Written by: IceFrog
AI Agent is very popular in the crypto world, and it is about to explode. In the face of the huge wealth effect, it has been shaped into a new revolutionary variable, which is very popular and has the limelight. However, the word "revolutionary" is a highly inflationary concept in the crypto world. It is scattered in white papers and social media of all sizes. For old investors, such words are difficult to make people passionate. The rise and fall of countless tokens in the crypto world in the past have proved that only a handful of new narratives have achieved the main upward trend of crypto, and most of them are just short-lived.
Thus, we must further explore a classic question in the crypto world: xxx, will this time be different? Everyone has different answers, but if we look back at the past, at least the fundamental conclusions of this question will not change significantly.
That is: the crypto world follows the attention economy. How far the new narrative can go depends on how much user attention and network effects can be spread. Even Bitcoin cannot escape this fundamental rule.
Naturally, to analyze this issue, we must start from the source and explore the answers.
1. What is an AI Agent?
AI Agent refers to artificial intelligence agents. The concept that is relatively well-known and commonly used is: AI Agent is an intelligent entity that can perceive the environment, make decisions, and execute actions. It is primarily based on LLM (large language models); in other words, it is the functional carrier of large language models. The interesting point is that, conceptually, it is a specific application of LLM, but in terms of naming, it is called Agent, emphasizing the power and ability granted to it for autonomous choice, action, and decision-making.
In terms of distinguishing definitions, while there is a chronological relationship between AI Agents and large models, the interaction between large models and humans requires specific tasks to be given through prompts, leading to answers. AI Agents, on the other hand, have specific goals set by humans. AI Agents will autonomously break down the execution steps based on your goals and prompt themselves to achieve the completion of those goals.
In terms of collaboration between humans and AI, the Agent model is also a relatively advanced collaborative method, comparable to early Siri—Microsoft Copilot—AI Agents. From this collaboration perspective, we can basically conclude that AI Agents are essentially a digital mapping of human thinking and behavior patterns. Therefore, their construction mainly consists of a question-and-answer entry + fully automated workflow (perception, decision-making, action) + knowledge base (human hippocampus). AI completes the vast majority of the work, rather than just assisting.
From a specific technical framework perspective, Lilian Weng, former Chief Safety Researcher at OpenAI, wrote a blog in June 2023 specifically elaborating on this, titled (LLM Powered Autonomous Agents), as follows:
In this article, the foundational framework for AI Agents proposed by Lilian Weng is = LLM + Planning + Memory + Tool Usage. The primary role of large models is to take on functions similar to human brain reasoning and planning.
Overall, AI Agents are standalone computing entities, fundamentally based on reasoning and planning functions of large models, combined with external environmental perception, tool usage, and actions, thus achieving the role of AI as a human agent to complete a series of relatively complex tasks.
2. How is the AI Agent industry progressing?
From 2023, AI Agents have entered the industry's view, discussions and advancements regarding Agents have accelerated. Most major companies regard 2025 as a key year for the commercial explosion of AI Agents, and the industry is entering a rapid development phase.
From the overall perspective of the industry chain, the upstream of the entire industry chain is still mainly composed of computing power and hardware suppliers like Nvidia, data suppliers, algorithm and large model development providers, while the midstream mainly consists of AI Agent integrators, and the downstream involves vertical applications or general Agents for different industries.
Currently, the AI Agent industry, aside from the existing infrastructure upstream, is mainly concentrated in the mid to lower reaches, especially the downstream applications, which are exhibiting a flourishing state.
In addition, from the perspective of downstream development, both the C-end and B-end have different progress. The C-end can significantly enhance user experience by adopting AI Agents, while the B-end can greatly reduce costs and increase efficiency.
From the actions of several major companies, AI Agents have gradually begun to accelerate this year in the second half, with hopes for overall acceleration next year.
This month, Google released Gemini 2.0, emphasizing that the model is primarily for the service of AI Agents. It also launched three AI Agent products: Project Astra (general), Project Mariner (browser operations), and Jules (programming).
At the end of October this year, Microsoft launched 10 AI Agents on its Dynamics 365 platform.
This month, Amazon announced that it will open an artificial intelligence laboratory in San Francisco, focusing on the realization of AI Agents.
OpenAI began a series of new product releases over a continuous 12 days this month. Sam Altman himself stated that next year will be the year AI Agents enter the mainstream. Even though OpenAI has not released related AI agents, it has launched a series of tools to support the development of AI agents based on foundational frameworks.
Whether from major companies or from the prosperity of the industry, the current AI Agents have indeed entered an acceleration phase. Their entry into Crypto is only a matter of time, but everything still belongs to the early stage.
3. Where does the integration of AI Agents and Crypto focus?
Rewinding the clock a few months, Andy Ayrey, who created the Truth of Terminal model, probably never expected that an AI Agent model, which was still an experimental product, would create a shocking wealth miracle in the crypto world. The derived GOATSE concept became a crypto hotspot, and the MEME token GOAT soared to $20 million within half a day, approaching a market value of $300 million within four days, and exceeding $1 billion in market value within a month, a miraculous thousandfold increase reappearing. AI Agents entered the crypto world in a very crypto way, creating an astonishing wave.
Subsequently, Ai16Z (a venture capital fund driven by AI Agents) quickly gained popularity with the concept of 'AI-driven DAO,' supported by Marc Andreessen, the founder of the traditional A16z. Within days, its value increased more than tenfold. Following this, another AI project, ACT, launched on Binance, further propelled AI Agents to a new peak. Not only did it generate unprecedented market buzz, but it also helped a large number of crypto users profoundly understand the wealth effects of this sector.
In the first two chapters of this article, a considerable amount of space was devoted to explaining what AI Agents are and the industrial development of AI Agents. The important significance lies in the fact that the source of AI does not come from the crypto world. In a broader sense, crypto is not the main battlefield for AI. However, if the development of AI Agents is not good in a broad sense, the attention to AI Agents will quickly dissipate, and the narrative surrounding AI Agents will, like many narratives, become ephemeral.
However, this time, within a visible range, it really is different. The main reason is:
The broader AI world has not entered a bubble-popping phase. Whether it is Nvidia, Microsoft, or Google, all major companies have increased their capital expenditures for 2025 in their third-quarter reports this year, coincidentally. The top four companies alone will invest over $170 billion next year, with one goal: AI.
From the development of AI Agents, although there are currently no market-exploding products like ChatGPT, both the actions of major companies and the industry's development show a steep upward trend in momentum and funding. There is a certain probability that a market-exploding AI Agent will emerge in 2025.
From these two points, from a broad technological perspective, the topics and attention on AI and AI Agents will inevitably remain among the hottest in the market. As mentioned at the beginning, attention in the crypto world is everything.
From the technical framework for implementing AI Agents, their combination with Crypto will lead to a breakthrough similar to the birth of Ethereum smart contracts. It is not just a technical enhancement, but a potential leap in economic paradigm changes, as it will fundamentally change the creation model of attention economy.
The current biggest challenge affecting mass adoption of blockchain may be its complex operations and entry barriers. Apart from compliance pain points related to fiat currency such as deposits and withdrawals, the operational difficulty and complexity on-chain and with wallets are at least several times that of Web2. If we use the natural language model of AI Agents, a simple command—whether for wallet management, filtering the best DeFi investments, cross-chain operations, or automatically executing trading plans based on external market conditions—can not only greatly simplify operational difficulty but also directly reduce the learning costs for new users by several levels.
Moreover, whether it is creator economy, market sentiment monitoring, smart contract auditing, governance voting, AI autonomous DAOs, or even MEME issuance, Agents can participate. Under certain conditions, they may be more serious and fair than most people and can better eliminate the influence of emotions.
From the narrative logic, AI + Crypto brings: AI can make reliable blockchains smarter, and blockchain can make intelligent AI more trustworthy.
The characteristic of blockchain is the immutability of data, while a significant drawback of AI lies in data quality. If the constructed AI Agent can train on-chain data and utilize its computing power, it may very well change the current incentive model.
Looking further ahead, perhaps in the not-so-distant future, every crypto user will have a digital avatar that will help manage their token assets, social interactions, etc. Each project party will have several AI Agents to assist with operations, from asset issuance, marketing, code construction, contract auditing, media operations, to even airdrop design and distribution can all be accomplished with the help of AI.
These long-term changes will alter the creation model of attention, shifting from larger-scale communities and humanity to AI Agents.
Of course, the above ideas are merely long-term speculations. Returning to reality, the current AI Agents in blockchain are still in an early, wild phase. Apart from the explosive popularity of AI MEME, AI Agents still belong to a stage where speculation outweighs building. Currently, there is no truly blockchain-based AI Agent framework in the market; even the pioneering ELIZA is still limited to conversational aspects and has not yet entered the core of the entire blockchain world.
From the three aspects of AI Agents—perception, decision-making, and execution—there is a need to rebuild a more systematic and fundamental infrastructure relying on decentralized features and smart contract characteristics, not to mention various tool platforms, data privacy, transaction security, etc. Fortunately, whether it is the attention that top-tier funds like A16z pay to AI Agents, the mainstream narrative surrounding AI Agents globally, or the astonishing wealth effects erupted from AI MEME, all provide a solid foundation for the development of AI Agents in the crypto world.
In the attention economy of crypto, the current AI Agents have capital attention, sustained heat from broad narratives, typical cases of wealth effects, and practical value for long-term construction.
Perhaps, we can be a bit bolder and say that this time, it really is different!
Passion!
