When the market is quiet, I would like to share some thoughts on the entire Web3 + AI Agent track.

Author: CaptainZ

When the market is quiet, I would like to share some thoughts on the entire Web3 + AI Agent track.

Since AI burst onto the scene a couple of years ago, it has already been confirmed as a technological revolution comparable to the invention of the internet. However, initially, everyone's focus was on LLM (large language models), and there was uncertainty about how AI would actually impact the real world.

If we compare LLM models to the brain, it is clear that AI only possesses analytical and reasoning capabilities, making it difficult to interact with the real world. Thus, we naturally think of adding various sensors and functional modules to LLMs, and those related to the internet are even simpler, as they can be connected via API interfaces to achieve various functions. Hence, AI officially evolved into AI Agents.

Large language models are certainly important; they determine the intelligence level of AI Agents. Currently, the top models, such as OpenAI's O3 and DeepSeek's R1, surpass the capabilities of many human PhD students in various aspects.

In 2024, various industries began exploring AI Agent + different sectors. For example:

• AI Agent + E-commerce (AI directly assists in product selection, generates copy and images)

• AI Agent + Internet (AI programming)

• AI Agent + Finance (AI analyzes data in real-time, provides intelligent investment advice)

• AI Agent + Supply Chain (AI optimizes supply chain management, monitors production lines)

• AI Agent + Education (AI provides real-time tutoring, offers personalized education)

Web3 has developed into a massive industry, so naturally, we also see

AI Agent + Web3

However, Web3 itself is divided into many subfields, so AI Agents will also have different approaches when integrating with industries. For example:

AI Agent + Meme

AI Agents directly replace humans in issuing meme coins, represented by Goat Fartcoin.

AI Agent + DeFi

AI Agents directly replace humans in analyzing market data (AIXBT), managing funds (DegenAI), and performing DeFi operations (Griffain).

AI Agent + Gaming

AI Agents directly participate in game commentary (Luna), play games (Freya), analyze sports betting (Dwain), and provide AI game frameworks (Digimon/Dreams).

AI Agent + Social

ACT initially studied how agents socialize with each other, later evolving into memes.

Of course, there are also AI Agent + DePin, etc., which I won't elaborate on here.

AI Agents have already been implemented in Web2 technologies; they are not just empty slogans. In the Web3 industry, it merely adds a token issuance process for speculation.

People often say that Web3's AI Agent technology is much worse than Web2, but AI Agent + Web3 is essentially about transplanting Web2 technology into Web3 for application; finding the intersection of industries is the key.

I personally categorize Web3 AI Agents into the following aspects: AI Agent Infra (infrastructure) and AI Agent (agents for specific application scenarios).

1. AI Agent Infra (Infrastructure)

1. Fine-tuned Large Language Models

LLMs are equivalent to the brain of AI Agents. An unrefined LLM is like a graduate who has not specialized in any field. A fine-tuned LLM based on industry data is like a university graduate who has chosen a major. Currently, only Lumo is working in this direction, with challenges in collecting, cleaning, and labeling industry data.

2. Frameworks

There are many competitors in this area, which was the market's main focus in the past two months, including ElizaOS, Arc, Swarms, etc. The AI Agent framework is essentially a set of rules for calling various functional modules of LLMs, making it easier to create agents with unified rules.

3. Launchpad

The aforementioned frameworks are often open-source code, requiring users to understand some basic coding and purchase their own servers for deployment. The AI Agent Launchpad, on the other hand, is a SaaS version of the framework, providing a no-code environment for creating agents (users only need to manually input a few parameters) and comes with its own server. While this simplifies the process of creating AI Agents, it also reduces flexibility. Notable launchpads include Virtual, Vvaifu, AvaAI, etc.

4. Special Functions

There are also other projects focused on special functions to enable AI Agents to operate better, faster, and stronger.

For example, Web3 is a unique industry that frequently interacts with large sums of money, requiring sufficient independence and verifiability in the operating environment of AI Agents (e.g., ensuring that the wallet is operated by the AI Agent and not by an individual). Notable examples include Phala's TEE environment.

2. AI Agent (Agents for Specific Application Scenarios)

This direction is very diverse, representing exploration and innovation at the application end, which can be further divided into:

• Analytical: Analyzing and reasoning within a specific subfield and outputting content.

• Operational: Analyzing a specific subfield and directly performing operations.

Here are some examples to illustrate:

1. AIXBT (@aixbt_agent)

Currently the most well-known DeFi AI analyst, skilled in providing brief analyses of various market projects and interacting with people on X.

2. Truth of Terminal (@truth_terminal)

A pioneer in the AI Agent track, known for nonsense and issuing meme coins.

3. Ava (@AVA_holo)

The flagship agent of Holoworld, adept at providing random market analyses through video.

4. YNE (@yesnoerror)

Analyzes scientific papers and corrects errors.

5. Buzz (@askthehive_ai)

I haven't used it, but according to the introduction, users can interact with the system through a natural language interface, and the agent can handle various tasks including trading, staking, liquidity management, and market sentiment analysis.

These are some of my analytical frameworks for the Web3 + AI Agent track. Although we are currently in a trough, AI Agents in Web2 are also thriving and are expected to be one of the hottest tracks in 2025, which will undoubtedly bring more exciting products and imaginative possibilities to Web3. It is one of the few directions that can continuously attract attention and builders.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。