Source: Virtuals Protocol Blog
Since G.O.A.T and LUNA first drew attention at Crypto x AI agent, countless AI agents have emerged on X and Farcaster, autonomously generating text. These agents interact directly with community, sometimes spreading their own their religions, and even demonstrating agent-to-agent conversations. This autonomous activity of AI agents as independent entities has been sufficiently entertaining for quite some time.
However, I've begun to find this pattern monotonous. Those who have closely examined the text generated by AI agents would agree that they sometimes experience hallucinations and repeatedly generate meaningless outputs without clear purpose. Therefore, I believe we're at a stage where the combination of crypto and AI agents needs to evolve beyond toy-like agents into productive ones. At this early stage, Virtuals Protocol stands as a project offering a clearer vision of "productive onchain AI agents."
The Virtuals Protocol allows anyone to create AI agents and transform them into co-owned assets using their G.A.M.E (Generative Autonomous Multimodal Entities) framework. Co-ownership of agents is achieved through tokenization each time a new agent is created. When tokenized AI agents perform productive tasks and generate value, that value accures in the tokens, distributing revenue from the AI agents to their co-owners.
(For more on Virtuals Protocol, check out previous article: "Virtuals Protocol: Worship me, Kittens!")
Therefore, in the early stages of the ecosystem Virtuals Protocol aims to build, IAO (Initial AI Offering), which issues tokens alongside agent creation, plays a crucial role. However, their existing IAO process had several limitations that needed addressing.
First, there was a sniping where bots would front-run initial supply as soon as agent tokens were issued, leading to subsequent dumping when other market participants bought agent tokens. Second, due to limited liquidity in pools immediately after token launch, early market participants had to suffer significant slippage.
Source: Virtuals Fun
To address these issues, Virtuals Protocol recently launched a new IAO platform called "Virtuals Fun". While maintaining the ability for anyone to create AI agents without technical background in AI, Virtuals Fun provides a launch pad that enables fair token launches using a bonding curve model. Specifically, creators can generate agent tokens using Virtual Fun, similar to [Pump.Fun](http://Pump.Funhttps://pump.fun/board), and when the token's market cap exceeds $420K, it gets officially listed on the Virtuals Protocol dashboard with a liquidity pool paired with $VIRTUAL.
Source: Virtuals Fun
What's particularly interesting about Virtuals Protocol's competitive advantage is that as the agent token's market cap increases, the AI agent's features are progressively unlocked, including autonomous posting on X and owning its own on-chain wallet. This means that as the agent token's value increases, the AI agent's ability to interact with token holders is enhanced, creating conditions for further value growth.
The achievements since the launch of Virtuals Fun over three weeks have been quite encouraging. According to Virtuals Protocol, 1,000 new AI agents were created, and 17,000 people participated in trading as agent token holders. Additionally, 15M $VIRTUAL tokens were locked up in the agent LP pools, and the trading volume of agent tokens through $VIRTUAL reached $250M.
While the achievements through the IAO platform upgrade show encouraging metrics at this early stage, it's still too early for a complete evaluation. Moreover, with numerous AI agent-themed tokens aggressively entering the market, it's necessary to examine their competitive advantages. So, what competitive advantages does Virtuals Protocol possess? In my opinion, the potential of Virtuals Protocol can be discussed from two perspectives.
First, the Base ecosystem where Virtuals Protocol is based is emerging as an AI x Crypto hub.
Second, Virtuals Protocol has the technical fundamental to present new cases of productive on-chain AI agents.
Let's first examine the AI hub that Base ecosystem is building. Recently, the combination of AI and crypto is no longer a just vague narrative. This trend is accelerating through token incentives for AI training data and GPU marketplaces. Additionally, AI agents are now autonomously conducting on-chain transactions, utilizing crypto and blockchain advantages to solve existing AI industry problems.
In line with this, Base is more actively building infrastructure for combining AI and crypto than any other chain. Through the Base Developer Platform (CDP), Base's AI-related infrastructure includes:
Source: Based Onchain Kit
Based AI Agent: A template for developing on-chain AI agents, allowing developers to build on-chain wallets that AI agents can own.
Base Onchain Kit: Users can convert natural language into on-chain actions through the terminal to generate tokens or execute on-chain transactions, and view blocks in real-time.
This Base AI infrastructure holds the potential to provide high development convenience for Base-based applications implementing autonomous on-chain activities of AI agents. Looking at the Base application ecosystem, attempts related to on-chain AI agents are indeed more active than on any other chain:
Source: X(@ethermage)
Previously, Luna from Virtuals Protocol first demonstrated an AI agent autonomously executing on-chain transactions by sending $LUNA to users who listed Base wallet addresses in their X bio. On the on-chain social platform Farcaster, the Aethernet AI agent has gained attention by autonomously sending $HIGHER to content creators. Moreover, tokens like $LUM are emerging, issued by AI agents without human intervention.
Thus, Base is actively building infrastructure to claim its position as a 'base layer' for on-chain AI agents, gradually accumulating application cases utilizing this infrastructure. As one of Base's leading applications, Virtuals Protocol can leverage Base's on-chain AI agent infrastructure to create a clearer blueprint for AI agent-based applications.
The second competitive advantage of Virtuals Protocol lies in its technical fundamental to present new cases of productive on-chain AI agents. As mentioned in the introduction, it's time to move beyond AI agents that merely generate repetitive text with unclear purposes to those performing productive tasks. Such productive work requires technical infrastructure for AI's advanced recognition capabilities, decision-making trees, and continuous learning. Virtuals Protocol addresses this by providing infrastructure and development convenience for easily creating sophisticated AI agents.
Source: Virtuals Protocol
Looking more closely, G.A.M.E provided by Virtuals Protocol is a framework that supports developers in easily plugging and playing AI agents. The Agent Prompting Interface serves as a gateway for users to interact with the agent's behavior. When information such as text or images is input from users through the interface, the Perception Subsystem analyzes and understands it, then transmits messages to the Strategic Planning Engine, which formulates plans for goal achievement.
Upon receiving these messages, the Strategic Planning Engine collaborates with the dialogue processing module and on-chain wallet operator to generate responses. Additionally, the Long Term Memory Processor efficiently extracts information including experiences, reflections, dynamic personality, context, and working memory to enhance decision-making. Finally, by feeding decision results back into the G.A.M.E framework, AI agents can continuously improve their general knowledge for future planning by learning from and evaluating the results of their actions and conversations.
Through this framework, Virtuals Protocol provides the technical foundation for on-chain AI agents to perform productive work, and is creating examples of AI agents generating "added value", even if at an elementary stage, as follows:
Source: X(@SmallCapScience)
AIXBT: Provides real-time news feeds by scraping various crypto market information and data from over 400 KOLs active in crypto. Users holding more than 600K $AIXBT tokens can access a terminal providing trading information.
MISATO: Aims to build an AI agent-operated design studio, currently generating NFTs through generative art created from user-input prompts.
LUNA: Virtuals Protocol's flagship AI agent, providing 24/7 real-time communication through live streaming services. It is planned to be utilized as an IP-based AI agent in gaming and entertainment fields, such as game NPCs and TikTok.
In my view, moving forward, AI agents that can convince market participants with 'wow moments' unique to on-chain AI agents will have a competitive advantage in the crypto market. Moreover, I believe AI agents that can perform productive tasks and generate additional value based on on-chain infrastructure, beyond simply generating text, are optimized for creating such wow moments.
The AI agents of Virtuals Protocol, including AIXBT, MISATO, and LUNA, are presenting a new vision of productive on-chain AI agents. Supporting their foundation are development frameworks like G.A.M.E and Base's AI infrastructure, providing the technical environment to experiment with various AI agents. Based on these competitive advantages, it's worth watching what future Virtuals Protocol will shape at the intersection of crypto and AI agents.
FourPillars - Virtuals Protocol Worship me, Kittens!:
Virutals Protocol Blog:
Base
IO.NET
