Written by: Ice Frog
AI Agents are booming in the crypto world and are expected to explode. In the presence of huge wealth effects, they are being shaped into a new revolutionary variable, highly sought after and without equal. However, the term revolutionary is highly inflated in the crypto world. It is scattered across various white papers and social media, and for seasoned investors, similar rhetoric has become hard to inspire passion. The rise and fall of countless tokens in the crypto world prove that only a few new narratives have led to significant upward trends in crypto, while the vast majority are merely fleeting.
Thus, we must further explore 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 to this question will not change much.
That is: the crypto world follows the attention economy, and how far new narratives can go depends on how much user attention and network effects can be spread; even Bitcoin cannot escape this fundamental law.
Naturally, to analyze this issue, we must start from the source and explore the answer.
1. What is an AI Agent?
AI Agent refers to Artificial Intelligence Agents, a concept that is relatively well-known and commonly used: an AI Agent is an intelligent entity capable of perceiving the environment, making decisions, and executing actions. It is mainly based on LLM (large language models), in other words, it is the functional carrier of large language models. Interestingly, intuitively, it is a specific application of LLM, but in terms of nomenclature, it is called an Agent, emphasizing the empowerment of autonomous choice, action, and decision-making rights and abilities.
In terms of definition differentiation, while AI Agents and large models have a sequential relationship, the interaction between large models and humans requires prompts to provide specific tasks, which then yield answers. AI Agents, on the other hand, operate based on specific goals provided by humans, breaking down execution steps autonomously and generating their own prompts to achieve those goals.
From the perspective of collaboration between humans and AI, the Agent model is also a relatively advanced collaborative method, analogous to early Siri—Microsoft Copilot—AI Agent. From this collaborative perspective, we can basically clarify that AI Agents are essentially a digital mapping of human thinking and behavior patterns, thus their structure mainly consists of a Q&A entry + fully automated workflows (perception, decision-making, action) + a knowledge base (human hippocampus), where AI completes 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 blog in June 2023, specifically elaborating on the topic with the title (LLM Powered Autonomous Agents), detailed as follows:
In this article, Lilian Weng proposed the foundational framework for AI Agents as LLM + Planning + Memory + Tool Usage, where the primary role of the large model is to assume functions similar to human brain reasoning and planning.
Overall, AI Agents are independent computational entities, fundamentally based on the reasoning and planning functions of large models, combined with external environmental perception, tool usage, and actions, thereby achieving the role of AI as a human agent, completing a series of relatively complex tasks.
2. How is the progress of the AI Agent industry?
Since 2023, the visibility of AI Agents entering the industry has accelerated the discussion and progress regarding Agents. Most major companies see 2025 as a key year for the commercial explosion of AI Agents, and the industry is entering a period of accelerated development.
From the overall perspective of the industry chain, the upstream of the entire industry chain is still dominated by computing power and hardware suppliers like Nvidia, data suppliers, algorithm and large model developers as foundational implementers, while the midstream mainly consists of AI Agent integrators, and the downstream targets vertical applications or general-purpose Agents for various industries.
Currently, the AI Agent industry, aside from existing upstream infrastructure, is primarily concentrated in the mid to lower reaches, especially the downstream applications that have shown a blossoming state.
Moreover, from the perspective of downstream development, both C-end and B-end have different progress, where 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 seems that AI Agents are gradually accelerating their launch in the second half of this year, with an overall acceleration expected next year.
Google released Gemini 2.0 this month, emphasizing that the model primarily serves AI Agents, and also launched three AI Agent products: Project Astra (general-purpose), 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 establish an artificial intelligence laboratory in San Francisco, focusing on the implementation of AI Agents.
OpenAI has begun a continuous 12-day product launch this month, with Sam Altman claiming that next year will be the year AI Agents enter the mainstream. Although OpenAI itself has not released any related AI agents, it has launched a series of tools to 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, and their entry into Crypto is only a matter of time, but everything still belongs to the early stage.
3. Where is the intersection of AI Agents and Crypto?
Rewind the clock to a few months ago, and it is hard to believe that Andy Ayrey, who created the Truth of Terminal model, never expected that an AI Agent model, still in its experimental phase, could create a shocking wealth miracle in the crypto world. The derived GOATSE concept became a crypto hotspot, with the MEME token GOAT skyrocketing to $20 million within half a day, approaching $300 million in market value within four days, and exceeding $1 billion in market value within a month, a miraculous thousand-fold increase reappeared. AI Agents entered the crypto world in a very crypto manner, sparking a tremendous wave.
Subsequently, Ai16Z (AI Agent-driven venture capital) quickly gained popularity with the concept of 'AI-driven DAO,' and with the support of A16z founder Marc Andreessen, it surged more than 10 times in just a few days; after that, another AI project ACT launched on Binance, directly pushing AI Agents to a new climax, creating unprecedented market buzz and allowing a wide range of crypto users to deeply understand the wealth effects of this track.
The first two chapters of this article have expended significant space to clarify what exactly AI Agents are and the development of the AI Agent industry. 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 AI Agents do not develop well in a broad sense, attention towards AI Agents will quickly dissipate, and the narrative surrounding AI Agents will, like many narratives, become fleeting.
However, this time, within a visible range, it is indeed different. The main reason lies in:
The broader AI world has not entered a bubble denial phase; whether it's Nvidia, Microsoft, or Google, these major companies have coincidentally increased their capital expenditures for 2025 in their third-quarter reports this year, with just the top four companies planning to invest over $170 billion in capital next year, with only one goal: AI.
From the perspective of AI Agent development, although there is currently no phenomenally market-exploding product like ChatGPT, the momentum and capital investment from large companies and industry development are on a steep upward trajectory, and there is a reasonable probability that a market-exploding AI Agent will emerge in 2025.
From these two points of view, broadly speaking, the topics and attention surrounding AI and AI Agents will remain among the hottest in the market, as mentioned at the beginning, the attention in the crypto world is everything.
From the perspective of the technical framework for implementing AI Agents, its combination with Crypto will give birth to a breakthrough similar to the emergence of Ethereum smart contracts, not only a technical enhancement but also a potential leap in economic paradigm shift, as it will fundamentally change the creation model of the attention economy.
The current biggest challenge for blockchain affecting mass adoption may lie in its complex operations and entry barriers. Setting aside compliance pain points related to fiat deposits/withdrawals, the operational difficulty and complexity on-chain and with wallets are at least several times that of Web2. If a natural language model of AI Agents is adopted, a simple command for wallet management, filtering the best DeFi investments, cross-chain operations, or automatically executing trade plans based on external market conditions can greatly simplify operational difficulty, significantly reducing the learning cost for new users.
Additionally, whether it's creator economies, market sentiment monitoring, smart contract auditing, governance voting, AI-autonomous DAs, or even MEME issuance, Agents can participate under certain conditions, and they may be more serious and fairer than most people, capable of eliminating emotional influences.
From a narrative logic perspective, AI + Crypto brings: AI can make trusted blockchain smarter, while blockchain makes intelligent AI more trustworthy.
The characteristic of blockchain is the immutability of data, while a 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; every project team will have several AI Agents to assist in operations, from asset issuance, marketing, code development, contract auditing, media operations, to even airdrop design and distribution, all can be accomplished with the help of AI.
These long-term changes will alter the creation model of attention, shifting from larger communities and human beings to AI Agents.
Of course, the above ideas are merely long-term aspirations. Returning to reality, current AI Agents in blockchain are still in an early, primitive stage. 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 on the market, and even the pioneering ELIZA only exists at the dialogue level, without entering the core of the entire blockchain world.
From the three levels of AI Agents—perception, decision-making, and execution—there is a need to rely on decentralized features and smart contract characteristics to build a more systematic and foundational infrastructure, not to mention various tool platforms, data privacy, transaction security, etc. It is encouraging that the attention from top funds like A16z towards AI Agents, the mainstream narrative of AI Agents globally, and the astonishing wealth effects generated by AI MEME have provided a solid foundation for the development of AI Agents in the crypto world.
In the attention economy of crypto, current AI Agents have capital attention, ongoing narrative heat, typical cases of wealth effects, and practical value for long-term construction.
Perhaps we can be a bit bold and say, this time, it really is different!
Passion!
