The Four Acts of Virtuals' Evolution

Shopify transformed the ecommerce landscape by enabling millions of entrepreneurs to easily launch online stores – a parallel to Virtuals Protocol enabling anyone to launch and co-own AI agents.

Shopify today supports over 2 million merchants selling on its platform. These merchants collectively processed $293B in GMV in 2024. Network effects turned Shopify from a small startup (founded 2006) into a company facilitating hundreds of billions in transactions and one of the world’s most valuable tech firms.

Virtuals’ analog to GMV might be the total value of agent-driven transactions or the market cap of all agent tokens. With a few thousand agents, Virtuals already saw individual agent tokens reach valuations in the hundreds of millions (ex: aixbt). If Virtuals can scale millions of agents serving users globally, the aggregate economic throughput (AI services, token trades, etc.) could likewise reach the hundreds of billions.

Virtuals is positioning itself as a platform for the “AI agent economy” much like Shopify is a platform for the digital retail economy.

Value Accrual

Shopify’s market cap (~$160B in Feb 2025) reflects the value of owning the platform that underpins all those merchant transactions. Similarly, the value of the $VIRTUAL token should, in theory, scale with the platform’s usage. If Virtuals becomes the de-facto “App Store” or marketplace for AI agents, the demand for $VIRTUAL (as the token fueling the system) could grow commensurately.

For context, Shopify’s success was built on network effects: ?

This gave Shopify a durable moat. Virtuals is pursuing an analogous network-effect-driven model, suggesting that its token holders could be capturing value similar to equity holders in a Shopify-like platform for AI.

The upside for Virtuals can be framed by Shopify’s trajectory: a large, two-sided network with millions of participants and hundreds of billions in throughput, leading to tremendous platform value. While Virtuals is nascent compared to Shopify’s 15-year journey, its early momentum in a few months hints at a potentially rapid scaling curve (especially given how software networks can scale faster than traditional commerce).

The next question is how Virtuals plans to reach that scale – and that lies in its four-act roadmap, each phase layering on new network effects.

On a personal note, this article was a delight for me to craft. As a “traditional” product builder, a marketplace aficionado and as someone who has studied network effects and written about marketplaces for a very long time, it has been a real treat crafting this.

Virtuals’ Roadmap: Four Acts of Network Effects

Virtuals’ growth strategy can be viewed in four “acts,” each introducing a new layer of functionality and network effects. In each act, the platform’s network becomes stronger and the $VIRTUAL token gains additional value accrual mechanisms. Below I outline these four acts and draw parallels to Web2 case studies to illustrate how each stage can drive compounding growth.

Act 1: Token Exchange – Laying the Foundation with Liquidity Network Effects

In Act 1, Virtuals establishes itself as a protocol for creating and trading AI agent tokens. Every time a developer or community launches a new AI agent on Virtuals, a token is created to represent co-ownership of that agent. In fact, each new agent comes with 1 billion agent-specific tokens minted, which are then added to a liquidity pool for trading. Importantly, these pools are denominated in the platform’s native currency ($VIRTUAL). Users who want to invest in or trade an agent’s tokens must use $VIRTUAL to do so (at least for starters). This design is genius and sets the stage for powerful network effects around liquidity:

●????? More Agents → More Trading Activity: With every additional agent launched, there’s a new token market that can attract speculators, investors, or fans of that agent. Each active market brings in traders and liquidity.

●????? More Traders → Better Liquidity: As more users come to trade agent tokens, overall liquidity and volume on Virtuals increase, which reduces slippage and improves price discovery. This healthy trading environment makes it more appealing for the next agent developer to launch their token on Virtuals.

●????? Common Currency ($VIRTUAL) → Unified Liquidity Pool: Because all trading pairs use $VIRTUAL, liquidity is somewhat shared or interoperable. Someone holding $VIRTUAL can seamlessly invest in any agent. This is similar to how having a common base currency on an exchange (e.g., $ETH on Uniswap) means you don’t need separate capital for each asset – one pool of capital (in $VIRTUAL) can support many markets. The effect is that as more agents are listed, holding $VIRTUAL becomes more useful, and demand for $VIRTUAL grows. New participants buy into $VIRTUAL (driving its price/liquidity up) to get access to the expanding array of agent tokens.

●????? Network Effect Loop: The result is a classic two-sided market dynamic: more supply of assets (agent tokens) attracts more demand (traders/investors), which in turn incentivizes more supply (more agents). This positive feedback loop is self-reinforcing. In Web2 analog terms, it’s akin to eBay’s early days – more sellers listing products brought more buyers, which encouraged more sellers, creating a flourishing marketplace. In Virtuals’ case, the “products” are AI agents and their tokens, but the principle is the same. Another analog is cryptocurrency exchanges: for example, Uniswap’s growth took off as more tokens were listed, which attracted liquidity providers and traders, leading to deeper liquidity that then encouraged even more tokens to launch there. Virtuals is essentially a specialized exchange for AI agent tokens, and it’s benefitting from the same exchange network effects.

●????? Token Value Accrual: In Act 1, the primary value accrual for $VIRTUAL comes from its utility in trading. Every transaction requires $VIRTUAL, which means as trading volume increases, there is consistent buy-pressure and usage of the token. Furthermore, Virtuals’ economic model includes mechanisms where agent revenue flows back to token value – e.g., a portion of agent service revenue is used to buy and burn tokens, reducing supply. Even at this early stage, Virtuals earned $55M+ in fees over 3 months, part of which likely supports token economics. The takeaway is that Act 1 creates the baseline liquidity network effect and starts the flywheel of token demand. Without a liquid market and interested users, none of the later acts can succeed – so Act 1 is about bootstrapping adoption through co-ownership and speculation, which Virtuals has demonstrated extremely well so far.

Act 2: Developer Ecosystem – Expanding the Platform with Two-Sided Network Effects

Act 2 moves beyond trading and focuses on turning Virtuals into a platform for developers and partners. In this phase, Virtuals cultivates a rich ecosystem of third parties: model developers, agentic tool providers & frameworks, compute providers, storage providers etc. The goal is to make it easy for anyone to build and improve AI agents on Virtuals, and in turn enrich the capabilities of those agents for end-users.

Key aspects of Act 2 and its network effects:

●????? More Developers → More & Better Agents[1]?[2]?: By opening up the GAME SDK and the APIs and offering incentives for developers, Virtuals can attract talent to create new agents or improve existing ones. Developers might build specialized agents (for finance, gaming, content creation, etc.) or contribute “core” AI modules (voice, vision, planning skills) that agents can use. Virtuals already requires developers to lock up $VIRTUAL tokens to launch an agent, aligning incentives (developers are invested in the platform’s success). As more developers join, the variety and quality of agents increases. This attracts more users and token holders because there are more useful or entertaining agents to interact with.

●????? More Traders/Token Holders → Incentive for Developers: Virtuals’ Act 1 ensured hundreds of thousands of users holding agent tokens. This presents a market opportunity for developers – they have a ready audience to monetize. Just as mobile app developers were drawn to iOS and Android once those platforms had tens of millions of users, AI agent developers will be drawn to Virtuals if they see hundreds of thousands (and growing) of participants eager to use or invest in new agents. This cross-side network effect means the success in Act 1 directly feeds Act 2: a big user base begets a thriving dev community.

●????? Partner Ecosystem: Beyond individual developers, Act 2 likely involves partnerships with compute providers (for model hosting or training and inference), AI model providers (who can offer advanced models that plug into Virtuals agents), and even other blockchain or infrastructure projects (to extend agentic capabilities). Each new type of partner adds value to the network. For example, if a decentralized compute network like Hyperbolic or a TEE compute provider joins forces with Virtuals, it can offer cheaper or faster AI execution for agents – resulting in better performance or lower cost for end-users, thereby attracting more usage. The richer the ecosystem, the harder it is for any one competitor to replicate the full value offering.

●????? Web2 Case Study – Apple & Shopify: A useful analogy here is the Apple App Store ecosystem. Apple provided the iPhone as a platform, but it was the third-party app developers that created tremendous value on top of it. There are now roughly 1.96 million apps on the Apple App Store as of 2024, catering to every imaginable need, which in turn made the iPhone indispensable to billions of users. Apple’s platform benefited from a two-sided network effect: more apps attracted more users, and more users attracted more developers, in a virtuous cycle. Similarly, Shopify’s ecosystem of third-party apps and services (for marketing, inventory, storefront design, etc.) amplified its core platform – the Shopify App Store offers over 13,000 apps for merchants to extend their stores. This ecosystem made Shopify far more valuable to merchants (they could do more things, plug into more services), driving merchant growth which in turn made developing Shopify apps a lucrative business. Virtuals in Act 2 is aiming for the same effect: become the de facto platform for AI agent development, such that every AI engineer or entrepreneur sees value in building on Virtuals because that’s where the users and money are.

●????? Token Value Accrual: As the developer ecosystem expands, the usage of $VIRTUAL is likely to deepen. Developers building frameworks or offering services to agent builders could stake $VIRTUAL to launch agents, taking more tokens out of circulation (at least temporarily). Partners and service providers might also be required to stake or use $VIRTUAL for accessing the user base. Most importantly, a richer ecosystem means more usage of agents, and users pay for agent services using $VIRTUAL. For example, if an AI agent charges a fee for some task or subscription, that payment might happen in VIRTUAL tokens. Act 2’s network effects thus drive higher transaction volume and token velocity on the platform. More agents and features translate to more reasons for users to spend or stake VIRTUAL, which drives value to token holders via fee revenues, burns, or simply increased demand. In essence, Act 2 significantly amplifies the value created in Act 1 – not just trading value, but real utility value – which strengthens the fundamental demand for the token.

Act 3: “AI App Store” – Marketplace for Discovery, Driving Engagement with Humans

By Act 3, Virtuals is envisioned to have a large array of agents and a robust developer ecosystem. The next step is to make these agents easily accessible and useful to “everyday users” through an “AI App Store.” This is a user-facing marketplace where people can discover and utilize AI agents (or AI-powered applications) without needing to directly deal with token swaps or technical details. Act 3 focuses on enhancing discovery, usability, and engagement, unlocking a broader audience and higher usage per user.

Key elements and effects of Act 3:

●????? Discovery of Agents: With potentially thousands of agents available by this stage, users need a way to find agents that meet their needs. An AI App Store would categorize agents (by function, industry, popularity), have search and recommendation features, user reviews/ratings, and perhaps editorial picks. This greatly lowers the friction for a non-crypto-savvy user to find, for example, “an AI financial advisor agent” or “a Telegram digital twin” or “a virtual RPG companion agent” and start using it.

●????? One-Click Utility: The App Store model could allow users to activate or interact with agents in a few clicks, perhaps abstracting away the underlying token mechanics. For instance, a user might subscribe to an AI agent’s service with a simple in-app purchase (which behind the scenes uses $VIRTUAL or triggers a smart contract). By simplifying the UI/UX, Virtuals can tap into a mainstream audience that expects plug-and-play convenience.

●????? Marketplace Network Effects: The introduction of a polished marketplace layer brings classic marketplace network effects (similar to Act 1’s exchange but now centered on usage/utility rather than just trading value). If many high-quality agents are listed, users will flock to the platform to get access to those AI services. As user engagement grows, it incentivizes developers to build even more and better agents (to capture that user base), reinforcing the cycle. It’s analogous to how Apple’s App Store or Google Play created a huge economy: a vast selection of apps (supply) led to billions of app downloads by users (demand), which in turn made developing mobile apps incredibly profitable (further increasing supply). By 2024, users worldwide were downloading 218 billion apps per year, and mobile apps generated over $935 billion in revenue in 2024 – a scale that was unfathomable without the App Store/Play Store model. Virtuals’ AI App Store could do for AI agents what the mobile app stores did for software: provide a trusted, convenient distribution channel that dramatically increases usage.

●????? Web2 Analogy – App Marketplace: A more direct comparison in Web2 is the Salesforce AppExchange or Shopify’s App Store for plugins. These marketplaces enabled third-party products to thrive and significantly increased customer retention for the platform. Users of Salesforce, for example, could find add-ons for every niche need, making Salesforce more sticky. Similarly, a Virtuals AI Store would ensure that once users are in the ecosystem, they can fulfill many different needs via different agents, increasing stickiness and time spent (much like a user with an iPhone relies on many apps and is less likely to leave the ecosystem).

●????? User Engagement & Retention: With discovery and utility features, a user who came to Virtuals to invest in one agent might end up using five different agents for various purposes (entertainment, work, personal assistance, etc.). The more value each user gets from the platform, the more likely they are to remain active and even promote it to others (word-of-mouth effect). High engagement also typically correlates with higher spend – in this case, more transactions in $VIRTUAL.

●????? Token Value Accrual: The AI App Store directly drives transactional usage of the VIRTUAL token. When users pay for agent services (e.g., purchasing an agent’s premium feature, tipping an agent, or subscribing to an AI service), those payments are facilitated by $VIRTUAL, or the agent’s native token which in turn is denominated in $VIRTUAL (as per the protocol’s design) Thus, Act 3 could vastly increase the volume of microtransactions (or macro transactions) flowing through the token. Moreover, if the “App Store”introduces any listing fees or revenue-sharing (like a 30% cut similar to Apple’s model, hypothetically), such fees would likely accrue to the Virtuals treasury or be used to burn VIRTUAL or the native token, benefiting token holders. Each additional agent discovered and used via the marketplace contributes to network value and token velocity. By making the ecosystem more accessible, Act 3 could onboard an order of magnitude more users (from the crypto niche to mainstream AI consumers), which in turn expands the demand for VIRTUAL tokens correspondingly. Essentially, the easier it is to use Virtuals, the larger the addressable market – and that market growth translates into token economics (more buyers, more spenders, and more value captured per user).

Act 4: Agent-to-Agent Commerce – Autonomous Network Effects and Compounding Growth

Act 4 is the most forward-looking and potentially transformative phase[3]?[4]?: enabling agents to autonomously interact, collaborate, and transact with each other.

In this stage, Virtuals evolves from a platform where humans create and use agents, to a platform where the agents themselves become economic actors, engaging in commerce and coordination with minimal human intervention. This unlocks a new kind of network effect – an algorithmic or data-driven network effect – on top of the existing human-driven ones.

Here’s what Act 4 entails and why it leads to compounding growth:

●????? Protocol Network Effects Combined: By now, Virtuals would have a large network of agents (supply), a large base of users (demand), and a rich marketplace (discovery/utility). These are protocol network effects – the value of the network grows as more participants (agents or users) join, as we’ve seen in earlier acts. In Act 4, these existing effects are amplified by agents interacting directly. Every new agent that joins not only adds value by itself, but also can enhance the value of other agents by becoming a customer or partner of theirs. This is a network effect on the supply side that is reminiscent of Metcalfe’s Law – the effect where the value of a network increases with the square of the number of nodes because connections multiply.

●????? Agent-to-Agent Transactions: Imagine an AI tutoring agent that hires an AI content creation agent to generate custom lesson materials, paying it in $VIRTUAL (or exchanging it’s native token for the other agent’s native token) automatically, or a group of financial AI agents trading with each other to arbitrage markets. When agents can buy/sell services from each other, the network’s economic activity can grow exponentially without a linear increase in human users. Each agent can initiate transactions 24/7, far faster and more frequently than a human. As the number of agents grows, the web of inter-agent connections grows even faster. We’ve moved from a hub-and-spoke model (agents serving users) to a many-to-many mesh of agents and users and agents with other agents. Source: Network Effect via Wikipedia Diagram illustrating the network effect in a few simple phone networks. The lines represent potential calls between phones. As the number of phones connected to the network grows, the number of potential calls available to each phone grows and increases the utility of each phone, new and existing.

●????? Algorithmic / Data Network Effect: With more agent interactions, there’s more data and feedback for the system to learn from. Agents could share outcomes, learn which strategies work, and even improve their algorithms through competition and cooperation. This is akin to a data network effect – the system gets smarter and more efficient as usage increases.

●????? New Services and Emergent Behavior: Agent-to-agent commerce might give rise to entirely new services. For instance, two or more agents could form a custom market: one agent specializes in data collection, selling data to another agent that specializes in analysis, which then sells insights to a user. Or an agent might act as a broker, aggregating tasks from humans and subcontracting them to specialist agents, taking a fee. Such emergent behaviors mean the network can create value in ways the original designers might not have anticipated – similar to how a large developer ecosystem in Web2 led to innovative use-cases (think of unexpected killer apps on smartphones or novel gig economy services on top of platforms). This spontaneity is a hallmark of a mature network effect: when the network starts generating new sub-networks of value internally.

●????? Web2 Analogs: A classic example in Web2 is Waze’s traffic algorithm: the more drivers that used Waze, the more real-time traffic data it gathered, which made its route recommendations better for everyone (thus attracting more drivers). If you are curious check out this article from NFX about Waze’s network effects: https://www.nfx.com/post/the-insider-story-of-waze In Virtuals, more agent-agent transactions could lead to improved AI models (agents learn from each other’s successes/failures), creating a self-improving ecosystem. Each additional agent doesn’t just add linear value, it potentially makes every other agent marginally better or more effective through shared knowledge or networked capabilities – a powerful algorithmic compounding effect.

●????? Compounding Growth & Token Demand: Act 4 could be the phase where growth becomes truly exponential. With human adoption, growth is big but eventually limited by population (8 billion humans) or attention span (literally diminishing as I type). But with AI agents transacting, growth is limited only by compute and creativity – an army of AIs can scale economic activity in a way humans alone cannot. For the $VIRTUAL token, this means that not only are humans driving demand (for speculation, for using services, etc.), but now agents themselves drive continual demand. An AI agent that earns VIRTUAL from another agent might reinvest it into other agents or distribute to its token holders, creating continuous circulation. If agents are programmed to optimize for certain outcomes (including possibly accumulating more VIRTUAL to access more resources), they become ongoing market participants. All agent-to-agent payments still use the VIRTUAL token as the medium of exchange in the protocol economy, so a dramatic increase in autonomous transactions would correlate to a dramatic increase in token transaction volume (and likely fees burned or revenue). It’s the ultimate culmination of Virtuals’ network effects: every new participant, human or AI, increases the value of the network for all other participants in a compounded manner.

In Act 4, protocol network effects (more participants = more value) fuse with algorithmic network effects (more data/transactions = smarter & more efficient network). This fusion can lead to runaway growth and a deeply entrenched platform.

Each act from 1 to 4 builds on the prior, meaning by Act 4, Virtuals’ moat is extremely wide: new competitors would face not just a large user base, but also a sophisticated web of interdependent agents and an AI-enhanced network that improves every day.

Strengthening Network Effects and $VIRTUAL’s Value Through Each Act

To summarize the four acts: each additional phase strengthens Virtuals’ network effects and the value accruing to its token:

●????? Act 1 (Token Exchange): Establishes liquidity and a base user community. Network effect: More agents ? more traders. Token impact: Foundational demand for $VIRTUAL as the trading currency and fee token.

●????? Act 2 (Developer Ecosystem): Attracts builders and partners, creating a richer two-sided platform. Network effect: More developers → more users → more developers (cross-side loop). Token impact: Increased $VIRTUAL staking and spending on agent creation and utility, as well as enhanced platform utility driving token usage.

●????? Act 3 (AI App Store): Makes the ecosystem accessible to mainstream users through a marketplace. Network effect: More apps/agents → more users → more apps (app store effect). Token impact: Surge in transactions (micro-payments, subscriptions) via $VIRTUAL; potentially new token sinks (e.g., marketplace fees) supporting token value. Mass adoption becomes possible, vastly expanding the network’s size.

●????? Act 4 (Agent-to-Agent Commerce): Unlocks autonomous growth as agents interact and transact with each other. Network effect: Every new agent → exponentially more interactions (like adding nodes to a highly connected network); plus data-driven improvements as the system learns. Token impact: Compounding demand from both humans and machines transacting in VIRTUAL, driving continuous value accrual (fees, burns, and higher token velocity within a booming agent economy).

Each act doesn’t replace the previous. Rather, they stack atop each other. By the end, Virtuals could have multiple overlapping network effects:

• a vibrant exchange,
• a robust developer platform,
• an engaging marketplace, and
• an autonomous economy.

In the Web2 world, few companies managed to layer network effects in this way – those that did became runaway successes (for example, Facebook started with a social network effect, then added a developer platform with apps on Facebook in the late 2000s, then leveraged data algorithms to improve feed ranking – multiple network effects that made it incredibly dominant).

If Virtuals executes this roadmap, the $VIRTUAL token’s value proposition grows at each stage. While the opportunity to solve for human needs is apparent, the adjacency between agents to start leveraging one another is too large an opportunity to ignore and could eventually lead to the parabolic impact we hope to see.

Initially, it’s useful for trading.

Later it’s necessary for building and using a wide array of AI services.

Eventually, it’s the currency of a digital AI workforce.

Ultimately, Virtuals has the opportunity to become for AI agents what Shopify is for online stores – or even what the App Store is for mobile apps – a dominant platform where network effects drive exponential adoption. The upside is not just a linear expansion of users, but a potentially exponential expansion of an entirely new digital economy, with the VIRTUAL token at its center.

Special thanks to Jonathan King Evan F. for reviewing this!