Autonomous agents no longer wait for human prompts. Valued at USD 6.8 billion in 2024 and forecast to expand at 30.3 per cent annually through 2034, these AI systems now span Web2 and Web3, independently managing assets, executing trades, generating content, and even hiring talent. Financially self‑sufficient and infinitely scalable, they are primed to redefine decentralised finance and the broader economy.
To guide us through this change we spoke with Victor Vernissage, a researcher at the intersection of Al, blockchain, and decentralized governance and a co-founder of multiple tech ventures, including Citadel.one, a next-generation crypto super app, and Humanode, an Al-driven biometric protocol that has authenticated over 500,000 users, explores how these autonomous Web3 agents are reshaping the DeFi ecosystem – offering both groundbreaking innovations and disruptive potential for economic and social structures.
What are the main types of AI agents today?
Despite the rapid growth of the autonomous AI agent market, the ecosystem is still fragmented and uneven. Agents vary not only in their technical capabilities but also in the environments where they operate and the tasks they are designed to perform. Broadly speaking, they fall into two categories: classical AI agents typical of the Web2 space, and crypto-native agents emerging within Web3.
Classical agents are actively developed by major tech companies. Google, for instance, is rolling out specialised agents for specific tasks, and GPT already includes built-in agent capabilities. Unlike general-purpose models, these agents are designed to fulfil narrowly defined objectives.
Crypto agents, on the other hand, follow a very different logic. They can be financially autonomous, but the space still feels experimental — more speculative and performative, and less mature overall. Growth in this area is largely driven by no-code frameworks that let users launch agents easily. People use them for everything from entertainment to investment tools. One of my teammates, for example, built a research-focused agent that analyses trends, makes scientific forecasts, and even assists with bets on emerging technologies.
How will agents impact employment and the economy?
Agents don’t just accelerate the flow of capital — they’re also beginning to replace humans in the workforce. On the one hand, this is part of the natural course of economic evolution. On the other hand, it means job loss for many. People may start using AI as their primary work tool, and that’s not necessarily bad. But if job displacement becomes widespread, a serious problem could emerge. For many, work is a central source of meaning, and finding a deeper, alternative purpose can be much harder. This could lead to depression and other negative consequences.
Adaptation mechanisms are already being explored. For instance, some protocols scan human identity and distribute basic income, offering financial support in an increasingly automated economy.
Who is actually building these agents?
Most traders are just trading. There are plenty of them in crypto, but far fewer people who are actually building things. Among the most notable teams are VirtuaLs, which provides a framework for launching agents and their associated tokens, and a16z, now rebranded as ElizaOS. Both briefly entered the top 100 cryptocurrencies but lost those positions in 2025.
Even before the major hype cycle that began in November, dozens of new teams had entered the space. Competition intensified quickly, but in terms of quality, most of the new crypto frameworks still lag behind their Web2 counterparts. In the corporate world, AI agents have long been used for automating sales, generating leads, and optimizing business processes. In crypto, however, development often leans more toward entertainment or speculation. Still, even these kinds of projects can reach billion-dollar valuations quickly, mainly because capital moves much faster in crypto.
Web2 companies have been at this for decades. The first agents appeared ten years ago but were mostly ignored because they were ineffective. That has changed. Today’s large language models can actually help with research, content creation, and real-world tasks. The Web2 ecosystem benefits from structure and long-term commitment. Web3, meanwhile, is driven by fast launches, viral appeal, and attention, often at the cost of real utility.
What are influencer agents, and how do they work?
One of the most striking and widely discussed cases in crypto is an agent called AIXBT. Its task is to scan all of crypto-Twitter, analyze trends, and publish hourly insights on what’s currently profitable — essentially identifying real-time alpha. But it doesn’t just post signals; it actively engages with users, replies to tweets, and has effectively become one of the leading voices in the space.
He’s actually more popular than many real people. In terms of reach and trust, AIXBT has outperformed all human influencers in the crypto space. It acts as a meta-influencer, aggregating and consolidating the voice of the community.
AIXBT was initially launched on the VirtuaLs framework, but its creator significantly extended its functionality by adding custom algorithms. As a result, each agent becomes unique — developers mix different models, design behavior rules, and fine-tune them with personalized prompts.
Twitter, however, isn’t the only domain where agents are active. There are more unusual examples. One of the most memorable is an agent called Spartan. Styled as a Spartan warrior, it speaks in a matching tone and uses computer vision to “see” what’s happening on the screen. Spartan physically moves the mouse, clicks through interfaces, and narrates its actions aloud in epic, battle-like commentary. One of its most iconic performances was choosing toilet paper on Amazon — carefully analyzing product features and then ceremoniously ordering the best option for its creator.
Why does crypto focus so much on hype over actual value?
The core currency in crypto isn’t technology — it’s attention. If a token gets noticed and bought, the project is already considered successful, regardless of its actual utility. Over the past two years, the market has been flooded with meme coins and AI agents, but very few of them offer real value. We’re still waiting for the focus to shift back to useful projects. For now, people just chase one trend after another.
Crypto investors rarely think long-term — they’re looking for quick returns rather than building sustainable infrastructure. That’s why even the simplest agents — those that just post on Twitter or perform absurd actions — often attract huge attention. Communities form around them, and native tokens are launched. Even if those tokens are essentially useless, the potential for speculation makes them appealing. Investors would rather bet on hype than support blockchain projects quietly operating in the background.
How do financially autonomous agents actually function?
First of all, these agents can be monetized through tokens, but the most interesting part is that they can actually earn money on their own. Each agent has its own crypto wallet. It can trade its own token, receive transaction fees, accept transfers, and manage funds independently.
What emerges is a class of digital entities that behave like economic actors. People really do send tokens to agents, hoping the agent will notice and mention them in a post. It’s a kind of marketing strategy. Things get truly interesting when the agent begins managing funds autonomously.
An agent can place an order — if crypto is accepted, it can pay. It can hire someone, for example, to integrate with a new platform. It can make trades on decentralized exchanges, accept investments, reinvest profits, and scale its operations. If the agent runs on an encrypted server, even the creator can’t interfere. That makes the agent autonomous, both financially and operationally.
This raises the question: are agents replacing people? Not exactly. The more transactions occur, the better the financial system works. Liquidity is a key factor. If money isn’t moving, the economy stagnates. Agents create new points of activity: they conduct research, make deals, manage capital. It’s not about displacement — it’s about expansion.
The economy is not a zero-sum game. It can be a positive-sum game, where everyone benefits. The more value is created, the more opportunities there are for growth. Agents help stimulate new industries, attract capital, and broaden the market.
Do creators always maintain control of their agents?
It all depends on how the agent is programmed. In some cases, the creator retains access; in others, they don’t. Why build an agent that doesn’t benefit the creator directly? First, for fun. That’s how many interesting projects get started. And second, the agent might end up hiring its creator if they turn out to be useful.
Fully autonomous agents are still rare. The first one was launched only in January of this year. But even without a direct link to its token, an agent can influence its value. A token is the agent’s reputation. It works like a meme coin: the agent is the meme, and the token is the way to participate in the hype.
In May 2024, Marc Andreessen, the founder of Andreessen Horowitz, awarded the first fully autonomous agent a $50,000 grant. This marked the beginning of a surge in attention: they became a subject of discussions, experiments, and investments. By November, a full-blown hype had emerged around the topic, largely driven by the appearance of frameworks that allowed people to launch their own agents, such as those for simple tasks like automatic Twitter posting. The first fully autonomous agent was launched in January 2025. One of the early examples was Terminal of Truth. After some time, it began to produce inaccurate and chaotic outputs, but this only heightened interest in the project.
