Remember when chatbots, alongside record-breaking OpenAI, were dominating headlines? Well, it's time for a switch-up: AI agents are here, drastically changing how we interact in Web3. Unlike chatbots that wait patiently for you to inquire about the next big AI coin or ask if $FARTCOIN is a good investment, AI agents are more like digital assistants who can get things done autonomously. They're shilling tokens, making jokes, and engaging with communities on X - all while you go about your business. Exciting? Absolutely. But are we facing an AI agent supercycle?
If, according to Salesforce's research on AI agents, surprisingly unsurprising "54% of consumers don't care how they interact with a company", the agentic future seems more tangible. And we're getting closer to it every day. At this point, AI agents have taken over Crypto Twitter, with a so-called aixbt agent officially achieving a spectacular milestone of having the biggest mindshare. Impressive. After all, it joined the platform just a month ago.
But here is the drill: even though we are swamped by ChatGPT wrappers on our social feeds, there can be a sophisticated complexity behind these agents. This narrative is strong for a good reason-AI agents have a more holistic approach to completing tasks. While simple chatbots rely on LLMs, agents use them to interact with us but, at the same time, rely on other processes to analyze data in real time and, by their own initiative, execute its purpose.
In layman's terms, AI agents are purpose-driven programs that make decisions autonomously. Contrary to how chatbots work, they are not dependent on human interaction to engage in activities. The process under an agent's hood can be summarized into four steps: gathering data, (reinforced) learning, decision-making, and execution. They're undeniably popping up everywhere in the crypto world, dominating the industry - from content creation to market analysis, the label "Automated" keeps appearing more frequently on crypto Twitter.
While the technology stack of AI agents is big, spanning storage, models, memory, hosting, and more, it can be categorized into three main layers. The data layer is where all the raw information flows. The agents use blockchain nodes to stay on top of what's happening onchain or pull real-world data through oracles and other APIs. The next layer-the AI/ML component-makes AI agents so compelling. Reinforced learning, a staple in the AI agent software stack, is the key to making it evolve continuously. Lastly, the blockchain layer enables these agents to interact with smart contracts, sign transactions, and pay gas fees, all within their own onchain wallet.
The amount of development that is needed to build AI agents from scratch is no joke. GaiaNet, an open-source AI agent developer tool, showcases the architecture of its solutions, proving just how more advanced agents are from traditional bots.
Regardless of your technical background or lack thereof, platforms like Virtuals Protocol make it easy to launch an AI agent with its token. The process at Virtuals mostly focuses on filling out an agent creation form that goes through name, ticker, description, and profile picture, as well as providing initial liquidity to accumulate a required amount to launch a pool on Uniswap, while the last step involves governance. Quick, easy, and simple.
On the other hand, Based Agent is making waves on Base - Coinbase's L2 chain. Their solution helps users handle all onchain tasks thanks to agents having their own wallets - contrary to how Truth Terminal does not own "its" wallet address. But Base is not the only one that makes onchain agents deploy seamlessly with a ready-to-use kit. Both Injective and Solana announced SDKs to launch AI agents, making this race even more fierce. The quality of such an agent is key here. In the recent Context podcast from the BoysClub, we hear the co-founder of Bonsai, Carlos Beltran, say that "the usefulness aspect hasn't really been explored yet, I think we're gonna start seeing some really useful ones in a month or so." However, being able to deploy this autonomous software onchain is only the first step. Building an ecosystem where they can thrive is the second.
The future of AI agents is looking pretty exciting, and many speculate that the next bull run will be because of-yes, you guessed it-AI agents.
You can think of the agentic web as a new onchain playground where AI agents collaborate to achieve their predefined goals. Very soon, the trick will be finding the sweet spot between automation and authentic engagement. And although there is still a long way before we enter a truly agentic web, there are important steps to make AI agents useful on a bigger scale.
Yatharth Jain, co-founder of a coordination layer for AI agents called Cluster Protocol, compares the current most popular agents to ChatGPT wrappers with a custom knowledge base due to their relatively limited scope. He emphasized the role of incentivized community members in keeping a healthy knowledge base repository. After all, "public contributions over knowledge base matters to keep the training ongoing over inference data to make the agent almost perfect for trading," Yarhath says.
The crypto space has been overloaded with AI agents for the past 2 months, which may feel like two years due to its intensity. As befits crypto, decentralized AI is moving extremely fast. While in 2025 the foundations for the agentic web will be laid out, many variables will impact how this upcoming year unfolds. But one thing is certain: the AI agent supercycle is here.