Author: Ice Frog Source: X, @Ice_Frog666666
The AI Agent has become a hit in the crypto world, and it is about to explode. Facing the huge wealth effect, it has been shaped into a new revolutionary variable, highly sought-after and unparalleled. However, the word "revolutionary" is a highly inflated concept in the crypto world, scattered in large and small whitepapers and social media. For old-timers, such rhetoric can hardly arouse Passion. The rise and fall of countless tokens in the crypto world in the past have proven that only a handful of new narratives have achieved the crypto bull run, and the vast majority are just fleeting.
So, we must go further to discuss a classic question in the crypto world: xxx, will it be different this time? Everyone has different answers, but if we look back, at least the fundamental conclusion of this question will not change much.
That is: the crypto world follows the attention economy, and how far the new narrative can go depends on the extent to which user attention and network effects can be disseminated, even Bitcoin cannot escape this fundamental law.
Naturally, if we want to analyze this problem, we must start from the source and explore the answer to the problem.
I. What is an AI Agent?
An AI Agent refers to an Artificial Intelligence Agent, and the currently relatively well-known and commonly used concept is: an AI Agent is an intelligent entity that can perceive the environment, make decisions, and take actions. It is mainly based on LLM (Large Language Model), in other words, it is the functional carrier of the large language model. The interesting point is that, in the intuitive concept, it is a specific application of LLM, but in the name, it is called an Agent, emphasizing the power and ability to autonomously choose, act, and make decisions.
In terms of the definition, although there is a sequential relationship between the AI Agent and the large model, the interaction between the large model and humans requires the use of prompts to give specific tasks and then get answers. The AI Agent, on the other hand, is given a specific goal by humans, and the AI Agent will decompose the steps to be executed, give itself prompts, and then achieve the goal.
In terms of human-AI collaboration, the Agent mode is also a relatively advanced form of collaboration, analogous to the early Siri - Microsoft Coplilot - AI Agent. From this collaboration, we can basically confirm that the AI Agent is essentially a digital mapping of human thinking and behavior patterns, so its structure is mainly a question-and-answer entry + a fully automated workflow (perception, decision-making, action) + a knowledge base (human hippocampus), and AI completes most of the work, rather than just an assistive relationship.
In terms of the specific technical framework, Lilian Weng, the former chief security researcher at Open AI, wrote a blog in June 2023 with the title "LLM Powered Autonomous Agents" to elaborate on this:
In this article, Lilian Weng proposes that the basic framework of the AI Agent = LLM + planning + memory + tool use, and the main role of the large model is to undertake functions similar to human brain reasoning and planning.
In general, the AI Agent is a separate computing entity, based on the reasoning and planning functions of the large model, combined with the perception of the external environment, the use of tools, and actions, to achieve the role of AI as a human agent and complete a series of relatively complex tasks.
II. How is the AI Agent industry progressing?
Since 2023, the AI Agent has entered the industry's vision, and the discussion and progress of the Agent have been accelerating. Most of the big companies see 2025 as a key year for the commercial explosion of the AI Agent, and the industry is entering a period of accelerated development.
From the overall perspective of the industry chain, the upstream is still dominated by hardware suppliers such as NVIDIA, data suppliers, and algorithm and large model developers, the midstream is mainly AI Agent integrators, and the downstream is vertical applications for different industries or general Agents.
Currently, the AI Agent industry, apart from the existing infrastructure in the upstream, is mainly concentrated in the midstream and downstream, especially the applications in the downstream have shown a blossoming state.
In addition, in terms of downstream development, both the C-end and the B-end have different progress. The use of AI Agents can greatly improve the user experience on the C-end, and on the B-end, it can greatly reduce costs and increase efficiency.
Looking at the actions of several major companies, AI Agents are accelerating their launch this year, and are expected to accelerate overall next year.
Google released Gemini 2.0 this month, not only emphasizing that the model is mainly for AI Agent service, but also released three AI Agent products: Project Astra (general), Project Mariner (browser operation), and Jules (programming).
Microsoft launched 10 AI Agents on its Dynamics 365 platform at the end of October this year.
Amazon announced this month that it will open an AI lab in San Francisco, focusing on the implementation of AI Agents.
Open AI has been launching a series of new products for 12 consecutive days this month, and Sam Altman himself has claimed that next year will be the year when AI Agents enter the mainstream, although Open AI itself has not released any related AI agents, but has launched a series of tools to support the development of AI agents on the basic framework.
Whether it's the big companies or the prosperity of the industry, the AI Agent has indeed entered an accelerated period, and its entry into the crypto world is just a matter of time, but everything is still in the early stage.
III. Where is the landing point of the combination of AI Agent and Crypto?
Turning the clock back a few months, Andy Ayrey, who created the Truth of Terminal model, probably never imagined that an experimental AI Agent model would create a shocking financial miracle in the crypto world, and the derived GOATSE concept has become a hot topic in the crypto world. The MEME token GOAT soared to $20 million in half a day, reached a market value of nearly $300 million in four days, and exceeded $1 billion in market value in a month, a miracle of a thousandfold increase. The AI Agent has entered the crypto world in a very crypto way and has sparked a huge wave of heat.
Subsequently, Ai16Z (an AI Agent-driven venture capital fund) quickly became popular with the concept of "AI-driven DAO", coupled with the support of Marc Andreessen, the founding partner of the traditional A16z, it has increased by more than 10 times in just a few days; later, another AI project ACT was launched on Binance, which directly pushed the AI Agent to a new climax, not only with unprecedented market volume, but also made the vast majority of crypto users deeply realize the wealth effect of this track.
The first two chapters of this article have spent a lot of space to explain what an AI Agent is and the industrial development of the AI Agent, the important significance of which is that the origin of AI does not come from the crypto world, and in a broader sense, crypto is not the main battlefield of AI. However, if the general development of AI Agents is not good, the attention to AI Agents will quickly dissipate, and the narrative of AI Agents will become a fleeting phenomenon like many other narratives.
However, this time, within the visible range, it is indeed different. The main reasons are:
The general AI world has not entered the bubble falsification period, whether it is NVIDIA, Microsoft, Google and other big companies, in their third quarter financial reports this year, they have unanimously increased their capital expenditures for 2025, and the top four companies alone will invest more than $170 billion in capital next year, with the only goal being AI.
From the development of AI Agents, although there is currently no phenomenon-level product like chatgpt that has ignited the market, but from the actions of the big companies and the development of the industry, the vigorous momentum and the capital betting are all rising linearly, and there is a certain probability that a market-igniting AI Agent will appear in 2025.
From these two points, from the broad technological perspective, the topic and attention of AI and AI Agents are still and must be among the top market heat, as mentioned at the beginning, attention is everything in the crypto world.
From the technical framework of the realization of AI Agents, its combination with Crypto will give birth to a key turning point breakthrough similar to the birth of Ethereum smart contracts, not only the technical superposition, but more likely a leapfrog economic paradigm shift, because it will fundamentally change the creation mode of the attention economy.
The biggest challenge currently facing blockchain's Mass Adoption may be its complex operations and high entry barrier. Aside from the compliance pain points involved in fiat currency deposits/withdrawals, the difficulty and complexity of on-chain operations and wallets are at least several times that of Web2. If an AI Agent with natural language mode is adopted, a simple instruction, whether it's wallet management, screening the best DeFi investments, cross-chain operations, or automatically executing trading plans based on external market conditions, can greatly simplify the operational difficulty and directly reduce the learning cost for new users by several levels.
Furthermore, whether it's the creator economy, market sentiment monitoring, smart contract auditing, governance voting, AI-autonomous DA, or even MEME issuance, Agents can be used to participate. Under certain predetermined conditions, they may be more serious, fair, and able to eliminate the influence of emotions than most people.
From the narrative logic, AI+Crypto brings: AI can make trustworthy blockchains smarter, and blockchains can make intelligent AI more trustworthy.
The characteristic of blockchains is the immutability of data, while a major flaw of AI is data quality. If the AI Agents built can be trained on on-chain data and leverage its computing power, they may very likely change the current incentive model.
Looking further ahead, perhaps in the not-too-distant future, every crypto user will have a digital avatar that will help manage your token assets, social interactions, etc.; every project will have several AI Agents to assist with asset issuance, marketing, code building, contract auditing, media operations, and even airdrop design and distribution, all of which can be aided by AI.
These long-term changes will transform the mode of attention creation, from larger-scale communities and humans to AI Agents.
Of course, the above ideas are just long-term speculations. Returning to reality, the current AI Agents in the blockchain space are still in a primitive stage. Apart from the explosive popularity of AI MEME, AI Agents are still in a stage where speculation outweighs building. There is currently no truly blockchain-based AI Agent framework in the market, and even the pioneering ELIZA is only at the dialogue level, not yet penetrating the core of the entire blockchain world.
From the three levels of AI Agents - perception, decision-making, and execution - all need to be rebuilt based on the decentralized features and smart contract properties to construct a more systematic and underlying infrastructure, not to mention the various tool platforms, data privacy, transaction security, etc. Encouragingly, whether it's the attention of top-tier capital like a16z, the mainstream narrative of AI Agents globally, or the astonishing wealth effect of AI MEME, they have all laid a solid foundation for the development of AI Agents in the crypto world.
In the attention economy of crypto, the current AI Agents have the backing of capital, the sustained heat of the mainstream narrative, typical cases of wealth effects, and the long-term practical value.
Perhaps we can be a little bolder and say, this time, it's really different!
Passion!
