AI is developing too fast. The future will definitely be a world of AI. If we add another core element, it will definitely be a world of AI+Crypto.
Today, AI has evolved to a new stage: AI Agent.
AI Agent is worth looking forward to both in terms of imagination and application scenarios.
The train of the times is whizzing by, and we need to get on it quickly.
I have also been learning about AI Agent recently. This article records my learning path, hoping to help everyone get started in the AI Agent track.
This article is the first in the introductory guide to the AI Agent track, and it is also to help everyone first establish an overall understanding and framework-based understanding. In the future, we will continue to delve deeper into this track, continue to improve, and grasp the AI wave.
01 What exactly is AI Agent?
Let’s put aside all the complicated concepts and directly compare the differences between AI Agent and existing large models (such as ChatGPT).
The current large models are more like powerful "natural language search engines" that can answer questions and provide suggestions, but cannot truly make proactive decisions and execution.
The capabilities of AI Agent go beyond the scope of existing large models and are no longer limited to "data processing", but are able to complete a complete closed loop from "perception" to "action".
Let’s use an intuitive example: Now if you ask ChatGPT how to invest in Crypto, ChatGPT will give you a bunch of suggestions, but AI Agent can help you track global market information in real time and dynamically adjust your investment portfolio to maximize your returns.
From this, we can abstract the concept of AI Agent: AI Agent (artificial intelligence agent) is a software entity based on artificial intelligence technology that can autonomously or semi-autonomously perform tasks, make decisions, and interact with humans or other systems.
The core difference here is: autonomous action.
How does AI Agent achieve autonomous action?
AI can transform complex logic into precise conditions (returning True or False depending on the context), which can then be seamlessly integrated into business scenarios.
The first is intent analysis: AI will understand what the user wants to do by analyzing the user's prompt words and context. It not only looks at what the user says, but also considers the user's previous usage history and specific circumstances, and then converts these needs into specific program instructions.
Secondly, AI can assist in making decisions: AI is like a smart assistant that can analyze complex problems that are difficult for humans to handle and turn them into simple yes or no answers or a few fixed options. This not only makes decision-making more accurate and efficient, but also works well with existing business systems.
According to the degree of autonomous action, AI Agent can be divided into two types:
One is that AI Agent is equivalent to a personal assistant and can assist users in handling some business.
The other approach goes a step further, where the AI Agent itself is an independent individual with its own identity or brand, providing services to many users.
In short, AI Agent can be said to be the next development stage and new product form of the big model, and AI Agent has a lot of room for imagination.
02 What is the relationship between AI Agent and Crypto?
AI and Crypto are not clearly separated; the two can be integrated.
More importantly, Web2’s AI Agent and Web3’s AI Agent are not the same.
Web3's AI Agent is a more advanced and complete AI Agent, perhaps it can be renamed: Crypto AI Agent.
With the power of Crypto, AI Agent has more features:
(1) Decentralization
With the combination of Crypto, AI Agent’s operations, data storage, and decision-making processes are more transparent and not controlled by a single entity.
Web2 AI Agent These agents are typically controlled by centralized companies or platforms, with data and decision-making processes concentrated in the hands of one or a few entities.
Once an AI Agent provides services to the outside world, there will be trust issues, so the AI Agent needs a running or verification environment provided by the blockchain.
AI Agents also require barrier-free usage, open and transparent data, interconnection and decentralization.
(2) Incentive mechanism
This is the strongest empowerment of Crypto. Through the token economic model, it provides a mechanism to directly incentivize developers and users to participate and contribute.
Web2 AI Agents mainly rely on traditional business models such as advertising revenue or subscription services to maintain operations.
Web2 startup teams or companies may not be able to make a profit for a long time and find it difficult to raise funds; but with Web3, they can directly obtain cash flow by issuing coins to provide support for project development. For example, the use of AI Agent requires Crypto payment.
A free market economy can give rise to more innovation.
(3) True eternal life
With smart contracts, AI Agent truly achieves "immortality".
As long as the smart contract is deployed on the blockchain, the AI Agent can automatically operate according to its rules and can theoretically run indefinitely.
Smart contracts can ensure that the AI Agent’s code and decision-making mechanisms persist permanently on the blockchain unless there is explicit logic to stop or change its behavior.
However, the data it relies on may need to be continuously updated or maintained. Without continuous data input or external interaction, the "immortality" of the AI Agent may be limited to its program logic and not dynamic.
In short, AI Agents need Crypto more than Crypto needs AI Agents.
03 The Evolution of AI+Crypto Narrative
There are two stages in the development of AI from big models to AI Agents. The combination of AI and Crypto can also be divided into two stages:
3.1 Large Model Stage: Infrastructure
There are three main evaluation dimensions for AI projects: computing power, algorithms, and data.
In fact, the role of Web3 is to add an incentive system to AI and tokenize computing power, algorithms and data.
Therefore, the combination of AI and Web3 can also be discussed from three dimensions: computing power, algorithms, and data:
(1) Computational Power:
Distributed computing network: Blockchain is naturally distributed. AI can use the distributed network of Web3 to obtain more computing resources. By distributing AI computing tasks to various nodes in the Web3 network, more powerful parallel computing capabilities can be achieved, which is especially useful for training large AI models.
Incentive Mechanism: Web3 introduces economic incentive mechanisms, such as token economy, which can motivate participants in the network to contribute their computing resources. Such mechanisms can be used to create a market where AI developers can purchase computing power for machine learning tasks, and providers are rewarded with tokens.
(2) Algorithms:
Smart Contracts: Smart contracts in Web3 can automatically execute AI algorithms. AI can design algorithms to run as smart contracts on the blockchain, which not only increases transparency and trust, but also enables automated decision-making processes, such as automated market predictions or content review.
Decentralized algorithm execution: In the Web3 environment, AI algorithms do not rely on a single central server, but are verified and executed by multiple nodes. This increases the algorithm's anti-interference and security, and prevents single point failures.
(3) Data:
Data privacy and ownership: Web3 emphasizes the decentralization of data and the ownership of data by users. AI combined with Web3 can use blockchain technology to manage data permissions and ensure data privacy. At the same time, users can selectively share data in exchange for compensation, which provides AI with a richer but controlled data source.
Data verification and quality: Blockchain technology can be used to verify data and ensure the authenticity and integrity of data, which is critical for the training of AI models. Through Web3, data can be verified before being used, improving the output quality and credibility of AI algorithms.
Data Market: Web3 can promote the development of data market, where users can directly sell or share data with AI systems in need. This not only provides AI with a diverse data set, but also ensures the liquidity and value of data through market mechanisms.
Through these points of integration, AI and Web3 can develop in synergy with each other:
- AI can obtain distributed computing power and high-quality data through Web3, while using smart contracts to improve the execution efficiency and transparency of algorithms;
- Web3 can use AI to enhance the intelligence of its system, such as intelligent resource management, automated contract execution, etc.
In view of these three dimensions, there are already many well-known projects on the market:
Computational Power Projects:
- Render Network: Although it focuses mainly on rendering, it can also provide AI computing power.
- Akash Network: Provides decentralized cloud computing resources that can be used for AI needs.
- Aethir: Focuses on decentralized cloud computing, which may involve the provision of AI computing power.
- ionet: A decentralized computing platform that supports AI reasoning and training.
Algorithms Projects:
- Cortex: A decentralized world computer capable of running AI and AI-driven DApps on the blockchain, focusing on integrating AI into smart contracts.
- Fetchai: A blockchain-based machine learning platform that launched Agentverse, a code-free management service that simplifies the deployment of AI agents for Web3 projects.
- iExec RLC: Provides a blockchain-based AI model marketplace that supports confidential computing and decentralized oracles.
Data projects:
- Vana: Vana is building a DAO for personal genetic data, allowing users to control and potentially benefit from a data marketplace.
- RSS3: Launched an open source AI framework that enables any large language model to be turned into an AI agent for Web3, involving the utilization and management of data.
Comprehensive Projects:
- Myshell: A decentralized AI consumption layer that aims to connect consumers, creators, and open source researchers. It opens a platform where anyone can create, share, and monetize their AI native applications.
In general, in the large-scale model stage, the combination of Crypto and AI is mainly at the infrastructure layer, laying the foundation for the long-term development of AI.
3.2 AI Agent Phase: Application Implementation
The emergence of AI Agents marks the entry of AI into the application layer.
AI Agent can also be divided into three development stages: Meme coin stage, single AI application stage and AI Agent framework standard stage.
1. AI Agent Meme Coin
AI Agent Meme Coin is a very special existence. Meme Coin itself is the product of community sentiment.
AI is developing too fast, and this technology seems very profound, which makes ordinary people very anxious. AI Meme Coin gives ordinary people the opportunity to participate.
Therefore, AI Meme Coin brings holders an emotional value of participating in the AI revolution, allowing ordinary people to participate in the AI wave.
The final result is: AI +MEME uses the wealth effect to accelerate the market education and dissemination of AI.
Let’s think about it from another perspective. Why does AI Agent need to issue currency?
On the one hand, it attracts funds and users through the wealth effect, injecting momentum into the subsequent development of the industry; on the other hand, the MEME-based issuance method itself is a means of community financing, providing cash flow for the project's own development.
We can look at the header label:
- $GOAT: The first popular AI Agent Meme coin;
- $Fartcoin: Attracts user attention by generating humorous content (such as "fart jokes");
- $ACT: Aims to create a digital ecosystem where users and AI interact equally;
- $WORM: Aims to combine digital biology with blockchain technology to create a unique digital asset that simulates the nervous system of biological worms;
2. Monolithic AI Application
AI Agent is integrating with various Crypto segments, presenting a flourishing trend.
With the development of AI Agent, the tokens issued by AI Agent are no longer simple Meme coins. They are supported by actual usage scenarios and gradually have the attributes of valuable coins.
(1) Genesis Project
- ai16z: The first AI Agent to go viral and established the first framework standard Eliza.
(2) Agent Gaming
- ARC: Developed an AI framework called RIG based on the Rust language, which supports decentralized applications (dApp) and smart contracts.
- FARM: Focuses on using AI to improve the realism and strategic depth of farming games.
- GAME: $GAME enables autonomous operation and intelligence of AI agents, and deeply integrates AI and games.
(3) Agent DeFi
- $NEUR: Focuses on token analysis and DeFi interaction, providing intelligent financial decision support.
- $BUZZ: Provides a natural language interface that enables users to conduct DeFi transactions and management more intuitively.
(4) Code audit
- AgentAUDIT: Use AI technology to automate code audits and improve code security and quality.
(5) Agent data analysis
- REI: Conducts large-scale data analysis through AI technology to provide insight and prediction services.
(6) Autonomous AI Agent
- LMT: An AI agent that learns and performs tasks autonomously, aiming to reduce human intervention.
- GRIFFAIN: An AI agent that can autonomously optimize its own behavior, especially for decision-making and strategy formulation in complex environments.
3. AI Agent Framework Standards
AI Agent framework standards are still in a state of chaos.
What is the AI Agent Framework Standard?
The AI Agent Framework standard simplifies the development and deployment process of AI Agents by providing a unified set of specifications and tools.
It allows developers to create an AI Agent that can interact with multiple clients (Twitter, Discord, Telegram, etc.), extend functionality through plug-ins, and leverage AI technology to enhance its intelligence.
These standards and basic libraries (such as memory storage, session isolation, context generation, etc.) ensure that the operation of AI Agents is efficient, safe, and user-friendly.
By connecting to various AI platform interfaces, the framework standard further enhances the capabilities of AI Agents, enabling them to leverage the latest AI technologies to provide better services.
In short, the AI Agent framework standard is infrastructure and platform, and it can form its own ecosystem, so the narrative space is naturally higher than that of single AI applications.
The main AI Agent framework standards are as follows:
- ai16z: Built the Eliza framework, which supports multiple platforms such as Discord, Twitter, Telegram, etc., allowing AI Agent to integrate seamlessly with these platforms.
- Virtual: Built the GAME framework, designed for games and virtual environments, allowing AI agents to operate autonomously or interact with players in these environments.
- Swarms: A multi-agent AI framework that allows developers to create and manage multiple AI agents. It is suitable for scenarios that require high-complexity coordination, such as simulating social behavior, automating complex business processes, or processing large-scale data.
- ZEREBRO: Built the ZerePy framework, equivalent to Optimism's OP Stack, making it easier and more standardized to develop and deploy single AI applications, allowing these agents to independently create and distribute content on social platforms.
Related ecosystems have emerged around these frameworks, and we need to focus on these ecosystems when studying related projects.
04 Conclusion
The AI Agent narrative has begun to explode.
There is a main narrative explosion in our industry every year. Around this main narrative, many star projects will emerge, and naturally there will be many opportunities.
For example, there was DeFi Summer in 2020, Inscription Summer in 2023, Meme Summer in 2024, and AI Summer is emerging in 2025.
Don't waste every rare opportunity to make money.
