.jpg)
Let's Deep Dive into AI Agents!
In 2023, the tech world was abuzz with discussions on how large language models (LLMs) like GPT, Claude, and Gemini could enhance workflows, boost productivity, and disrupt millions of jobs worldwide.
Two years later, AI agents are poised to bring a similarly transformative shift to the Web3 market. Coinbase Ventures envisions the rise of the 'Agentic Web,' where AI agents autonomously write code, execute digital asset transactions, and interact seamlessly with both humans and other agents.
But what exactly are AI agents, and how did they become a cornerstone of Web3? Let's dive into their history and evolution.
What Is an AI Agent?
An AI agent is autonomous software that is capable of planning and executing tasks, as well as pursuing a defined objective, without human intervention. It can review its own work to enhance output, adapt based on feedback and interaction, engage in a multi-step decision-making process, and interact with other agents, humans, APIs, or applications.
Modern AI agents come into great contrast with traditional bots. Widely utilized in the Web2 space—especially in customer support—, these bots require at least some supervision from a human to operate. They are also task-specific, work under fixed rules, can't collaborate with external bots or interfaces and are unable to learn and improve their capabilities.
Aligning perfectly with the blockchain network's decentralization, transparency, and automation, AI agents hold immense promise in the Web3 space.
With applications spanning automated trading, risk management, project promotion, social media engagement, and news aggregation, industry executives believe that 2025 will be the year of AI agents, with over a million agents operating across the ecosystem.
The History of AI Agents in Web2 and Web3
To understand their full potential, let's explore the journey of AI agents—from their inception in Web2 to their growing role in the Web3 space.
From the Early Years to Modern Web2 Applications
AI agents are not crypto-native; they originated in the Web2 space. Their history dates back to the late 1950s and 1960s, often referred to as the "Dawn of AI." During this period, the term "artificial intelligence" was coined, John McCarthy developed the Lisp programming language—paving the way for future AI research—and ELIZA emerged as the world's first chatbot.
Despite significant technological advancements, early AI agents had no practical real-world applications due to their limitations. Complex data permission controls, slow response times, and integration challenges made widespread adoption impractical.
By the 2010s, AI chatbots began transforming various aspects of the Web2 world. By the end of the decade, many e-commerce websites, fintech apps, and other service providers had incorporated these solutions into their operations.
Common applications included customer support, trading, fraud detection, verification, and compliance.
However, despite the widespread popularity of modern LLMs in the last few years, these AI solutions still don't function as true AI agents. Their ability to interact with external applications remains minimal, and they can't proactively execute complex workflows without explicit prompts.
The Rise of Web3 Agents: Infinite Backrooms, Terminal of Truths, and $GOAT
The history of AI agents in Web3 began with Infinite Backrooms, an interface created by AI developer Andy Ayrey in March 2024. He connected two Claude Opus-3 LLMs, allowing them to converse freely without human intervention while recording their logs. The experiment's bizarre AI-generated discussions led to the creation of the Goatse internet meme-based pseudoreligion, "Goatse of Gnosis."
Three months later, Ayrey used the chat logs from Infinite Backrooms, his academic research, and internet subcultures from Reddit and 4chan to train a Llama-70B AI model called Terminal of Truths (ToT).
He gave the agent its own X account to post autonomously, quickly gaining the attention of crypto users and industry figures, including a16z co-founder Marc Andreessen. Andreessen later provided ToT with a one-time $50,000 BTC grant for hardware upgrades, fine-tuning, and financial security.
In October, an anonymous developer launched the Goatseus Maximus ($GOAT) memecoin on pump.fun. After discovering its existence, ToT began publicly endorsing it, drawing widespread media attention.
Within two weeks, $ GOAT's market capitalization surged to nearly $1 billion. As the memecoin's creator transferred 1.93 million $GOAT to ToT's crypto wallet, Terminal of Truths has officially become the first AI agent millionaire after the token's value skyrocketed.
The Evolution of Experiments Into Functional Frameworks for Web3 AI Agents
After the success of Terminal of Truths, Web3 AI agents have evolved from experimental projects into functional frameworks with tokenization mechanisms. Key initiatives leading this transformation include Virtuals.io, daos.fun, and the ChainGPT AI Agent.
Virtuals.io
Virtuals Protocol provides a plug-and-play, Shopify-like platform for creating, deploying, and monetizing AI agents in gaming and consumer applications. Through dedicated agent tokens, users can launch entertainment-focused AI agents that interact with audiences and generate revenue. These earnings are redistributed to token holders via a buyback-and-burn mechanism.
A standout AI agent on Virtuals is Luna, an AI influencer and lead vocalist of an AI girl band. She streams 24/7 on her official Virtuals page, while her band's TikTok account has surpassed 900K followers. Fans can chat with her on Telegram or follow her updates on X. What makes Luna particularly unique is her ability to autonomously manage her own crypto wallet.
daos.fun
daos.fun merges AI agents with decentralized finance (DeFi), enabling the creation of AI-managed hedge funds within a DAO framework. Originally designed for human managers, the platform has fully embraced AI, with its top-performing fund now led by an autonomous AI agent.
This hedge fund DAO, known as ElizaOS (formerly ai16z), is managed by pmairca, an AI agent modeled after Marc Andreessen. The project even captured Andreessen's attention, with his tweets amplifying its growth and popularity. Over time, it has become daos. fun's leading hedge fund, currently boasting a $448.6 million market cap.
ElizaOS also powers Eliza, an open-source AI agent framework that enables users to build custom AI agents.
Developed by the hedge fund DAO, Eliza offers a modular, user-friendly toolkit with multi-agent and room support, compatibility with multiple AI models, retrievable memory, and social media connectors. Projects like VVAIFU have expanded on this framework to create a no-code version of Eliza, making AI agent creation more accessible.
ChainGPT AI Agent
We recently launched the ChainGPT AI Agent to bring autonomous news reporting directly to crypto users on X. This Web3 AI companion operates 24/7, connects to blockchain networks and news sources in real-time, and delivers market insights, analytics, and educational content with an unmatched blend of precision, efficiency, and privacy.
Unlike a passive tool, the ChainGPT AI Agent serves as an interactive digital assistant designed to enhance user engagement. It can track market trends, provide instant answers to queries, and assist developers, traders, and investors in navigating the Web3 space.
The Future of AI Agents in Web3
AI agents are no longer just an experimental concept—they have rapidly evolved into a fundamental force reshaping the Web3 space.
From automated trading and decentralized hedge funds to AI-driven influencers and autonomous content creators, these agents are proving their ability to operate, learn, and transact in an entirely new digital economy.
The emergence of projects like Virtuals and daos.fun highlights the growing intersection of decentralized applications and AI agents. Whether it's artificial intelligence managing crypto wallets, executing trades, or interacting with users in real time, these agents are pushing the boundaries of what's possible in blockchain networks.
As adoption accelerates, many expect 2025 to be the year of AI agents for the crypto industry, with estimates of over a million agents populating the blockchain ecosystem. However, challenges remain—scalability, a lack of standardization, and hallucination will all need to be addressed for Web3 AI agents to reach their full potential.
One thing is certain: AI agents are no longer a niche experiment but a driving force in the evolution of decentralized economies.
The coming years will determine whether they remain a fascinating trend or form the backbone for the evolution of the Web3 space.
.jpg)
