There isn’t much content related to the new narrative of Web3 x Agent domestically, so it’s a pleasure to collaborate with the potential AI project HajimeAI on Solana for this research. Reflecting on January this year, the approval of Bitcoin spot ETFs in the U.S. triggered strong bullish sentiment among investors and positive capital inflows. The crypto market continued to grow in the first half of 2024, with a total market cap increasing by 37.3%. Several crypto narratives gained strong momentum, with Memecoin, AI, and RWA standing out as the strongest. Based on last week’s coin price performance, AI remains quite strong, just behind Memecoin.

Web3 x AI: This year, various Web3 VCs have heavily invested in this sector, even purchasing old projects like TAO (Bittensor) on the open market, so I don’t believe this narrative is fleeting. On the contrary, it will continue to innovate with the development of the traditional AI sector.

For example, the latest trend in traditional AI, AI Agents, has also been introduced into the Web3 world. In the first half of the year, a large number of Web3 x AI Agent projects were launched, such as Spectral and Olas Network, and many established projects are also following this narrative, such as Fetch.AI ($FET) and Phala ($PHA). At the recent ETHCC, many Web3 developers and top VCs began focusing on the AI Agent narrative.

This article starts with what I consider to be the two most representative new projects to quickly understand this new narrative and its opportunities.

Table of Contents:

1. What is an AI Agent?

2. New Changes in Web3 x AI Agents

3. Spectral Research

4. GaiaNet Research

5. Overview of Other Early Projects

1. What is AI Agent?

Simply put, an AI Agent is a type of agent based on large language models (LLMs) that can perceive the environment to autonomously understand, think independently, make decisions, and execute actions. Similar to the human process of “doing things,” the core functions of an agent can be summarized in three steps: perception, planning, and action.

So, how does an AI Agent differ from AI chatbots like ChatGPT? In terms of purpose and capabilities:

Chatbots are designed to interact with humans. Since AI chatbots are created to assist humans, they do not take autonomous actions.

Agents, on the other hand, are designed to complete tasks independently and have the ability to take autonomous actions. You don’t need to constantly tell them what to do; just provide them with a goal, and they will find a way to complete it automatically.

For example, an AI Agent is like a more intelligent assistant. If you are feeling unwell and tell it, “I’m not feeling well,” it will monitor your temperature and other health indicators, analyze data from the internet, and provide a conclusion like, “You have a fever.” It will then automatically generate a sick leave request for you and send it to your boss. Additionally, it might detect that you are running low on fever medication and automatically add it to your shopping cart, so you only need to pay and have it delivered to your doorstep within 15 minutes.

I have tried several popular Web2 AI Agents, such as Perplexity, CrewAI, AutoGPT, and MultiOn. Common features include batch document content extraction and internet information integration (such as generating research reports). I found MultiOn particularly interesting and will focus on it here. Its main feature is simplifying internet user interactions, freeing users’ hands, and changing the way users interact with the internet.

For example, when I tasked MultiOn with “finding the most-viewed Web3-themed video on YouTube,” it automatically completed the entire process from “opening YouTube in the browser,” “searching for Web3-themed videos,” “filtering by highest views,” and finally provided me with a video with 26 million views ✅.

2. New changes in Web3 x AI Agent

First, what can Web3 bring to AI Agents? In other words, what are the benefits of moving AI Agents to the blockchain?

Anti-censorship

AI Agents based on LLMs can experience bias due to centralized LLMs, which may limit the dissemination of true information. Using decentralized LLMs to build AI Agents can address this issue.

Decentralization/Ownership

Similarly, unless one invests a massive amount of data resources to build an LLM, the core data of an AI Agent will remain with centralized AI providers.

Monetization

AI Agent Launchpad? Token governance of an AI Agent, Initial Agent Offering (IAO), provides a way for AI Agent developers and investors to monetize.

Composability/LEGO

AI Agents can interact, trade, and enable functionality with other agents on the same network. If we consider the composability in DeFi, it would not be surprising if a single Web3 AI Agent could handle tasks such as selecting investment options, finding the best liquidity DEX, completing token swaps, and monitoring yield.

Conversely, what can AI Agents bring to Web3?

1. Conversational AI Agents on the blockchain can do more than Web2 Agents’ domain-specific knowledge collection and organization; they can also retrieve and aggregate Web3 information, greatly simplifying the on-chain research process for Web3 users.

2. Just considering the concept of AI Agents, MultiOn is particularly reminiscent of the Intent-Centric approach. It aims to free users’ hands and change how they interact with Web3, enabling mass adoption. This interaction includes but is not limited to swaps and airdrop interactions. Imagine telling an AI Agent, “Help me complete the Linea airdrop interaction,” and the AI Agent retrieves airdrop tutorials from KOLs online and completes the process automatically using a blockchain wallet. Or, “Help me build an ETHBot that buys ETH with 30% of my USDT balance when ETH falls below MA200,” providing 24/7 market monitoring. If every solver used an AI Agent that could perform on-chain interactions, it would complete a crucial piece of the intent protocol puzzle.

Regarding Intent-Centric, you can refer to my previous articles.

Here, I categorize the current Web3 x AI Agent space into two types: Web2-style on-chain conversational AI Agents and more Web3-native AI Agents.

The first type can be understood as AI Agents created on a specific Layer-1, suitable for Web3 users to learn domain-specific knowledge and conduct on-chain research, but does not include on-chain operations.

The second type involves on-chain logic, helping users perform specific on-chain interactions, including on-chain operations.

Of course, my classification is basic and focuses on whether an AI Agent has on-chain interaction capabilities, without considering whether it is based on reliable, verifiable decentralized AI models.

Below, I will introduce two representative projects, GaiaNet and Spectral, for a comparative analysis.

3. Spectral (with on-chain operation capabilities)

Originally a credit scoring protocol based on Ethereum, Spectral aimed to provide a new way for lenders to assess borrower credit risk. In Q4 2023, it transformed into a machine intelligence network, allowing users to build on-chain AI Agents and create an on-chain agent economy

1.Business Model
Spectral has four key products:

Spectral Syntax: A collection of agents developed by Spectral. Users can instruct an agent on what to do, and the agent will convert natural language intentions into executable code to help with tasks such as creating NFTs, launching Memecoins, automating trading, and retrieving on-chain information. For instance, the MoonMaker Agent can automate the entire process from naming, logo design, to CA deployment.

Spectral Nova: A decentralized platform that provides machine learning inference directly to smart contracts. Top scientists, businesses, developers, and engineers can build AI models here and earn revenue from user payments. Additionally, model creators can launch incentivized challenges where solvers (bounty hunters) tackle these challenges to win rewards or share in the revenue. Creators, solvers, validators, and consumers interact on Spectral’s machine intelligence network, creating a flywheel through incentives.

Agent Wallets, launching in Q3 2024, integrates AI Agents into wallets to help users perform on-chain operations and simplify the Web3 user experience. For example, it supports gasless transactions using USDC, with the Gas asset Agent automatically handling the swap.

Inferchain, a Layer1 project centered around AI Agents, will integrate Agents from Syntax and Nova to facilitate interoperability among these Agents. It is set to launch in Q4 2024.

1.Development History

2021-2022: Completed funding rounds as Web3 credit risk assessment infrastructure.

March 2024: Launched Syntax, officially transitioning to on-chain AI Agents.

May 2024: TGE, initiated the first season of airdrops.

June 2024: Partnered with crypto wallet Turnkey; Agent Wallet set to launch in Q3 (Turnkey previously raised $15 million, with Sequoia and Coinbase participating).

2.AI x Web3 Intersection

Inferchain is the final piece of the Spectral ecosystem, achieving its ultimate vision of enabling easy development of on-chain AI Agents that are interconnected, leading to transparent, decentralized, and verifiable AI applications in the Web3 space.

3.Value Created

Addresses the high trial-and-error costs and reliance on a single source of information in centralized AI.

Allows non-technical users to quickly create on-chain Agents.

Facilitates communication among on-chain Agents.

4.Token Economics

$SPEC: Tokens can be used for payments and accessing community-developed AI Agents. They are also used for decentralized governance and staking mechanisms:

In Spectral Syntax, staking $SPEC grants users the ability to create AI Agents and access community-created AI Agents.

In Spectral Nova, validators must stake $SPEC as collateral to verify challenges completed by solvers.

SPEC is listed on exchanges such as Bybit, Gate.io, and MEXC, with a circulating market cap of $85 million and a fully diluted valuation (FDV) of $800 million. Currently, only the first-season airdrop and market maker shares are in circulation, accounting for 10.3% of the total supply.

⚠️ In May 2025, unlocking for core contributors and investors will begin. If the crypto market continues to rise in the coming months, be aware of the potential sell-off risk after the unlock.

6.Team Background

Sishir Varghese - Co-founder and CEO

Previously co-founder and managing partner at AlphaChain, and strategic partner at Loopring.

Mihir Kulkarni - Product Lead

Previously institutional product operations manager at Coinbase.

7.Funding

Round 1:

Galaxy, ParaFi Capital, Maven 11, Alliance DAO, Rarestone Capital, etc.

Round 2:

General Catalyst, Social Capital, Jump Capital, Samsung Next, Circle Ventures, Franklin Templeton, Section 32, etc.

Includes major VC investments such as Franklin Templeton, Samsung, and Google.

8.AI Agent Usage

Q2 data released in June 2024:

Registered users: 65,362

Number of contracts generated by SYNTAX: 1,055,568

Average interactions per user: 25

Number of Memecoins created using MoonMaker: 5,043

4. GaiaNet (does not have on-chain operation capabilities)

GaiaNet is a distributed AI infrastructure that aims to gradually become a decentralized AI Agent ecosystem.

1.Business Model

Nodes are equivalent to Agents, meaning Node = Agent. In my own experience of setting up nodes, when I interact with the AI Agent corresponding to a node, the output results consume computational resources.

GaiaNet is built on Ethereum and aggregates various domain knowledge bases in the form of nodes to create an AI Agent network. This allows individuals and enterprises to quickly build AI Agents based on their specific domain expertise and provide these Agents to the demand side for profit.

The three core components of the GaiaNet network:

• GaiaNet Nodes

A comprehensive software stack that enables individuals and enterprises to quickly deploy AI Agents integrated with their domain expertise.

• GaiaNet Domains
A collection of nodes registered under internet domain names and managed by domain operators. The idea is that each domain is specific to a particular field, such as finance, healthcare, or education, and contains AI Agents with specific functions.

User process:
Users pay the node (AI Agent), with the payment held in a smart contract on the blockchain. A portion of the fee is taken by the domain operator, and the remaining is used to provide services to the user

• GaiaNet DAO

Stakers stake tokens with domain providers, providing “trust”;

↓

Domain providers manage qualified nodes, providing “guarantee”;

↓

Users choose AI Agents under the domain provider, and fees are shared among nodes, domain providers, and stakers.

Additionally, there is a role for component developers. Non-AI Agent developers can earn from AI Agent developers by fine-tuning components such as NFTs, models, knowledge bases, and plugins.

2.AI x Web3 Integration

Bringing the Web2 AI Agent ecosystem onto the blockchain.

3.Value Generated

Developers can more easily deploy Agents. GaiaNet nodes support all open-source LLMs, multi-modal models, text-to-image models, and text-to-video models, allowing for custom additions and model fine-tuning. Based on domain providers, each industry and field will have specialized knowledge AI Agents created.

4.Token Economics

No token issued yet. In GaiaNet, users pay with tokens to use other people’s Agents.

5.Team Background

Matt Wright - Co-founder and CEO

Graduated from UCLA, formerly Community Lead at Consensys, co-founder of EVM Capital, and worked at JP Morgan.

Shashank Sripada - Co-founder

Background in venture capital in Web2, founded and joined multiple VC firms. Graduated from the London School of Economics with some political background.

Sydney Lai - Development and Promotion Lead

Co-founder and CTO of EVM Capital, graduated from UC Berkeley.

6.Funding

Seed round on May 28, 2024, raised $10 million

Investors include Mantle Network, ByteTrade, EVM Capital, Mirana, Lex Sokolin (Generative Ventures co-founder), Kishore Bhatia (Superscrypt co-founder), and Brian Johnson (Crypto Lead at Republic Capital).

7.AI Agent Usage

User count unknown, currently 35 available conversational AI Agents covering fields like finance, crypto, and programming. Total number of nodes is 18,594. Due to the testing phase and multi-threaded node operation, it is estimated that many nodes are used for bulk airdrop collection.

5. List of other early projects

Zotto: Users can create their own AI Agents to implement trading intentions, such as mirroring smart money addresses and executing multi-condition trades. It recently launched on Testnet and is worth watching closely.

AgentLayer: A Layer 2 built on the OP Stack, with a focus on AI Agents. It aims to facilitate interoperability between both self-developed and community-developed AI Agents.

Olas Network: An off-chain AI Agent ecosystem where single or multiple Agents collaborate to complete tasks, with outputs transmitted to the blockchain.

Theoriq: Theoriq aims to be a modular and composable foundational layer for AI Agents.

AgentCoin: Transitioned from a mature Web2 AI Agent product, evo ninja, to create a general-purpose Web3 AI Agent.

Giza: Developing a Web3 AI Agent framework that uses ZKML for off-chain inference and on-chain execution.

Olas Network: A Web3 AI Agent ecosystem where single or multiple off-chain Agents collaborate to complete user-defined tasks and transmit outputs to the blockchain.

statement:

1. This article is reproduced from [PANews], the copyright belongs to the original author [Baize Research Institute], if you have any objections to the reprint, please contact the Gate Learn team, and the team will handle it as soon as possible according to relevant procedures.

2. Disclaimer: The views and opinions expressed in this article represent only the author’s personal views and do not constitute any investment advice.

3. Other language versions of the article are translated by the Gate Learn team and are not mentioned in Gate.io, the translated article may not be reproduced, distributed or plagiarized.