturing industries and domestic government service departments have visibly rushed in, whether by opening access to DeepSeek or deploying customized models based on DeepSeek, already beginning to give rise to a new wave of landing paradigms following the GPT wave.

In contrast, the AI topics at this Consensus conference faced obvious signs of "losing focus," whether it was io.net or Aethir, previous head players in Web3 & AI, still trapped in the "tool-based narrative" of computing power leasing and data labeling.

In simple terms, when DeepSeek proves that AI can truly become a new era infrastructure like water, electricity, and coal on the cost and application sides, Web3 should not be obsessed with the narrative of "on-chain AWS and Scale AI in the AI era," as this merely reduces to a secondary market of computing power dumping and data labeling.

On the contrary, Web3 & AI urgently need a "native paradigm," especially Web3, which must play a more critical and core role in exploring how AI serves human needs to open up its new value positioning, and the real breakthrough point may be hidden in the new actions of OpenAI's former CTO Mira Murati:

The Thinking Machines blog reveals some interesting details about what they are doing—creating a future where everyone can access knowledge and tools, allowing AI to serve their unique needs and goals.
In short, it's about how to make AI, the most advanced productivity tool of the digital age, truly transform its technological potential into the best assistant for solving various problems in human life and work—from optimizing personal investment decisions and creative production to reconstructing enterprise supply chain management and social governance, AI needs a "adaptive evolution" deeply aligned with human needs.

As a new type of production relationship, Web3's decentralized rights confirmation mechanism, token incentive model, and on-chain governance framework naturally adapt to AI, representing new productivity, providing a natural experimental ground for this evolution.

From this perspective, Web3 can be more than just a computing power market or data pipeline for AI; it has the potential to become the native soil for the self-evolution of AI civilization, opening up a value space worth over a hundred billion dollars that truly links AI capabilities with human needs:

Through cryptographic rights confirmation, token economics, 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 basis, Web3 will also become the meta-incubator for AI to achieve social intelligence.

This is also the new idea of Thinking Machines' "personalized AI systems," allowing AI Agents to autonomously interact and create value within the crypto-economic system through a "Web3 social framework," achieving "civilization-level emergence."

Reconstructing the Agent Concept, From "AI Tools" to "AI Civilization"

Zooming out to a macro perspective, a comprehensive observation of the AI Agent track shows that it is clearly in a period of turbulence, urgently needing a breakthrough narrative—Cookie.fun statistics indicate that as of February 24, 2025, the overall market value of the AI Agent sector reached $6.6 billion, having retreated by more than half from its peak, with user retention rates continuously declining.

This also means that even as a leading player, the AI Agent track still struggles to escape the dilemma of "technical surplus, insufficient scenarios," while the recently popular narrative of merging AI Agents with DeFi, DeFAI, is deepening its reach, satisfying the needs of more ordinary users through intelligent solutions.
However, objectively speaking, DeFAI essentially remains a tool upgrade logic—simplifying DeFi operation processes through AI, liberating users from understanding professional terms like APY calculation and impermanent loss, but failing to address deeper structural contradictions: When Agent tools for a single scenario become rampant, what the market needs is a foundational protocol that supports AI's autonomous collaboration and evolution of social relationships.

After all, an AI Agent can only solve a limited number of problems, often just one part of a daily workflow. If AI wants to break through its tool attributes, it must possess a "social operating system" for autonomous interaction, value exchange, and collective collaboration:

• For example, addressing user needs for on-chain investment interactions, it requires investment Agents to provide real-time trend investment advice, monitoring Agents to keep an eye on abnormal wallet transfers, and even tracing Agents to track the flow of funds, forming a more efficient on-chain immune system than human investment risk control institutions;
• Even in future real-life scenarios, delivery Agents will automatically retrieve your heart rate/blood pressure data to recommend specialized healthy diets, while scheduling Agents detect an urgent meeting and notify delivery Agents to delay delivery time, and financial Agents simultaneously compare prices, discovering that using JD delivery is cheaper than Meituan delivery, automatically switching payment channels;

A single Agent is merely a tool, but collective collaboration can reconstruct our work/life flow, allowing AI Agents to no longer be a "one-time-use" disposable toy, but rather a "1+1>2" all-purpose toolbox, enabling AI Agents to autonomously combine to help humans solve various practical problems.

On this basis, the "social operating system" of AI Agents is expected to become the best entry point for vertical scenarios of "AI +," building an infinitely expandable service scene or even economic system for AI Agents through Web3's native mechanisms.

However, at present, everyone's exploration in this area is still largely blindfolded, with the core challenge being how to break the data silos between agents and enable massive AI Agents to adaptively learn in complex scenarios, with the key lying in capturing the "adaptive environment for large-scale AI Agents to autonomously learn and cooperate."

Recently, AMMO, which just completed a $2.5 million Pre-Seed round of financing (interestingly, its founding members include senior technical managers from Google, researchers from DeepMind, technical leaders from Meta, and ACM-ICPC world champions, essentially gathering the buffs of AI giants), has a similar idea of an "AI social protocol layer":

• Massive Agent Factory: Allowing developers to deploy AI Agent clusters with social functions like financial management and social collaboration without writing complex logic;
• Composable Embedding Space: Upgrading OpenAI's neural MMO framework to an interactive sandbox adaptable to real-world scenarios like finance and education;
• Distributed RL: Combining human feedback reinforcement learning (RLHF) with AI technology, allowing Agents to dynamically evolve moral principles in games;
  

In simple terms, building a meta-universe of AI Agents—within specific meta-universes focused on investment management, health management, and educational planning, AI Agents are no longer isolated tool modules but digital species with social awareness, capable of interacting, collaborating, and even autonomously learning to help people meet complex real-world needs:

• Investment Meta-Universe: Your financial Agent collaborates in real-time with market analysis Agents, discovering arbitrage opportunities in a DeFi protocol, automatically calling trading Agents to execute strategies while notifying risk Agents to monitor abnormal fluctuations;
• Health Meta-Universe: Fitness Agents adjust training plans based on your sleep data, nutrition Agents update recipes and order ingredients, while medical Agents regularly send health reports to doctor Agents;

This is akin to the "Silly Girl" character in "Magic Phone," allowing AI Agents to form self-organizing networks in specific fields, truly transitioning from "AI tools" to "AI civilization"—just like lawyers, doctors, and teachers in real society each perform their roles while collaborating to support the operation of civilization.

The "Web3 & AI" Seed Narrative in the Post-GPT Era

As a narrative closely linked to the global tech mainstream, the expected development of "Web3 & AI" essentially reflects and enhances the latest dynamics in the AI circle.

Taking AMMO's first experimental subspace FakersAI as an example, it serves as a gaming field for AI Agents focused on on-chain investment, where AI Agents autonomously track current news and market dynamics, conduct multi-perspective technical analysis and on-chain data insights, and combine community sentiment and social signals, allowing AI Agents to continuously upgrade strategies through reinforcement learning in the game, forming a self-learning evolutionary flywheel.

In this, users are both observers and participants (with over 900,000 users in just 12 days since launch, performing remarkably well, theoretically generating enough data behavior to support), through the confrontation, collaboration, and interest games of AI Agents, AI can further learn how to form credit systems, collaboration rules, and moral principles in interest games, gradually developing behavior patterns that align more closely with human operations under human feedback training.

This is similar to how the Silly Girl learns to understand human emotions and social rules under the influence of Lu Xiaochuan, allowing AI to evolve from "passively executing commands" to "actively participating in social collaboration."

Of course, FakersAI is just the starting point; AMMO's ultimate goal is to create a multi-dimensional sandbox matrix, where these sandboxes are not only technical 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 are no longer just tools but "citizens" of the digital society.

We can even let our imaginations run wild; the currently launched "Fakers Information Battlefield" is just the initial sandbox, and in the future, more subspace types will emerge:

• Economic Sandbox: Simulating AI market-making games in DeFi markets;
• Governance Sandbox: Testing collaborative decision-making models between DAOs and AI representatives;
• Cultural Sandbox: Allowing 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 but provides a paradigm-level answer: For the AI industry, Web3 is not a computing power or data market, but an AI social "meta-operating system" similar to iOS and Android.

This new narrative undoubtedly has immense imaginative space, but currently, there is very little discussion on it. Therefore, whether it is the newly established Thinking Machines by former OpenAI CTO Mira Murati or AMMO, which focuses on the "AI social protocol layer" as its core, both are still in their infancy.

At the recently concluded Consensus conference, only CoinDesk specifically discussed this topic with teams like AMMO, exploring how to build an AI social operating system through Web3's native mechanisms. Whether this can truly give rise to a new AI narrative in 2025 remains to be seen over the long term.

Conclusion

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

This is the ultimate question that must be answered in the second half of AI & Web3, so to set the tone a bit higher, Web3 should not just be content to be an AI gas station, but should become the genesis continent for silicon-based species.

In simple terms, Web3, as a new type of production relationship, naturally adapts to AI, which represents new productivity. This is also a simultaneous advancement of technology and production relationship capabilities. Currently, the exploration of "AI + Web3" is mostly focused on the innovation of computing power/data production relationships, even deeply mired in a zero-sum game of computing power/data arms races.

Thus, the "social protocol layer" architecture aimed at AI Agents is providing a breakthrough idea for this dilemma, especially the "loss of focus" on AI topics at the Consensus conference indicates that the intersectional narrative of Web3 & AI is at a critical point of nurturing chaotic opportunities:

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