Written by: Ice Frog
AI Agents have gained popularity in the crypto world, and it looks like they are about to explode. In the face of this immense wealth effect, they are shaped into a new revolutionary variable, hot and trending. However, the term revolutionary is highly inflated in the crypto world, scattered across various white papers and social media. For veteran investors, such phrases have become hard to ignite passion. The rise and fall of countless tokens in the crypto world prove that only a few new narratives have achieved the main upward trends in crypto markets; the vast majority are merely fleeting moments.
Thus, we must further explore a classic question in the cryptocurrency sphere: xxx, will this time be different? Everyone has different answers, but if we trace back, at least the fundamental conclusions to this question will not change much.
That is: the crypto world follows the attention economy. How far new narratives can go depends greatly on how much user attention and network effects can spread; even Bitcoin cannot escape this fundamental law.
Naturally, to analyze this issue, we must start from the source to explore the answers.
1. What is an AI Agent?
AI Agent refers to Artificial Intelligence Agent, a concept relatively familiar and commonly used by the public: AI Agent (Artificial Intelligence Agent) is an intelligent entity capable of perceiving the environment, making decisions, and executing actions. It primarily relies on LLM (large language models); in other words, it is a functional carrier of large language models. Interestingly, intuitively, it is a specific application of LLM, but in terms of terminology, it is referred to as an Agent, emphasizing the power and capability granted for autonomous selection, action, and decision-making.
In terms of definition, while AI Agents and large models have a sequential relationship, the interaction between large models and humans requires prompts to give specific tasks, leading to answers. AI Agents, however, operate on specific goals set by humans, breaking down execution steps independently, prompting themselves, and ultimately achieving the goals.
From the perspective of human collaboration with AI, the Agent model is also a relatively advanced collaborative approach, akin to the early Siri—Microsoft Copilot—AI Agent. In this collaboration, we can basically clarify that AI Agents, in essence, are a digital mapping of human thought and behavior patterns. Therefore, their construction primarily involves a question and answer interface + fully automated workflows (perception, decision-making, action) + knowledge bases (human hippocampus), with AI completing the vast majority of the work, rather than merely assisting.
From a specific technical framework perspective, former OpenAI chief security researcher Lilian Weng wrote a dedicated blog in June 2023 titled (LLM Powered Autonomous Agents), detailing as follows:
In this article, Lilian Weng proposed the foundational framework of AI Agents as LLM + Planning + Memory + Tool Usage, with the primary role of large models being akin to the reasoning and planning functions of the human brain.
Overall, AI Agents are independent computing entities, fundamentally based on the reasoning and planning functions of large models, combined with the perception of external environments, tool usage, and actions, thereby achieving the role of AI as a proxy for humans to accomplish a series of relatively complex tasks.
2. How is the progress of the AI Agent industry?
Since 2023, AI Agents have entered the industrial view, discussions and advancements regarding Agents have accelerated, and most major companies see 2025 as a key year for the commercial explosion of AI Agents; the industry is entering an accelerated development phase.
From the overall perspective of the industrial chain, the upstream of the entire industrial chain still mainly consists of computing power and hardware suppliers, data suppliers, algorithm and large model developers, etc., while the midstream primarily comprises AI Agent integrators, and the downstream targets vertical applications or general Agents for different industries.
Currently, the AI Agent industry, aside from existing upstream infrastructure, mainly focuses on the mid to downstream, especially the downstream applications, which have shown a flourishing state.
Additionally, from a downstream development perspective, both C-end and B-end have different progress; the adoption of AI Agents on the C-end can significantly improve user experience, while on the B-end, it can greatly reduce costs and increase efficiency.
From the actions of several major companies, it appears that AI Agents are gradually accelerating this second half of the year, with an overall acceleration expected next year.
This month, Google released Gemini 2.0, emphasizing that the model is primarily designed for AI Agents, and 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.
Amazon announced this month that it will open an AI laboratory in San Francisco, focusing on the implementation of AI Agents.
OpenAI has launched a series of new products over 12 consecutive days this month, with Sam Altman claiming that next year will be the year AI Agents enter the mainstream. Although OpenAI itself has not released relevant AI agents, it has introduced a series of tools that support the development of AI agents on a foundational level.
Whether from major companies or the prosperity of the industry, current AI Agents have indeed entered an acceleration phase; their entry into Crypto is just a matter of time, but all of it still belongs to the early stage.
3. Where does the integration of AI Agents and Crypto focus?
Rewinding the clock a few months, it's hard to believe that Andy Ayrey, who created the Truth of Terminal model, could have imagined that an AI Agent model, still considered experimental, would create a shocking wealth miracle in the crypto world, with its derived GOATSE concept becoming a crypto hotspot. The MEME token GOAT surged to 20 million USD in half a day, approached 300 million USD market value in four days, and surpassed 1 billion USD market value within a month, a miraculous thousandfold increase. AI Agents entered the crypto world in a very crypto way, sparking a tremendous wave.
Subsequently, Ai16Z (a venture capital fund driven by AI Agents) quickly gained popularity with the concept of 'AI-driven DAO,' backed by Marc Andreessen, the founder of the traditional A16z, and increased over tenfold within a few days; later, another AI project ACT launched on Binance, directly pushing AI Agents to a new peak, not only unprecedented in market volume but also deeply enlightening a broad range of crypto users about the wealth effect of this sector.
The first two chapters of this article have spent a considerable amount of space clarifying what AI Agents are and the industrial development of AI Agents. The important significance lies in the fact that the origin of AI does not come from the crypto world; in a broader sense, crypto is not the main battlefield for AI. However, if the broad development of AI Agents is poor, the attention on AI Agents will quickly dissipate, and the narrative of AI Agents, like many narratives, will become a fleeting moment.
However, this time, within a visibly different scope, it is indeed not the same. The main reason is:
The broader AI world has not entered a bubble-deflating phase. Whether it's NVIDIA, Microsoft, Google, or other major companies, they have all increased their capital expenditures for 2025 in their third-quarter reports this year. The top four companies alone will invest over 170 billion USD next year, with a single goal: AI.
From the perspective of AI Agent development, while there are currently no phenomenon-level products like ChatGPT igniting the market, whether from actions of major companies or industrial development, its vigorous momentum and financial bets are on a steep upward trajectory, and there is a certain probability that a market-explosive AI Agent will emerge in 2025.
From these two points, from a broad technological perspective, the topic and attention of AI and AI Agents must remain among the hottest in the market rankings. As mentioned at the beginning, attention in the crypto world is everything.
From the technical framework for implementing AI Agents, its integration with Crypto will give rise to pivotal breakthroughs similar to the birth of Ethereum smart contracts, not merely a technological enhancement but a potentially transformative economic paradigm shift because it will fundamentally change the mode of attention economy creation.
The current biggest challenge affecting mass adoption of blockchain may lie in its complex operations and entry barriers. Excluding compliance issues related to fiat currency, the complexity and difficulty of on-chain operations and wallets are at least several times that of Web2. If a natural language model of AI Agents is utilized, 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, not only simplifies operational difficulties greatly but also directly reduces the learning costs for new users by several levels.
Moreover, whether it's 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, fair, and capable of eliminating emotional influences than most people.
From a narrative logic perspective, AI + Crypto brings about: AI can make trustworthy blockchains smarter, while blockchain can make intelligent AI more trustworthy.
The characteristic of blockchain is the immutability of data, while one major flaw of AI lies in data quality. If an 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 cryptocurrency user will have a digital twin that will help manage their token assets, social interactions, and more; each project team will have several AI Agents to assist in operations ranging from asset issuance, marketing, code development, contract auditing, media operations, and even airdrop design and distribution, all facilitated by AI.
These long-term changes will alter the mode of attention creation, transforming it from larger communities and humans into AI Agents.
Of course, the above ideas are merely long-term visions. Returning to reality, current AI Agents in blockchain are still in an early, rough stage. Aside from the explosive popularity of AI MEME, AI Agents are still in a phase more of speculation than building. Currently, there isn't a truly blockchain-based AI Agent framework on the market. Even the early arrival, ELIZA, is limited to conversational interfaces and has yet to penetrate the core of the entire blockchain world.
From three aspects of AI Agents—perception, decision-making, execution—there's a need to rely on decentralized characteristics and smart contract traits to build a more systematic and foundational infrastructure, not to mention various tool platforms, data privacy, transaction security, etc. Fortunately, whether it's top-level capital like A16z focusing on AI Agents, the broad mainstream narrative around AI Agents globally, or the astonishing wealth effect brought about by the explosion of AI MEME, all have provided a solid foundation for the development of AI Agents in the crypto world.
In the attention economy of cryptocurrency, current AI Agents have capital attention, sustained enthusiasm for broad narratives, typical cases of wealth effect, and practical value for long-term construction.
Perhaps we can boldly say that this time, it really is different!
Passion!
