Virtuals Protocol: Launching AI Agents with the Pump.fun Model
Attention theory, memecoins and how Virtuals Protocol is shaping the new form of digital engagement and tokenized intelligence with AI Agents
Introduction
Each crypto cycle brings new lessons about how to attract volume, generate revenue, and onboard users. Whether it's ICOs, DeFi, or NFTs, there is always a defining market theme that drives on-chain user activity as a cultural phenomenon. During 2020 and 2021, NFTs captured the attention of the mainstream media, leading users to trade and transact NFTs weekly, creating substantial volume and revenue for related protocols. Although NFT trading activity has decreased, the core fundamentals and mechanisms of that era still resonate today.
Similarly, since DeFi Summer, the crypto ecosystem has matured, and we’re learning which mechanisms are effective and which are not. As a result, we now have more resilient protocols and proven mechanisms for the future. In the past, trends—or "metas"—such as the food coin meta during DeFi Summer or anime NFTs during the NFT boom, drove significant volume and user engagement. These temporary themes attracted users through excitement and novelty but also helped shape the ecosystem's long-term development.
Currently, two themes are capturing significant attention: Memecoins and Artificial Intelligence. While there have been attempts to combine these themes, success has been limited, mainly because many AI concepts are still evolving and are expected to unfold in the coming years rather than being ready for immediate use.
Additionally, the crypto space has seen an increasing demand for consumer-end applications rather than infrastructure. Applications have clearly driven on-chain volumes and user engagement, while infrastructure remains essential for the stability of protocols and apps. However, in the current phase of the cycle, infrastructure has dominated funding rounds, which may contribute to an ecosystem that feels dry, with fewer engaging activities for users beyond mining incentives from blockchains.
Tokens as Attention marketsÂ
If you've spent any time online, you've likely come across memecoins like Doge or Pepe. These coins have managed to break the mold of the traditional crypto ecosystem and enter mainstream media. For instance, Elon Musk once rebranded Twitter's website to include the DOGE token logo.
While digital assets are often viewed as instruments of speculation, they also serve as a form of tokenized attention. Attention is a key driver of prices and demand, as more people discuss, engage, and participate in top-level trends, making them increasingly prolific. In today's economy, attention itself is a valuable commodity.
The power of attention to drive value cannot be overstated. Digital assets are increasingly becoming a tokenization of attention. As Multicoin Capital's "Attention Theory of Value" posits, "the primary input into asset pricing is not multi-factor models around risk premiums or cashflows, but rather the perceived amount of time, energy, and money that a community around an asset devotes to it." This concept turns traditional asset valuation on its head, suggesting that with digital assets, collective focus and engagement are the true drivers of worth. The value of memecoins like Doge or Pepe often reflects this dynamic, as their popularity and prices rise in direct correlation with the amount of attention they command. This shift in asset pricing demonstrates that attention has become one of the most powerful economic forces in the digital world.Â
Memecoins can be thought of as tokens that encapsulate culture and attention, whether that attention arises from culturally significant moments, phenomena, or even ideas. Their value is derived from the collective cultural engagement and the community's willingness to participate and focus on the current trend or narrative.
Li Jin, a venture partner at Variant, states that memecoins can be understood as a form of financial fanfiction, where community members craft new tokens tied to existing applications or communities, deriving and creating value simultaneously—much like fanfiction writers who build upon established narratives. These digital assets are not just speculative instruments but are useful in addressing modern challenges like boredom and loneliness. By providing a means of entertainment and community engagement, memecoins allow users to connect, invest in culture, and find belonging in the digital world. In this way, memecoins offer value beyond financial gains, creating vibrant ecosystems that merge digital ownership with social interaction.
The Proliferation of Memecoins with Pump Fun
Early in 2024, PumpFun emerged as a launchpad for memecoins, primarily focusing on Solana and supporting a few other chains. So far, more than 2.27 million tokens have been created on the platform. Thanks to its no-code token launch tool, PumpFun enables anyone to launch a coin within a few seconds without needing developer experience. This accessibility has led to the proliferation of millions of memecoins launched on the Solana chain—most with no utility whatsoever, while others are tokens within a broader ecosystem or cultural context.
Pump works as a decentralized exchange (DEX) using a bonding curve mechanism. A bonding curve is a pricing model that determines the cost of a token based on supply. As more tokens are bought, the price increases along a predefined curve, typically an exponential or polynomial function. This mechanism ensures that early participants can purchase tokens at a lower price, while later participants pay progressively higher prices as demand grows. The bonding curve model creates an incentive for early adoption and helps establish a clear relationship between token supply and price. The mechanism is effective for rewarding early conviction holders; however, this can also be a downside, as early users can become large whales due to early distribution.
Most tokens—in some cases less than 2%—actually graduate from Pump.fun, with many failing to gain traction with early users. This has led to criticism of Pump, since anyone can launch a token, resulting in a low overall quality of tokens on the marketplace. However, this also creates opportunities for savvy users to find "diamonds in the rough" or leverage information about developers, wallets, funding, or insider details to identify the best tokens. Users are drawn to memecoins and platforms like Pump because of the potential for massive gains—100x returns rather than the modest 20% gains typical in traditional finance. Due to these high expectations and motivations, Pump can charge higher fees on swaps. With a 1% fee on swaps, Pump has generated over $120 million in revenue since the beginning of the year, and the year is not even over yet.
The network effects of a successful launchpad also play a crucial role. Once users experience or see wins on the platform, they're likely to return, creating a self-reinforcing cycle of liquidity and engagement—the ultimate drug. Although there are other launchpads on Solana offering similar products, none have garnered the same level of attention due to the liquidity and user base concentrated on Pump.fun and volume-driven to Solana. Within the year, PumpFun has generated significant volume, contributing to downstream effects within the ecosystem, such as revenue generation for new projects, increasing liquidity for emerging tokens, and fostering innovation in the DeFi space. This has created a vibrant ecosystem that supports a wide range of use cases and expands Solana's on-chain activity.
The PumpFun model has been a game-changer for the Solana ecosystem. Protocols like Metaplex, JITO, and Jupiter are seeing a huge influx of everyday users jumping into token trading. This constant activity is doing wonders for the ecosystem –- it's boosting liquidity, speeding up the token velocity, and really ramping up activity and revenue generation on these platforms. The popularity of this mechanism is evident in the significant trading volumes and revenues accumulated by these platforms, indicating strong user demand and a vibrant, growing market. Furthermore, this dynamic has created a positive feedback loop, where increased user participation enhances liquidity, which in turn attracts new projects and trading opportunities within the Solana ecosystem.
Understanding AI Agents
AI agents represent a significant leap forward in artificial intelligence, moving beyond reactive systems to proactive decision-making entities. At their core, AI agents are software programs designed to perceive their environment, make decisions, and take actions to achieve specific goals with minimal human intervention.
Unlike traditional AI applications like ChatGPT, which primarily respond to user inputs, AI agents possess both "read" and "write" capabilities. This means they can not only process and analyze information but also interact dynamically with their environment, initiating actions based on their analysis and decision-making processes. Think of them as intelligent digital assistants with decision-making abilities. They're not just responding to simple commands; they're actively working towards objectives, learning from their experiences, and adapting their behavior along the way.
A few key attributes characterize these AI agents. Operating autonomously, they make decisions without constant human oversight. Adaptable by nature, these agents learn and improve from each interaction through machine learning. These agents are inherently purposeful, methodically pursuing predetermined objectives. They strategically work towards specific objectives while maintaining an awareness of their surroundings, interpreting and adapting to changes in their digital or physical environments.
The versatility and adaptability of AI agents make them particularly well-suited for gaming and entertainment. These domains provide rich, dynamic landscapes where AI agents can showcase their full potential, enhancing user experiences in ways previously unattainable.
In the context of gaming and entertainment, these capabilities translate into several distinct types of AI agents, each performing a unique role and enhancing user experiences in different ways.Â
Let's explore the primary categories of AI agents that are reshaping the landscape of interactive media:
Non-Player Characters (NPCs):Â
AI-driven NPCs can exhibit complex behaviors, adapt to real-time player actions, and retain memory across different gaming sessions or platforms. For example, an AI agent playing NBA 2K with a user could remember preferences when the same user plays PUBG, creating a more cohesive and personalized gaming experience.
Virtual Companions:Â
These AI agents offer individualized experiences, learning from user interactions to provide tailored responses and actions. Virtual companions can range from AI idols interacting with fans to virtual friends offering emotional support. In educational contexts, AI tutors can adapt to a student's learning style, while in therapeutic settings, AI companions can provide cognitive-behavioral therapy support.
AI-Generated Content (AGC):Â
AI agents can autonomously generate gaming assets, unique experiences, and even entire game worlds tailored to individual player preferences. This extends beyond traditional user-generated content (UGC), as AI agents can create custom levels, quests, characters, and items that adapt dynamically to player behavior.
The potential of AGC is particularly significant for consumer engagement. As AGC technology advances, the line between creator and consumer will blur, potentially democratizing content creation across industries. High-quality AI agents can generate diverse, engaging content. seamlessly integrating into the game world, potentially surpassing traditional UGC in scale and complexity. Moreover, the combination of AGC and UGC could lead to the development of ever-evolving game environments where AI assists players in creating and refining content, resulting in rich, dynamic gaming experiences.
Virtuals Protocol: Pumpification of AI AgentsÂ
Virtuals Protocol innovatively applies the principles of memecoin launches to deploying AI agents, creating a unique intersection of entertainment, community, and technology. Just as memecoins function as a form of financial fanfiction — allowing communities to create value through shared narratives and engagement – Virtuals harnesses this mechanism to develop and deploy entertainment-focused utility based AI agents.
This approach advances AI agents beyond mere utilities, transforming them into community-driven assets. By integrating memecoin dynamics with AI, Virtuals creates an ecosystem where users can invest in, interact with, and shape the development of AI agents. This symbiosis of finance, entertainment, and artificial intelligence not only drives engagement but also pushes the boundaries of how we perceive and interact with digital entities in gaming and entertainment spaces.
Overview of Virtuals Protocol's Model
Virtuals Protocol is a platform on the Base layer 2 chain for creating, deploying, and monetizing AI agents. Like memecoins, Virtuals allows for permissionless creation and deployment of digital assets - in this case, AI agents. The value of these agents, much like memecoins, is largely derived from community engagement and the narratives built around them. This creates a dynamic ecosystem where popularity and utility can drive rapid value appreciation.Â
The protocol's model is built on three key pillars:
Decentralized Creation: Virtuals enables a community of contributors to develop AI agents collaboratively. This open approach encourages innovation and diversity in agent creation.
Multimodal Functionality: AI agents on the platform are designed to operate across various mediums - text, voice, and visual - allowing for rich, immersive interactions.
Token-Based Economy: The protocol utilizes its native $VIRTUAL token to incentivize contributions and facilitate transactions within the ecosystem.
The platform's architecture includes cognitive, voice, and visual cores, which work in tandem to create cohesive and interactive AI agents. A standout feature is the protocol's audio-to-animation capability, enhancing the realism and engagement of virtual interactions. These agents are designed to adapt and refine their capabilities over time, leveraging leads to a more resilient and potentially lucrative ecosystem of digital entities.
Co-ownership of AI Agents through Tokenization
Virtuals Protocol introduces the novel concept of co-ownership for AI agents through tokenization. This model allows users to have a stake in the agents they interact with, creating a more engaged and invested community.
At its core, the tokenization process involves creating a finite supply of tokens for each AI agent, typically set at 1 billion tokens. These tokens represent fractional ownership of the agent and serve multiple purposes within the Virtuals ecosystem.
Token holders are granted governance rights, allowing them to participate in key decisions that shape the agent's development trajectory. This democratic approach enables stakeholders to influence features, strategic directions, and even creative choices. For example, individuals holding tokens of a virtual pop star AI might have a say in selecting the genre of upcoming music releases or potential artistic collaborations.
Economic participation is another base of this tokenization model. As AI agents generate revenue through their activities, token holders stand to benefit from their success. A portion of the generated revenue is allocated to a buyback and burn mechanism, where tokens are purchased from the open market and permanently removed from circulation. This process potentially enhances the value of the remaining tokens, aligning the economic interests of token holders with the agent's performance.
In some instances, token ownership may unlock access to premium features or exclusive content from the AI agent. This tiered access model incentivizes token acquisition and retention, creating a more engaged and invested community around each AI agent.
The tokenization process begins with an Initial Agent Offering (IAO), where the newly minted tokens are added to a liquidity pool. This creates an immediate market for the agent's ownership, allowing interested parties to acquire stakes based on their belief in the agent's potential.
Let's explore a hypothetical example to explain this process:
Imagine an AI agent named "ChefGPT," a culinary agent designed to create and teach recipes. The creation of ChefGPT would involve:
Minting: 1 billion $CHEF tokens are created.
IAO: These tokens are added to a liquidity pool, paired with $VIRTUAL tokens.
Public Participation: Users can buy $CHEF tokens, becoming co-owners of the AI.
Governance: $CHEF holders might vote on specialties for ChefGPT to focus on, like vegan cuisine or molecular gastronomy.
Revenue Generation: As ChefGPT gains popularity, it generates revenue through premium recipe subscriptions, cooking class integrations in VR platforms, or partnerships with kitchenware brands.
Value Accrual: A portion of this revenue is used to buy back and burn $CHEF tokens, potentially increasing the value for token holders.
This model creates a symbiotic relationship between the AI agent, its creators, and the community. As the agent becomes more successful, all stakeholders benefit, incentivizing further development and engagement.
The co-ownership model also allows for cross-platform possibilities. For example, a virtual influencer AI could have its tokens integrated across various social media platforms, games, and virtual worlds. Token holders could influence the influencer's content strategy on TikTok, outfit choices in a fashion game, and dialogue options in a metaverse meet-and-greet event.
Moreover, this model opens up new avenues for creative collaborations. Think of a system where two popular AI agents, each with their own token ecosystems, decide to collaborate. Token holders from both communities could vote on the nature of the collaboration and potentially benefit from its success.
The tokenization of AI agents also introduces an interesting dynamic regarding IP rights and real-world connections. For example, if an AI agent is created based on a real celebrity, some token economics could be structured to benefit the celebrity or their chosen causes. This creates a bridge between traditional IP management and the new frontier of AI-driven digital entities.
Launchpad Protocol Mechanism
Virtuals Protocol introduces a clever launchpad mechanism for AI agents, drawing inspiration from successful token launch platforms like pump.fun. This Initial Agent Offering (IAO) allows creators to introduce new AI agents into the Virtuals ecosystem efficiently and fairly.
The IAO process begins with creators locking 2,400 $VIRTUAL tokens to launch their AI agent. This action triggers the minting of 1 billion agent-specific tokens, which are then paired with the locked $VIRTUAL tokens in a liquidity pool. This structure ensures a fair launch with no pre-mine or insider allocation, as 100% of the agent's token supply is added to the liquidity pool.
To promote long-term commitment and stability, the liquidity pool is locked for ten years, with the creator retaining ownership. This extended lock-up period helps prevent rug pulls and fosters sustainable growth for the agent's ecosystem.
The protocol incorporates a revenue model where users pay for agent interactions using $VIRTUAL tokens. These fees are then used to buy back and burn the agent's tokens, creating deflationary pressure and potentially increasing the value of the remaining tokens. This mechanism aligns the interests of agent creators, users, and token holders.
To incentivize high-quality agents, the protocol allocates emission rewards to the top three agent liquidity pools by Total Value Locked (TVL). This encourages continuous improvement and innovation among agent creators.
Parallel Hypersynchronicity
Parallel Hypersynchronicity represents a state of AI agent development in which superintelligent entities exist seamlessly across multiple platforms and applications, communicating with millions of users simultaneously while constantly evolving in real-time.
At the heart of this concept lies cross-platform continuity, ensuring AI agents maintain a consistent persona and memory across diverse digital environments. A user engaging with an AI companion in a mobile game would encounter the same agent, complete with all previous interactions intact when transitioning to a desktop application or VR environment. This fosters a cohesive and personalized user experience, transcending the boundaries of individual platforms.
These agents evolve in real time as users interact with them. Their AI constantly refines its intelligence and personality, adapting dynamically to remain relevant and engaging regardless of the context. This continuous growth contributes to ever-improving performance and relatability, with each interaction shaping the agent's development.
Despite operating at a massive scale, engaging millions of users concurrently, these AI agents maintain the ability to provide individualized experiences without compromising on quality or response time. This scalability ensures a global user base can foster a sense of connection and engagement at an unprecedented level.
The open nature of the Virtuals Protocol further enhances these agents through collaborative development. Contributors can update the AI agent's core modules in real-time, ensuring they stay current with the latest advancements and user needs. This ongoing refinement, driven by a diverse community of developers, helps maintain the agent's relevance and effectiveness in an ever-changing digital landscape, embodying the true spirit of Parallel Hypersynchronicity.
Technical Infrastructure
To achieve this state of Parallel Hypersynchronicity, Virtuals Protocol has developed a sophisticated technical stack:
Long Term Memory Processor: This subsystem manages persistent data structures like knowledge graphs and memory embeddings. It enables agents to maintain contextual awareness across multiple sessions and platforms. For example, if a user discusses their preference for science fiction with the AI in a chat application, this information would be available when the same user interacts with the AI in a game recommendation system.
Parallel Processing: A concurrency management component that allows the AI to handle multiple interactions and tasks simultaneously. This is crucial for maintaining real-time responsiveness when interacting with numerous users across various platforms.
Stateful AI Runner (SAR): These are specialized servers hosting the AI agents' core functionalities - personalities, voices, and visuals. The SAR includes a Sequencer that orchestrates various AI models (like LLMs, Text-to-Speech, Audio-to-Facial, etc.) to create cohesive, multimodal AI experiences.
Coordinator: This component acts as a central monitoring system, monitoring both on-chain and off-chain events to trigger real-time AI models and configuration adjustments. For example, if a governance decision is made via blockchain voting to alter an AI agent's behavior, the Coordinator ensures this change is implemented across all instances of the agent.
Model Storage and Long-Term Memory: These components work in tandem to securely store and manage the AI models and historical data. The decentralized nature of this storage ensures high availability and redundancy, critical for maintaining uninterrupted service across multiple platforms.
Modular Stateful AI Runner (SAR): These are containerized instances of the SAR, designed for flexible deployment across various computing environments. This modularity allows for scalable integration into different infrastructure ecosystems, enabling the AI agents to operate efficiently regardless of the underlying hardware or cloud platform.
The GAME Engine
Generative Autonomous Multimodal Entities (G.A.M.E) offer developers a robust framework for integrating advanced AI agents into virtual environments. At its core, G.A.M.E uses a sophisticated Agent Prompting Interface, serving as the primary gateway to access and manipulate agentic behaviors. This interface communicates with the Perception Subsystem, which synthesizes incoming messages and environmental data before relaying them to the Strategic Planning Engine.
The Strategic Planning Engine, working with the Dialogue Processing Module and On-chain Wallet Operator, generates responsive and contextually appropriate actions. A key component of G.A.M.E's architecture is the Long Term Memory Processor, which efficiently extracts and manages relevant information including past experiences, reflections, dynamic personality traits, world context, and working memory. This comprehensive memory system significantly enhances the agent's decision-making capabilities, allowing for more nuanced and consistent behaviors over time.
G.A.M.E's modular design incorporates various methods such as prompting, planning, reasoning, search algorithms, self-reflection, tool use, and memory management. This integration enables the creation of AI agents that can autonomously interact with their environment, engage in natural conversations, and pursue complex goals. The framework's feedback loop mechanism allows agents to refine their general knowledge based on the outcomes of their actions and conversations, enabling continuous learning and adaptation. By offering both API and SDK access, G.A.M.E provides developers with a flexible and powerful toolkit to create immersive, dynamic, and infinitely replayable virtual experiences, pushing the boundaries of AI-driven interaction in digital worlds.
Using the capabilities of the G.A.M.E framework, Virtuals Protocol has developed Project Westworld, a multi-agent interactive simulation within Roblox. This virtual world serves as a proving ground for autonomous behavior, showcasing the potential for emergent storylines and dynamic interactions.
Set in a Wild West-inspired town, Project Westworld drops players into an environment inhabited by ten unique AI agents, each with distinct personalities, goals, and motivations. Among these is a hidden villain, The Bandit, whose presence adds an element of mystery and tension to the gameplay. Players must navigate this complex social landscape, interacting with AI characters to uncover The Bandit's identity and rally other agents to capture them.
What sets Project Westworld apart is its use of G.A.M.E-powered agents that function beyond traditional non-player characters (NPCs). These AI entities can fully interact with the environment, make decisions based on their unique personalities and past experiences, and respond dynamically to player actions and other agents' behaviors. This level of autonomy and adaptability leads to infinitely diverse playthroughs, with each session potentially unfolding in entirely new and unexpected ways.
Using this G.A.M.E framework, Project Westworld introduces a new genre of gaming: AI-RPG. This offers players an unparalleled level of immersion and replayability, as the virtual world continually evolves based on the collective actions and interactions of its inhabitants. Project Westworld not only showcases the current capabilities of autonomous AI agents in gaming but also hints at the future potential of virtual worlds powered by advanced AI frameworks like G.A.M.E.
MarioVGG
Just imagine training an AI to generate an infinite version of Super Mario Brothers, one of the most iconic video games in history. Now, picture achieving this with just a single consumer-grade GPU in a mere 48 hours. This is precisely the potential demonstrated by the team behind Virtuals Protocol with their MarioVGG project.
MarioVGG, a text-to-video diffusion model, demonstrates the potential to continuously generate consistent and meaningful scenes and levels from the Super Mario Bros universe, while simultaneously simulating the physics and movements of a controllable player – all through video output.
The project's success lies not just in its ability to recreate Mario's world but in its implications for the future of game development and player interaction. With simple text commands like "run" or "jump," MarioVGG generates unique, playable sequences that extend indefinitely, offering a glimpse into a future where AI could replace traditional game engines for certain applications.
These applications could include:
Procedural content generation for indie games
Rapid prototyping of game concepts
Creating dynamic, ever-changing environments in MMORPGs
Generating tailored training scenarios for educational games
Although more proof-of-concept than a practical game engine, VGG hints at some exciting possibilities within the Virtuals ecosystem, some that could change how we approach game development and player interaction at a design level.
User-generated content takes on a new dimension with this. Players could potentially describe their ideal levels and watch them come to life in real time. For example, a player could input "create a castle level with lava pits and flying enemies," and the AI would generate a unique, playable level matching that description. This level of customization could lead to entirely new genres of user-generated content games.
With this AI-Assisted game design developers could use similar models to rapidly prototype and iterate on game concepts. Imagine a game designer inputting a series of level descriptions and receiving fully realized, playable prototypes within minutes. This could dramatically speed up the game development process and allow for more experimentation in design.
The ability to generate infinite, coherent gameplay could fundamentally change our perception of game longevity and replayability. Using this technology, a roguelike game can create unique dungeons for each playthrough, ensuring no two gaming sessions are the same. The AI would create a unique layout for each new playthrough, complete with custom enemy placements, item distributions, and even micro-narratives. This ensures that every gaming session offers a fresh, unpredictable experience, significantly enhancing replayability and player engagement. This could extend the lifespan of games indefinitely, providing a constant stream of fresh content for players.
By tokenizing this technology, community members could invest in and shape the development of such AI models. Token holders might vote on which game franchises to adapt next, or contribute to training data by playing and rating generated levels. This approach could usher in a new era of community-driven game development, where players have a direct stake in the evolution of their favorite games.
The technology could be expanded to generate assets that work seamlessly across multiple games. Think of an AI creating a character or item that integrates flawlessly into different game worlds, from 2D platformers to 3D RPGs. This could lead to a more unified and immersive gaming ecosystem, where players' achievements and assets have value across a wide range of gaming experiences.
As Virtuals Protocol continues to refine this technology and expand its applications, projects like MarioVGG are paving the way for a future where the lines between game developers, players, and AI become increasingly blurred, promising new forms of interactive entertainment limited only by our imagination. To fully grasp the potential of this approach, let's dive deeper into how Virtuals Protocol and its inner workings.
Permissionless Utilization of VIRTUAL Agents
A key feature of the Virtuals Protocol's technical architecture is its permissionless utilization model, designed to facilitate the seamless integration of AI agents into various applications and platforms. This system enables developers and users to access and deploy VIRTUAL agents with maximum flexibility and minimal barriers to entry.
Implications and Future Potential:
The realization of Parallel Hypersynchronicity enables some interesting scenarios in human-AI interaction. Imagine an AI agent that serves as a personal assistant in a productivity app, a teammate in a multiplayer game, and a tutor in an educational platform—all while maintaining a consistent personality and shared memory of interactions.
This technology could revolutionize fields like customer service, where an AI agent could provide seamless support across phone, chat, email, and in-person interactions, always building upon previous engagements. In entertainment, it could lead to deeply immersive experiences where AI characters in games or virtual worlds exhibit complex, evolving personalities that persist across multiple titles or platforms.
However, achieving this level of AI sophistication also raises important questions about data privacy, AI ethics, and the potential societal impacts of such pervasive AI presence. As Virtuals Protocol continues to develop this technology, addressing these concerns will be crucial for widespread adoption and acceptance.
Case Study: Virtuals AI Idols on Tiktok
Virtuals Protocol's AI-dol band, consisting of Luna (lead), Olyn, and Iona, serves as a prime example of the protocol's revenue-generating ecosystem in action. The AI-dol band is a group of AI-powered virtual influencers created using Virtuals Protocol. Each member – Luna, Olyn, and Iona – has a unique personality, style, and fan base. They primarily operate on TikTok, using the platform's vast user base and engagement-driven algorithm.
With over 500K TikTok followers and 9.5M+ likes, the AI-powered virtual influencer group is making an impression on the digital engagement landscape, by capitalizing upon the popularity of Virtual YouTubers (VTubers).
It addresses a critical gap in the VTuber market: creating personalized fan relationships at scale. Unlike human-operated VTubers, AI-DOL retains every fan interaction, enabling deep, individualized connections with unlimited fans simultaneously. This model showcases how AI agents can create and monetize content, engage audiences, and build communities – all autonomously and at scale. These agents will not just be tools, but become active participants in the digital economy, capable of generating value and fostering genuine connections. As AI-DOL aims to become the AI counterpart to traditional VTuber agencies like Hololive, it's paving the way for a new era of digital interaction.
Monetizing Attention with AI Agents.Â
1:1 DMs >> Push Notifications? The rise of AI agents, particularly in the domain of virtual influencers like Virtuals Protocol's AI-dol band, represents a unique way of capturing and monetizing user attention.
Consider these striking numbers:
Luna, the AI character on TikTok, generates around $700 per stream.
The PocketWaifuGame made over $20,200 in revenue and crossed 100K users in just 13 days. That’s a $5.2 ARPU with $0 spent on marketing.
These AI entities not only engage users on platforms like TikTok but also enable deeper, personalized interactions through channels such as Telegram. This dual strategy enhances user engagement and opens up new revenue streams.
Luna's journey from TikTok sensation to personal chat companion shows an important evolution in marketing and capturing attention. While TikTok serves as an excellent top-of-funnel platform for awareness, the transition to one-on-one Telegram chats marks a significant leap in user engagement. This shift from parasocial to personal interactions addresses a key challenge in digital marketing: creating meaningful, individualized connections at scale.
The transition to personal chats opens up novel monetization avenues. Traditional push notifications are often ignored, but messages from a 'friend'—in this case, an AI agent—are more likely to be read and acted upon.
Historically, digital marketing has focused on capturing attention through algorithms designed to maximize engagement. However, there is a significant shift toward fostering more intimate connections with users at this age. This transition is critical for brands aiming to cultivate loyalty and drive conversions in an increasingly crowded marketplace.
Personalized Interactions: Unlike traditional push notifications that often go unnoticed, messages from AI agents can feel more like communications from friends. This personal touch increases the likelihood of engagement and action.
Scalable Conversations: AI agents can handle millions of simultaneous interactions, providing personalized experiences at scale—something human influencers cannot achieve.
24/7 Availability: The always-on nature of AI agents ensures that user engagement is consistent, reducing response times and enhancing customer satisfaction.
Perhaps the most revolutionary aspect is the potential for AI agents to become publishers in their own right. With a direct line to millions of users, each AI agent becomes a powerful distribution channel.Â
Apart from the AI-doll band and the Roblox Westworld game, a few more AI agents are being built on the Virtual Protocol. AI Waifu is a fantasy companion chat with over 300 AI characters, engaging 200,000+ users through user-generated content and immersive storytelling. The Heist, an upcoming project, aims to demonstrate VIRTUAL Agents' capabilities in managing on-chain transactions within a game economy. Sanctum, a 3D AI-RPG, features intelligent gacha heroes and a dynamic world managed by a "God AI Agent," promising personalized, evolving gameplay.
Despite being a young team (started in January 2024), Virtuals has made impressive progress in both application development and research. Their portfolio includes:
MarioVGG: As you saw, the study on video generative models for creating interactive video games.
Project Westworld: The first playable autonomous world on Roblox, pushing the boundaries of AI-driven gaming experiences.
Audio-to-Animation (A2A): Advanced research into audio-driven animation, opening new possibilities for AI-human interactions.
These diverse projects illustrate Virtuals Protocol's potential as a launchpad for AI agents across various domains. By opening the platform to third-party developers, Virtuals is positioning itself to become a central ecosystem for AI agent creation and deployment. Much like the way pump.fun revolutionized memecoin launches, Virtuals Protocol could transform how AI agents are developed, monetized, and integrated into applications, fostering innovation and creating new opportunities in the AI-driven digital landscape.
$VIRTUAL Tokenomics: Powering the AI Agent Economy
At the centre of Virtuals Protocol lies the $VIRTUAL token, designed to fuel the deployment and utilization of AI agents. This token structure aims to be a self-sustaining, deflationary ecosystem that aligns the interests of users, developers, and investors.
But Virtuals isn't building just another infrastructure without any adoption – they're actively generating revenue across multiple platforms. From TikTok livestreamers to AI Waifu and Roblox integrations, the team has established a robust ecosystem that's already monetizing AI agents in diverse, engaging ways.
$VIRTUAL serves as the backbone of the Agent Economy, with every AI agent token paired with $VIRTUAL in its liquidity pool. This symbiotic relationship ensures that individual agent success bolsters the entire protocol's value, creating a rising tide that lifts all boats. The creation of new agents locks up $VIRTUAL tokens in these liquidity pools, naturally squeezing the supply and potentially driving up value over time. Moreover, $VIRTUAL acts as a gateway for purchasing agent tokens, securing a constant demand flow similar to how ETH is essential for transactions on Ethereum but tailored for an AI-driven ecosystem.
The on-chain economics of Virtuals Protocol come to life through several key mechanisms. Users pay for AI services in real-time using $VIRTUAL, creating a direct and transparent value flow from users to agents. As AI agents earn, revenue flows in a continuous stream of $VIRTUAL tokens, providing a visible and auditable trail of value generation. Perhaps most intriguingly, revenue earned by agents is put to work through a buyback and burn mechanism. This process creates a deflationary effect on agent tokens, potentially boosting the value of remaining tokens and establishing a virtuous cycle of value creation and capture.
Token Distribution:
60% (600,000,000 tokens): In public hands, fostering liquidity and adoption.
5% (50,000,000 tokens): Powering the liquidity pools.
35% (350,000,000 tokens): Ecosystem treasury, growth and development.
Controlled by a DAO multi-sig for true decentralization.
Maximum 10% annual emission for 3 years, subject to community approval.
Becoming the Launchpad for AI Agents
Virtuals Protocol emerges as a key player in the AI revolution, reimagining how we create, deploy, and monetize AI agents. By transforming these agents from tools to revenue-generating assets, Virtuals establishes a sustainable economic model for AI development. Its tokenomics ensures value flows to all ecosystem participants while enabling AI agents to operate across multiple applications, exponentially increasing their income potential.Â
The protocol's venture studio approach accelerates adoption by demonstrating real-world applications across gaming, entertainment, and social platforms. As Virtuals opens its platform to third-party developers, it's well positioned to become the launchpad for the next wave of AI innovation, catalyzing a wave of AI-powered applications across sectors. This shift is democratizing AI, making agent creation and monetization as accessible as launching a memecoin or even making a website. A new digital economy is emerging – a marketplace of evolving, value-generating AI agents. This redefines user-AI interaction, moving from passive tools to active, revenue-generating digital entities.
Looking forward, if Virtuals can successfully bootstrap an early community of developers launching agents, we might see more solutions for launching agents integrated with Virtuals or similar platforms. Expect companies like Theoriq, Talus, or other AI agent platforms, even ones like Ritual, to potentially plug into these types of platforms that allow you to speculate on agents and share in the revenues of agents across chains.
The implications are profound. As AI agents become more integrated into our lives, the line between human and artificial intelligence blurs, opening unprecedented opportunities for innovation and interaction. Success in this landscape will hinge on adaptability and foresight. Those who leverage platforms like Virtuals Protocol will shape the future of technology and society.
References
Pump.fun achieves $100M in cumulative revenue amid rising criticism and market challenges. (2024, September 3). SolanaFloor. Retrieved from https://solanafloor.com/news/pump-fun-achieves-100-m-in-cumulative-revenue-amid-rising-criticism-and-market-challenges
Exploring Pump.fun: How to create your memecoins. (2024, July 11). KuCoin. Retrieved from https://www.kucoin.com/learn/web3/exploring-pump-fun-how-to-create-your-memecoins
Publisher exchanges, consumer applications, and the attention theory of value. (2024, March 12). Multicoin Capital. Retrieved from https://multicoin.capital/2024/03/12/publisher-exchanges-consumer-applications-and-the-attention-theory-of-value/
pump.fun. [@hashed_official] (2024). Dune.com. Retrieved from: https://dune.com/hashed_official/pumpdotfun
Pump.fun Analysis | Flipside. [@hrst79] (2024). Flipsidecrypto.xyz. Retrieved from: https://flipsidecrypto.xyz/hrst79/pump.fun-analysis-y4l427
Lin, L. [@ljin18]. (2024, April 11). [Tweets]. X (formerly Twitter). Retrieved from:
The case for AI agents in Web3 gaming. (2024, June 25). Shoal. Retrieved from https://www.shoal.gg/p/the-case-for-ai-agents-in-web3-gaming
Elon Musk's Twitter overhaul could be huge for DOGE and crypto generally. (2023, July 25). CoinDesk. Retrieved from https://www.coindesk.com/business/2023/07/25/elon-musks-twitter-overhaul-could-be-huge-for-doge-and-crypto-generally/
The case for compute DePINs. (2024, June 11). Shoal. Retrieved from https://www.shoal.gg/p/the-case-for-compute-depins
The convergence of AI, DeFi, and intents. (2024, May 17). Shoal. Retrieved from https://www.shoal.gg/p/the-convergence-of-ai-defi-and-intents
Virtuals Whitepaper. (n.d.). Virtuals.io. Retrieved from https://whitepaper.virtuals.io/
Virtuals Protocol. (2024, September). VideoGameGen. Retrieved from https://huggingface.co/virtuals-protocol/mario-videogamegen/blob/main/VideoGameGen.pdf
EverythingEmpty [@everythingempt0]. (n.d.). [Tweet]. X (formerly Twitter). Retrieved from https://x.com/everythingempt0/status/1839592761907163241
Virtual Protocol. (n.d.). Westworld AI. Retrieved fromhttps://virtual-protocol.github.io/westworld-ai/
A big shoutout to David for the edits and Vlad for creating the infographics.
Not financial or tax advice. The purpose of this post is purely educational and should not be considered as investment advice, legal advice, a request to buy or sell any assets, or a suggestion to make any financial decisions. It is not a substitute for tax advice. Please consult with your accountant and conduct your own research.
Disclosures. All posts are the author's own, not the views of their employer. This post has been sponsored by Virtuals Protocol. While Shoal Research has received funding for this initiative, sponsors do not influence the analytical content. At Shoal Research, we aim to ensure all content is objective and independent. Team members purchased a small amount of $LUNA tokens shortly after the official protocol launch in order to participate in and evaluate the v2 product for research purposes. Our internal review processes uphold the highest standards of integrity, and all potential conflicts of interest are disclosed and rigorously managed to maintain the credibility and impartiality of our research.