AI Agent Track Beginner's Guide

The development speed of AI is astonishing, and it will undoubtedly be dominated by AI in the future. If we add one more core element, it is undoubtedly the combination of AI and encryption technology.

Currently, AI has entered a new stage: AI Agent. This field is worth looking forward to both in terms of imaginative space and specific application scenarios.

The tide of the times is rolling forward, and we need to catch this train in time.

Recently, I have been studying knowledge related to AI Agents. This article documents my learning path and hopes to help everyone better enter this field.

This is the first article of the AI Agent track introductory guide, aimed at helping readers establish an overall understanding and framework. I will continue to delve deeper into this field, constantly improving and seizing the opportunities brought by the AI wave.

What is an AI Agent?

Putting aside complex concepts, we can directly compare the differences between AI Agents and existing large models like ChatGPT.

The current large models are more like powerful "natural language search engines" that can answer questions and provide suggestions, but they cannot truly make proactive decisions and execute actions.

The capabilities of AI Agents go beyond the scope of existing large models, no longer limited to "data processing", but able to complete a full loop from "perception" to "action".

A straightforward example: Now, if you ask ChatGPT how to invest in cryptocurrency, it will give you a bunch of suggestions, but an AI Agent can help you track global market information in real-time and dynamically adjust your portfolio to maximize returns.

From this, we can abstract the concept of an AI Agent: AI Agent( is a software entity based on artificial intelligence technology, capable of autonomously or semi-autonomously performing tasks, making decisions, and interacting with humans or other systems.

The most core difference here is: autonomous action.

How does the AI Agent achieve autonomous action?

AI can convert complex logic into precise conditions ) to return True or False based on context (, and then it can be seamlessly integrated into business scenarios.

First is intent analysis: AI will understand what the user wants to do by analyzing the user's prompts and context. It not only looks at what the user has said, but also considers the user's previous usage records and specific situations, and then translates these needs into specific program instructions.

Secondly, it assists in judgment: AI is like a smart assistant that can transform complex problems that are difficult for humans to handle into simple yes or no answers, or a few fixed options, after analysis. 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 Agents can be divided into two types:

One type is the AI Agent, which acts as a personal assistant and can help users handle certain tasks.

Another approach goes a step further, where the AI Agent itself is an independent entity with its own identity or brand, providing services to many users.

In summary, AI Agent can be regarded as the next development stage of large models and a new product form, with a vast space for imagination.

What is the relationship between AI agents and cryptography?

AI and cryptocurrency technology are not distinct; the two can be integrated.

More importantly, the AI Agent of Web2 is different from the AI Agent of Web3.

The Web3 AI Agent is a more advanced and complete AI Agent, perhaps it could be renamed: Crypto AI Agent.

With the power of cryptographic technology, AI Agent has gained more features:

) Decentralized

After integrating with cryptographic technology, the operations, data storage, and decision-making processes of AI Agents become more transparent and are not controlled by a single entity.

Web2 AI agents are typically controlled by centralized companies or platforms, with data and decision-making processes concentrated in one or a few entities.

Once an AI Agent provides services externally, there will be trust issues. Therefore, the AI Agent needs the operating or verification environment provided by the blockchain.

AI agents also require a barrier-free usage method, data transparency, interoperability, and decentralization.

( incentive mechanism

This is the most powerful enablement of cryptographic technology, providing a mechanism for directly incentivizing developers and users to participate and contribute through a token economic model.

Web2 AI Agents primarily rely on traditional business models, such as advertising revenue or subscription services, to sustain operations.

Web2 startup teams or companies struggle to be profitable for a long time and find it difficult to secure funding; however, in Web3, by issuing tokens, they can directly obtain cash flow to support project development, such as the use of AI Agents requiring cryptocurrency payments.

A free market economy can foster more innovation.

) true eternal life

With smart contracts, AI Agents have truly achieved "eternity".

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 code and decision-making mechanisms of AI Agents permanently exist on the blockchain, unless there is explicit logic to stop or change their behavior.

However, the data it relies on may need continuous updates or maintenance. Without ongoing data input or external interaction, the "immortality" of the AI Agent may be limited to its program logic and lack dynamism.

In summary, compared to the fact that encryption technology requires AI Agents, AI Agents need encryption technology even more.

The Narrative Evolution of AI+Cryptography Technology

The transition from large models to AI agents is two phases, and the integration of AI with cryptographic technology can also be divided into two phases:

Large Model Stage: Infrastructure

AI projects mainly have three evaluation dimensions: computing power, algorithms, and data.

In fact, the role of Web3 is to add an incentive system to AI, tokenizing computing power, algorithms, and data.

Therefore, the intersection of AI and Web3 can also be discussed from three dimensions: computing power, algorithms, and data.

算力###Computational Power###:

Distributed Computing Network: Blockchain inherently possesses distributed characteristics. AI can leverage the distributed network of Web3 to access more computational resources. By distributing AI's computational tasks across various nodes in the Web3 network, more powerful parallel computing capabilities can be achieved, which is particularly useful for training large AI models.

Incentive Mechanism: Web3 introduces economic incentive mechanisms, such as token economies, that 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, while providers receive token rewards.

算法(Algorithms):

Smart Contracts: Smart contracts in Web3 can automatically execute AI algorithms. AI can design algorithms to run in the form of 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 moderation.

Decentralized algorithm execution: In a Web3 environment, AI algorithms can operate without relying on a single central server, instead verifying and executing through multiple nodes together. This enhances the algorithm's resilience to interference and security, preventing single points of failure.

Data(Data):

Data Privacy and Ownership: Web3 emphasizes the decentralization of data and user ownership of data. The combination of AI and Web3 can leverage blockchain technology to manage data permissions, ensuring data privacy, while users can selectively share data in exchange for rewards, providing AI with richer yet controlled data sources.

Data Verification and Quality: Blockchain technology can be used for data verification, ensuring the authenticity and integrity of data, which is crucial for training AI models. Through Web3, data can be verified before being used, improving the quality and credibility of AI algorithm outputs.

Data marketplace: Web3 can facilitate the development of data marketplaces, allowing users to directly sell or share data with AI systems in need. This not only provides diverse datasets for AI but also ensures the liquidity and value of data through market mechanisms.

Through these convergence points, AI and Web3 can mutually develop and collaborate.

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 enhance the intelligence of its systems through AI, such as intelligent resource management and automated contract execution.

In relation to these three dimensions, several well-known projects have emerged in the market:

Computational Power ### Class Projects:

• Render Network: Although primarily focused on rendering, it can also provide AI computing power.
• Akash Network: Provides decentralized cloud computing resources for AI needs.
• Aethir: Focused on decentralized cloud computing, potentially involving the provision of AI computing power.
• ionet: A decentralized computing power platform that supports AI reasoning and training.

Algorithms ( Class Project:

• Cortex: A decentralized world computer capable of running AI and AI-driven DApps on the blockchain, focused on integrating AI into smart contracts.
• Fetchai: A blockchain-based machine learning platform that has launched the no-code management service Agentverse, simplifying the AI agent deployment for Web3 projects.
• iExec RLC: Provides a blockchain-based AI model marketplace, supporting confidential computing and decentralized oracles.

Data ) Data ( Class Project:

• Vana: Vana is building a DAO for personal genetic data, allowing users to control and potentially profit from a data market.
• RSS3: Launched an open-source AI architecture that enables any large language model to become a Web3 AI agent, involving the utilization and management of data.

Comprehensive Project:

• Myshell: A decentralized AI consumption layer designed to connect consumers, creators, and open-source researchers. It opens a platform where anyone can create, share, and monetize their AI-native applications.

Overall, in the stage of large models, the combination of cryptography and AI mainly lies in the infrastructure layer, laying the foundation for the long-term development of AI.

) AI Agent Stage: Application Implementation

The emergence of AI Agents marks the stage of AI's application layer landing.

AI Agents can also be divided into three development stages: Meme coin stage, single AI application stage, and AI Agent framework standard stage.

(# AI Agent Meme Coin

AI Agent Meme Coin is a very special existence; Meme coins themselves are products of community sentiment.

The development of AI is too fast, and this technology seems very profound, causing ordinary people to feel very anxious. AI Meme coins provide ordinary people with the opportunity to participate.

Therefore, AI Meme Coin brings a sentimental value to holders, allowing ordinary people to participate in the AI revolution.

The final result is: AI + MEME has accelerated the market education and dissemination of AI through the wealth effect.

Let's think from another perspective: why does the AI Agent issue tokens?

On one hand, attracting funds and users through the wealth effect injects momentum for the subsequent development of the industry; on the other hand, the MEME-based issuance method itself is a form of community financing that provides cash flow for the project's own development.

We can take a look at the top assets:

• $GOAT: The first AI Agent Meme coin that became popular;
• $Fartcoin: Attracting user attention by generating humorous content ) such as "fart jokes" (;
• $ACT: Aims to create a digital ecosystem for equal interaction between users and AI;
• $WORM: Aims to combine digital biology with blockchain technology to create a unique digital asset that simulates the neural system of biological worms.

)# Monolithic AI Application

AI Agent is integrating with various segments of cryptocurrency technology, presenting a flourishing situation.

With the development of AI Agents, the tokens issued by AI Agents are no longer just simple Meme coins; supported by actual use cases, they gradually possess the attributes of value coins.

Genesis Project

• ai16z: The first AI Agent to break out and established the first framework standard Eliza.

Agent Gaming

• ARC: Developed an AI framework called RIG based on the Rust language, supporting decentralized applications ###dApp### and smart contracts.
• FARM: Focused on using AI to enhance the realism and strategic depth of agricultural games.
• GAME: $GAME empowers AI agents with autonomous operations and intelligence, deeply integrating AI with gaming.

Agent DeFi

• $NEUR: Focused on token analysis and DeFi interactions, providing intelligent financial decision support.
• $BUZZ: Provides a natural language interface, allowing users to conduct and manage DeFi transactions more intuitively.

Code Audit

• AgentAUDIT: Automates code auditing using AI technology to improve code security and quality.

Agent Data Analysis

• REI: Conduct large-scale data analysis through AI technology, providing insights and predictive services.

Autonomous AI Agent

• LMT: An autonomous learning