AI智能體軌跡：綜合分析

The AI agent track has led the market in the past week, with a market capitalization exceeding $15 billion, becoming the biggest dark horse in the crypto market at the beginning of 2025. Many still regard AI agents as part of #Memecoin, but in fact, AI agents have long since detached from their initial meme attributes, combining technology and market in a positive way, and have already or will completely change some of our on-chain trading behaviors. This article will thoroughly analyze AI agents from various aspects, including their development, current track classification, future development directions, #token application scenarios, and how to better profit from AI agents.

AI代理賽道在過去一周領先市場，市值超過150億美元，成為2025年初加密市場最大的黑馬。屬性，以積極的方式將技術與市場結合起來，並且已經或將徹底改變我們的一些鏈上交易行為。本文將從各個方面深入分析AI代理，包括AI代理的發展、目前的賽道分類、未來的發展方向、#token應用場景以及如何更好地從AI代理中獲利。

1. Concept of AI Agent

1. AI Agent的概念

The essential characteristics of AI agents: they can autonomously plan and execute tasks, working towards established goals without human intervention. The difference from traditional internet robots is that they can make dynamic multi-step decisions, adjust behaviors based on interactions, and interact with other agents, protocols, and external applications.

人工智慧代理的基本特徵是：它們可以自主規劃和執行任務，在無需人工幹預的情況下朝著既定目標努力。與傳統網路機器人的區別在於，它們可以做出動態的多步驟決策，根據互動調整行為，並與其他代理商、協定和外部應用程式互動。

In simple terms, the AI we currently use only targets specific tasks, operates according to predefined rules, and provides a fixed set of responses, usually pre-programmed fixed outputs, lacking the ability to learn and improve. Typically, they can only generate text-based responses and generally cannot collaborate with external interfaces or other robots.

簡單來說，我們目前使用的人工智慧僅針對特定任務，根據預先定義的規則進行操作，並提供一組固定的回應，通常是預先編程的固定輸出，缺乏學習和改進的能力。通常，它們只能產生基於文字的回應，並且通常無法與外部介面或其他機器人協作。

In contrast, AI agents can make dynamic multi-step decisions and adjust based on feedback and interaction, enabling self-learning, thinking, and achieving independent operational goals. Correspondingly, they have numerous application scenarios and can interact with other agents, APIs, and applications.

相較之下，AI智能體可以根據回饋和互動做出動態的多步驟決策和調整，從而實現自我學習、思考並實現獨立的營運目標。相應地，它們有眾多的應用場景，可以與其他代理、API和應用程式互動。

2. Development History of AI Agents

2. AI智能體的發展歷史

The First Generation of AI Agents (First Wave)

第一代人工智慧代理（第一波）

The concept of AI agents did not emerge in 2025; it began to sprout at the end of 2023 and led to the creation of Autonolas and Bittensor in early 2024.

AI代理的概念在2025年才出現；它於 2023 年底開始萌芽，並於 2024 年初催生了 Autonolas 和 Bittensor。

Taking Autonolas (OLAS) as an example, it is a platform that supports the creation and management of autonomous agent services. These services are known as agent services, operating independently off-chain as part of a multi-agent system (MAS) to achieve common goals collaboratively. Autonolas enables developers to build and deploy autonomous agents that seamlessly collaborate off-chain while leveraging blockchain technology to enhance on-chain functionality. There are also some established agents on Autonolas, but it is evident that, in terms of market enthusiasm and capital flow, the first generation of AI agent projects did not gain anything from this wave of AI agents, and even saw a decrease in market capitalization, which is the fate of most currently popular AI agent projects.

以Autonolas（OLAS）為例，它是一個支援自主代理服務創建和管理的平台。這些服務被稱為代理服務，作為多代理系統 (MAS) 的一部分在鏈外獨立運行，以協作實現共同目標。 Autonolas 使開發人員能夠建立和部署自主代理，這些代理可以在鏈下無縫協作，同時利用區塊鏈技術增強鏈上功能。 Autonolas上也有一些老牌代理商，但很明顯，從市場熱情和資金流向來看，第一代AI代理商項目並沒有從這波AI代理商中獲得什麼收穫，甚至出現了市值的下降。就是目前最熱門的AI代理專案的命運。

The Evolution of AI Agents (Second Wave) and the Overlay Effect of Memes, Gaining Community Attention

AI智能體的演化（第二波）與迷因的疊加效應，獲得社群關注

In June 2024, Andy trained a Llama-70 B AI model based on chat records from Infinite Backrooms, his research papers, and content from 4 Chan and Reddit. This model was named Terminal of Truths (ToT). On October 10, 2024, an anonymous developer launched the $GOAT token on Solana's meme coin launchpad pump.fun.

2024 年 6 月，Andy 根據 Infinite Backrooms 的聊天記錄、他的研究論文以及 4 Chan 和 Reddit 的內容訓練了 Llama-70 B AI 模型。此模型被命名為真理終端（ToT）。 2024 年 10 月 10 日，一位匿名開發者在 Solana 的 meme coin launchpad Pump.fun 上推出了 $GOAT 代幣。

This was the first AI-related meme coin marketed by autonomous AI agents, and within the following month, Goat, backed by capital, saw its market capitalization exceed $1 billion. Meanwhile, a platform allowing users to create, deploy, and monetize AI agents, Virtuals Protocol, launched its first entertainment-focused AI agent - Luna, marking its first step towards a market cap of $10 billion. Community attention and capital backing ignited instantly. Compared to the first generation of AI agent projects, AI agents during this period were more closely connected to the community, starting from memes and allowing users to understand AI agents at a lower threshold.

這是第一個由自主人工智慧代理行銷的與人工智慧相關的迷因幣，在接下來的一個月內，Goat 在資本的支持下，其市值突破了 10 億美元。同時，允許用戶創建、部署人工智慧代理並透過其貨幣化的平台Virtuals Protocol 推出了首款專注於娛樂的人工智慧代理——Luna，標誌著其向100 億美元市值邁出了第一步。瞬間引發社會關注和資本支持。相較於第一代AI代理項目，這段時期的AI代理與社區的聯繫更加緊密，從梗開始，讓用戶以更低的門檻了解AI代理。

Framework + Launchpad, Enriching the Story of AI Agents (Third Wave?) Achieving a $10 Billion Market Cap

框架 + Launchpad，豐富人工智慧代理的故事（第三波？）實現 100 億美元的市值

The success of Virtual Protocol, starting with the launch of the first agent - Luna in early October, saw the platform token Virtual's market cap rise from $50 million to nearly $5 billion, a hundredfold increase, and the continuously increasing number of agents led to an explosive growth of Launchpad and AI agent platforms, officially opening the path to a $10 billion market cap for AI agents. Although the current enthusiasm and market cap of AI agents are declining, with the robust development of applications combining AI agents with DeFi, social, and consumer sectors, there is still a significant opportunity to reach a market cap of $100 billion (not investment advice).

Virtual Protocol 的成功，從 10 月初推出第一個代理 Luna 開始，平台代幣 Virtual 的市值從 5000 萬美元上升到近 50 億美元，增長了百倍，而代理數量的不斷增加導致了Launchpad和AI代理平台的爆炸性成長，正式開啟了AI代理市值100億美元的道路。儘管目前AI代理的熱情和市值正在下降，但隨著AI代理與DeFi、社交、消費領域結合的應用的蓬勃發展，仍有很大機會達到1000億美元的市值（非投資建議） 。

3. Classification of AI Agent Tracks

3. AI智能體軌跡分類

Overview and Development History of Leading Projects

主導專案概況及發展歷程

ai16z/eliza/degenai

ai16z/伊麗莎/德格奈

https://x.com/ai16zdao

ai16z currently has a market cap of $1.7 billion

ai16z目前市值為17億美元

eliza currently has a market cap of $70 million

伊麗莎目前的市值為 7000 萬美元

degenai currently has a market cap of $50 million

Degenai 目前市值為 5000 萬美元

Founder @shawmakesmagic, achievements

創辦人@shawmakesmagic，成就