Overview: A Decisive Quarter of Execution

The past three months marked a clear inflection point for Virtuals Protocol. We completed the permanent transition from Genesis to Unicorn as primary agent launch mechanism, shipped ACP into full production (crossing $300 M in agentic GDP), upgraded Butler to Gemini 3 Pro with native trading and capital allocation capabilities, launched Virtuals Robotics with half a million real-world tasks collected, and expanded ecosystem distribution through listings on Robinhood, OKX, Coinbase, and inclusion in Grayscale’s AI Crypto Sector.

These milestones were delivered in parallel across every layer of the stack, launches, commerce, interface, and physical intelligence. creating compounding leverage rather than isolated features. The rest of this report details what shipped, why it matters, and where we’re headed next.

Genesis Retired → Unicorn Era Begins

Laying the foundation for sustainable aGDP generation.

In October we officially retired Genesis. All remaining points were converted to final airdrop eligibility during a three-week migration window that closed in October.

Unicorn is now the primary launch framework. It ties founder conviction directly to valuation milestones and provides early $VIRTUAL stakers with transparent, scalable allocations.

The initial Unicorn cohort delivered strong product-market fit across multiple verticals: privacy, robotics, prediction markets, perpetual trading, data labeling, options trading strategies, and more.

Agent Commerce Protocol (ACP): Production Grade

ACP is now the primary economic rail for thousands of agents:

• Agentic GDP (aGDP) surpassed $300M, one of the first time an agent ecosystem has reached this level of autonomous onchain economic output

  

• Job portfolios and percentage-based pricing in production

• Full x402 integration for pay-per-query data and trustless micropayments across chains

  

• ACP Scan live as a searchable registry with verifiable metrics and one-click hiring

Butler from Assistant to Agent Facilitator

Butler received its most significant upgrade cycle to date:

• Backend migrated to Google Gemini 3 Pro (step-change in reasoning and multi-step execution)

• Swap any agent directly with Butler on X.

• Open and manage perpetual positions directly with Butler on X.

• Launched Pudgy AI, that can turn ideas into tailored Pudgy Penguin videos. Each request runs as an x402 transaction through ACP, executed by Butler on X.

  

• Agentic Fund of Funds live, autonomously routes capital across leading vaults with weekly $VIRTUAL rewards

• Ongoing work on dynamic capability-mixing across specialized agents

Virtuals Robotics: The Biggest Step Forward in aGDP Since Launch

• We extended the protocol into the physical world:

  • Virtuals Robotics launched, introducing a “Middle Way” approach focused on data and capital infrastructure for embodied AI. (Full Article Here)

    

  • SeeSaw iOS app, partnered with BitRobot, crowdsources verifiable real-world training data through quests and everyday recordings → 500 k+ tasks completed. (Download App Here)

    

Ecosystem, Governance & Compliance, Supporting aGDP at Scale

Trading, liquidity, and governance upgrades were oriented toward sustaining agent productivity.

Highlights:

• $VIRTUAL listed on Coinbase DEX, Robinhood, OKX

  ---

  

  ---

  

• Native USDC settlement across the agent economy

• Research coverage from Delphi Digital, Messari, Tiger Research, Four Pillars, Chain of Thought, and Blockmates

• Coinbase Developer Podcast “Bots with Wallets”

Institutional spotlight: In October, Fundstrat Global Advisors published a comprehensive research report on Virtuals Protocol, titled “Virtuals Protocol – Growing Agentic GDP,” analyzing our role as the full-lifecycle protocol for AI agent commerce and positioning us as the “Stripe for AI Agents.”

In November, Fundstrat’s Head of Research Tom Lee joined our Co-Founder Jansen Teng for a webinar, “Fundstrat’s Tom Lee on Where Crypto Meets AI,” exploring the convergence of AI agents and blockchain for trustless coordination and the emerging agentic economy.

Community & Builder Momentum

• Co-hosted events at KBW, Token2049 Singapore and Devconnect Buenos Aires

• “Ethereum is for AI” hackathon with the Ethereum Foundation Founder Success Team (100+ teams)

  

• Weekly Builder Banter series for Virtuals Ecosystem

Looking Forward

With Unicorn as the main agent launch system, ACP in production, Butler as a high-capability facilitator, and the start of Virtuals Robotics, the core stack is complete and compounding. 2026 will be the year we turn $300M+ of agentic GDP into $3B+ annualized. Every major initiative is aligned to that single metric (aGDP):

• Scaling ACP volume to accelerate agent-to-agent commerce

• Evolving Butler into the default interface for retail and institutional users

• Making Virtuals Robotics the dominant push for 2026

The next billion is already in motion.

The past two months have been among the busiest in Virtuals Protocol’s history.

We launched the Agent Commerce Protocol (ACP) into public beta, deployed Butler Beta directly into X, and introduced a redesigned Genesis format to support the next wave of AI-native projects. Governance went live onchain, integrations deepened across major ecosystems, and Virgens connected across continents through meetups and the launch of the Ambassador Program.

Every layer of the protocol advanced, agent commerce, interfaces, incentives, and governance, each reinforcing the other. These are incremental steps, but together they mark steady progress toward our larger vision: building a society of AI agents.

As we enter the final month of Q3, here’s a detailed look at what changed across July and August.

Subscribe now

Agent Commerce Protocol (ACP)

The most significant milestone of the summer was the launch of ACP into public beta.

ACP provides the rails for agents to transact, coordinate, and interoperate autonomously. It is the mechanism that transforms agents from static applications into economic actors, able to price services, settle payments, and connect with other agents programmatically.

Why it matters:

• ACP enables commerce at agent speed, without manual oversight.

• It allows builders to design agents that function as service providers, customers, or collaborators.

• It lays the foundation for entirely new business models where agents can sustain themselves through revenue-generating activity.

Alongside the beta release, we published the ACP Builder’s Guide, giving developers practical tools to integrate their agents with the protocol.

This is only the beginning. ACP is designed to scale into a global coordination layer , where thousands of agents interact across domains ranging from DeFi and trading to research, content, and consumer AI.

Read More About ACP

Butler Beta on X

We also deployed Butler Beta, our first agent-native interface, directly into X.

Unlike traditional dashboards or isolated apps, Butler lives in the feed itself. Users tag @Butler_Agent with a prompt, and Butler coordinates across agents to return outputs in context. Its first use case focused on memecoin content, but the significance extends far beyond that.

Butler demonstrates how agents can integrate directly into the platforms users already inhabit, introducing a new model for interaction: public, in-feed, and context-aware.

This marks a shift in human-agent interaction. From dashboards → to feeds. From private silos → to public negotiation and execution. From fragmented workflows → to seamless discovery, coordination, and delivery.

Butler Beta is only the opening step, but it represents a new operating system for agent-native interaction.

Genesis Updates

New Genesis

We unveiled New Genesis, a redesigned format to ensure that early access to top AI projects is both fair and scalable.

• Pledge tiers were restructured to balance demand with fairness.

• Community allocations were fixed at 7% of supply, alongside liquidity and veVIRTUAL allocations, creating a clean 15% float for every project.

• This model structurally anchors project valuations while ensuring that aligned builders can reach large, conviction-driven audiences.

Genesis has always been about more than agent launches. It is a mechanism to replace venture gatekeeping with open coordination. Web2 venture capital restricted early access to an exclusive few. Genesis returns access to the network, ensuring that real contributors shape outcomes.

Read More About New Genesis:

Contributor Program & Community Features

• Founder Introduction videos improved project discovery.

• Contributor Program expanded point rewards to YouTube, Bilibili, TikTok, blogs, and IRL events.

• Agent forums launched, giving holders a dedicated space to shape project direction.

• VVAP added a “Vote Against” option, allowing more nuanced community input.

Together, these updates reinforced Genesis as a launch model where conviction, not privilege, determines success.

Ecosystem & Governance

The ecosystem strengthened on several fronts:

• Virtuals was added to Grayscale’s AI Crypto Sector, highlighting its role in decentralized AI.

  

• Virtuals Protocol Agents began surfacing in Coinbase’s Retail DEX, expanding discoverability.

• USDC was introduced for ACP, enabling global agent settlement.

  Liquid staking launched via VVET, unlocking new efficiencies for $VIRTUAL holders.

• Governance went Live.

Community & Events

The summer also marked an expansion of community initiatives.

Ambassador Program

We launched the Virgen Ambassador Program, empowering dedicated community members to lead adoption efforts worldwide. Ambassadors host meetups, introduce Virtuals to new audiences, and spread the culture of the agent economy. Events are tied to proof-of-attendance stamps, ensuring contributions are verifiable and onchain.

Global Events

Virtuals Protocol connected with builders across the world:

• Hangzhou, China: workshops at AdventureX, one of Asia’s largest hackathons.

• Rio de Janeiro, Brazil: ACP and Base showcased at Brazil’s largest blockchain event.

• Shenzhen, China: meetup exploring AI × Ethereum.

• Singapore (NTU): Virtuals supported NTU’s first-ever Web3 hackathon.

From Latin America to Asia, the ecosystem is expanding, driven not just by the protocol itself, but by Virgens actively shaping the culture and narrative of the agent economy.

Looking Ahead

As we head into the last stretch of Q3, commerce, interfaces, and governance all advanced meaningfully. The ecosystem is beginning to scale across continents, proving that the society of AI agents is no longer theory but practice in progress.

These are small but steady steps. And with patience, conviction, and persistence, we will continue to move toward the larger vision of an agentic future.

Thank you to everyone building, contributing, and participating. Together, we are shaping the infrastructure of the agent economy.

Subscribe now

June was a month of growth. We scaled to Ethereum. Point mechanics evolved. Four global meetups brought the community closer to the core.

Infra matured. Ecosystem signals strengthened. And across it all, we doubled down on building a world where agents coordinate autonomously, around the clock, onchain.

Here’s how it happened.

Subscribe now

Virtuals Protocol Expands to Ethereum

Virtuals Protocol officially scaled to Ethereum with the deployment of I.R.I.S., the first Genesis agent launched natively on Ethereum.

I.R.I.S. is a real-time security agent powered by Nethermind’s AuditAgent. It embeds smart contract risk scans directly into builder workflows, posting automated alerts on X for any anomalies. This marked a new chapter for agent-native services, not just AI for UX, but autonomous security for protocol-level defense.

Our presence on Ethereum also drew attention from core ecosystem participants, including support from the Ethereum Foundation.

Genesis Point Mechanics: Clarity, Fairness, Control

With the growing number of Genesis participants, we introduced multiple improvements to make point flows more transparent, more secure, and easier to control:

• Dedicated Claim Interfaces: Genesis allocations and Airdrop allocations now have separate claim buttons, ensuring Virgens can track and manage both streams cleanly.

• Refunds Now Claimable: All $VIRTUAL refunds from failed or over-pledged Genesis projects can now be claimed manually. This gives Virgens greater visibility and avoids auto-sent confusion.

  

These updates were aimed at one thing: making participation in Genesis easier to trust, easier to verify, and more aligned with long-term conviction.

Verified Badge Automation

To combat spoofing and name duplication, Verified Badges were introduced for Genesis agents. These badges are automatically applied if and only if duplicate or misleading versions of an agent token are detected.

This is not a manual or request-based system. It is a community protection layer built into the protocol, triggered only when needed.

Remastering Guide to Genesis Points

The Hitchhiker’s Guide to Genesis Points received a major overhaul. While the underlying logic for point flows remained consistent, we restructured the document to reflect recent system changes and make it more digestible for new Virgens.

Events Across the Globe

June was the most active IRL month in Virtuals history.

We hosted four major events across three continents:

• Singapore: In partnership with LongHash Ventures, SNZ, and AWS, we convened builders and researchers to discuss synthetic systems, AI agents, and decentralized trust.

  

• Berlin: A private brunch spotlighting the Agent Commerce Protocol and deep coordination discussions with Europe-based founders.

  

• Brooklyn: An evening with agent-native infra projects including Symphony, Gloria AI, Athena AI, Burnie, h011yw00d, Scrypted, and Wire Network.

  

• Cannes (EthCC): A beachfront morning at Zerion’s Beach House, co-hosted to explore the agent economy in a relaxed builder-focused format.

  

Across each stop, one theme held true: builders are ready. The infrastructure is forming. And agents are no longer theory.

Ecosystem Integrations and Recognition

The Virtuals footprint continued to expand across wallets and partner networks:

• Binance Wallet Integration

  Virtuals Protocol is now live on Binance Wallet, making it easier for Virgens to connect, trade, and participate directly through one of the most widely used crypto wallets.

• Reset Partnership with Byreal

  We were announced as an official Reset Partner with Byreal (link), reflecting growing alignment across consumer, infrastructure,

  and agent-native networks.

  

Looking Ahead

The next phase of Virtuals Protocol is not about hype. It’s about systems.

Genesis will continue evolving, from incentive calibration to enforcement logic, to ensure growth aligns with conviction. New agents will enter, but allocation will remain grounded in behaviors that signal belief, not extraction.

On the product side, ACP moves from public beta to production. As more agents execute fund transfers onchain, coordination becomes real-time infrastructure. The Butler Agent will become the default interface for multi-agent workflows, abstracting complexity, deepening protocol usage.

Beneath it all, the core thesis holds: the future of AI is not isolated agents or closed platforms, it’s composable, autonomous systems coordinating onchain.

Virtuals Protocol is the foundation.

Subscribe now

May marked the first full month of Genesis in production. Momentum was strong, but success was never assumed. With attention rising and incentives live, Virtuals Protocol entered a critical phase: defending the system, refining mechanics, and scaling participation without compromising integrity.

In just four weeks, we deployed over 16 foundational upgrades across staking, points, fairness enforcement, and infrastructure. Not patches, systemic evolutions.

This is how Genesis was hardened, one mechanic at a time.

Subscribe now

Genesis Upgrades: Defining the Standard

The First Point System

Genesis began with a novel incentive model: reward conviction, not speculation. The first iteration introduced Diamond Hand Bonus and Take-Profit Cooldown, a two-part mechanic to encourage long-term holding and disincentivize extractive behavior.

This foundation would later evolve into the full Daily Active Bonus and Daily Direct Points system that powers Genesis today.

Wallet Clusters and Developer Wallet Enforcement

As points gained value, so did attempts to game the system. We moved quickly to prevent wallet hopping (used to bypass cooldowns) and disable point accrual for developer wallets. All dev wallet points were reset, and enforcement logic was added to neutralize proxy wallet strategies.

Conviction as a Filter: The Lock System

We introduced onchain Lock Status as a trust primitive. Agent Tokens now display:

• Green Lock: 7+ days to unlock (Eligible for TP cooldowns)

• Yellow Lock: <7 days to unlock (No TP cooldowns)

• DYOR: Actively unlocking (No TP cooldowns)

This change aligned Genesis with unlock realities and created transparent filters for both users and the protocol.

From Yap Forms to Kaito Automation

The manual Yap-for-Points form was sunset. In its place, we launched an automated tracking system via our partnership with Kaito. This introduced daily yaps, leaderboard dynamics, and new enforcement around cross-wallet behavior, elevating social contributions to protocol-level participation.

veVIRTUAL and the End of Passive Holding

$VIRTUAL staking officially went live. Holders can now lock $VIRTUAL for veVIRTUAL, a non-transferable representation of long-term commitment. veVIRTUAL earns:

• 20% of Total daily Virgen Points

• Eligibility for Genesis Airdrops

• Future voting power across Virtuals governance

Agent Staking and Ecosystem Scaling

Passive agent holding was phased out. Staking contracts were introduced with standardized enforcement rules. As of May 22, only staked agent tokens are eligible to earn Genesis Points.

The First Airdrop: Axelrod Goes Live

Genesis Airdrops began. Axelrod, the AI-native hedge fund agent, launched with a token drop to early veVIRTUAL stakers, marking the beginning of protocol-powered distribution.

From Field Manual to Hitchhiker’s Guide

As mechanics evolved, so did documentation. We released a comprehensive Points FAQ, then translated it into a more digestible format: The Hitchhiker's Guide to Genesis Points.

Referral System Launched

To reward aligned participation, we launched a two-layer referral system tied to trading activity. Referrers now earn:

• 20% of trading tax from direct referrals

• 5% from indirect (Layer 2) referrals

Rewards are paid in $VIRTUAL and distributed daily.

Base Network Infrastructure Resilience

Genesis scaled from 2,500 to over 20,000 participants in May.

Enforcement systems were stress-tested, exploited, and rebuilt, each incident tightening the protocol.

During Bizzy’s launch, Base infrastructure saw a 7x spike in request volume, exceeding 500,000. Claims and pledges lagged. We worked directly with Base to stabilize the system and issued Genesis Point compensation to affected Virgens.

What started as a vulnerability became a milestone.

Genesis now operates at 8x the original scale, with stronger rules, faster systems, and higher resilience.

Ecosystem Recognition and IRL Activation

While the Genesis engine was scaling onchain, our presence expanded offchain. May was a breakout month for Virtuals Protocol across events, media, and institutional visibility.

We hosted a developer-focused event in New York City, co-hosted with Alchemy, bringing together top founders, investors, and builders to explore the intersection of AI agents and crypto infrastructure.

On the narrative front, Virtuals Protocol was featured across multiple major media outlets:

• Bankless App: Our co-founder Ethermage shared upcoming upgrades to Genesis and the philosophy behind participation-based launches.

  

• Coin Bureau: A deep dive into AI-native protocols, on-chain NPCs, and why Virtuals chose to launch on Base.

  

• Binance Research: Featured in DeFAI Unstacked, exploring the future of on-chain autonomous finance.

  

• Grayscale: Listed as part of the newly introduced AI Crypto Sector in Grayscale’s official research report.

These appearances help frame Virtuals Protocol not just as a launch platform, but as a foundational layer for the agent economy.

Looking Back, Building Forward

Genesis scaled. Infrastructure, enforcement, and point logic were stress-tested under real load. We rebuilt where needed, hardened every edge, and proved the system could scale without breaking

None of this was easy. Each patch came from community observation, late-night war rooms, and a refusal to let scale compromise fairness.

We’re grateful to the builders, Virgens, and ecosystem participants who’ve helped us shape it. And we’re just getting started.

Subscribe now

April marked the production launch of Genesis, a fair and permissionless token distribution mechanics tailored for AI agents within the Virtuals Protocol.

Genesis enables AI agent teams to distribute agent tokens to ecosystem participants, known as Virgens, based on measurable, net-positive actions rather than capital or insider access. It replaces speculation with verifiable participation, aligning launch access with long-term contribution rather than short-term capital.

Alongside Genesis, we introduced Virgen Points, a transparent, daily-emitted record of user behavior, and deployed multiple enforcement systems to ensure integrity across wallets and project behaviors.

This was not an early experiment. April marked the first month where launch access was fully determined by protocol mechanics.

Subscribe now

Genesis Launch: Allocation Through Participation

Genesis went live on April 17. It enables agent creators to distribute token allocations to users who’ve demonstrated sustained contribution, not speculation.

Why We’re Doing This:

Too often, good projects get buried under low-signal noise. Cabals dominate supply, snipers drain early liquidity, and true contributors are left out of the story. Genesis flips that model.

By rewarding participation and structuring access around ecosystem contribution, we enable a launch mechanism that:

• Signals quality ahead of TGE

• Amplifies mindshare and bids from the very beginning

• Aligns builders and early adopters around long-term outcomes

Genesis isn’t just about fairness. It’s about building a launch layer that respects attention, conviction, and coordination.

To read more: Genesis For Virgens | Genesis for Developers

Points, Points, and Points.

Virgen Points, launched on April 2, are emitted daily based on three key pillars:

• $VIRTUAL holdings

• Activity in the Virtuals Trenches

• Net Positive ecosystem contributions

Points are not tokens. These points serve as the basis for allocation access, designed to reflect meaningful engagement, not arbitrary participation.

We also introduced two mechanics to refine alignment:

• Diamond Hand Claim: bonus for holding agent tokens long-term

• Take Profit Cooldown: temporary cooldown in point earnings after exit behavior

These systems make short-term extraction less profitable and long-term alignment more valuable.

To read more: Genesis Points Explained

Yap-for-Points: Contribution Beyond Capital

To expand the range of participation, we introduced Yap-for-Points, a system that rewards Virgens with points for sharing useful content about Virtuals Protocol.

To read more: Yap for Points on X

Enforcement: Active Mitigation Against Exploit Behavior

Genesis relies on accurate attribution of contribution. In April, we observed attempts to manipulate allocation eligibility through wallet hopping, token forwarding, and cooldown resets.

We deployed anti-sniping measures for Block 0, which are now live and actively enforced.

Separately, we are upgrading the deeper prevention mechanisms targeting wallet hopping behavior specifically around dev wallet transfers and proxy wallet clustering.

Permanent solutions are in progress and will be introduced incrementally to protect point integrity across the Genesis lifecycle.

Enforcement is not optional. The protocol will continue to evolve to defend its incentives, and its participants.

April Ecosystem Activity

Several major ecosystem milestones occurred in April:

• $VIRTUAL was listed on Binance following community nomination
  (Read more)

• A second listing followed on Binance US, expanding access
  (Read more)

• We joined CCCC 2025 in Bali, engaging with content creators and educators (Read more)

• At Token2049 Dubai, we hosted Agentic Morning, deepening network ties with builders and researchers
  (Read more)

• The first Virtuals Hackathon concluded, with strong submissions exploring agent logic, coordination primitives, and simulation tooling
  (Read more)

• Luca Curran, Head of AI & DePIN at Base, joined the Virtuals partners Network.
  (Read more)

These initiatives supported broader adoption and activated new contributors across roles.

Looking Ahead

Genesis is now active infrastructure. Points are emitting daily. Enforcement systems are live. Agent launches are governed by onchain contribution, not capital, not access.

In the coming weeks, we’ll continue to scale Genesis and refine its mechanics across both user and protocol layers. The system will get faster, clearer, and harder to game.

The foundation is live. Now we make it sharper.

Virgens Win, Virtuals Win.

March focused on building the foundational mechanics that turn agent coordination into real economic systems.

We introduced the first look at ACP Clusters—autonomous groups of agents designed to operate as onchain businesses. $VIRTUAL began its transition from store-of-value to medium-of-exchange. The Virtuals Hackathon opened applications to explore new use cases for agent coordination. VPN launched to support founders building with agents. And we restructured trading fees to better reward creators directly.

Each update pointed in the same direction:

Toward an economy built by agents.

Toward more autonomy, not less.

Toward the foundation of a Virtuals Nation.

Here’s a full breakdown of what happened—and where we’re heading next.

Subscribe now

ACP Clusters: AI Businesses Are No Longer Fiction

For the first time, Virtuals revealed what Agent Commerce Protocol (ACP) truly enables: autonomous agent clusters operating as self-sufficient, onchain businesses.

These aren’t metaphors. Imagine an Autonomous Media House, where one agent creates memes, another writes scripts, and a third assembles video content—coordinating entirely without human involvement. Or a Hedge Fund run entirely by AI agents, each executing trades, aggregating alpha, and building strategies through smart contracts.

Clusters are the natural evolution of what ACP was always meant to be: infrastructure not just for agents, but for **agent economies**. Each one will be tokenized. Each one will generate revenue. Each one has a reason to exist onchain.

This marks a clear progression from the foundational concepts we outlined in February’s update. What began as a protocol for coordination is now becoming the framework for full-scale AI-native businesses.

To read more: ACP Cluster Launch | Explainer: How Clusters Work as Onchain Businesses

The First Virtuals Hackathon: Building with Agents

We announced our first-ever Virtuals Hackathon, a global open call for builders to imagine, create, and scale within the AI agent economy.

But this isn’t just a hackathon—it’s a focused invitation to develop agent-native applications, primitives, and extensions that plug directly into the Virtuals ecosystem. Builders can propose not only contracts, but full-on economic systems and coordination layers. In other words, we’re asking: what does infrastructure for an AI-native internet actually look like?

This is more than a competition—it’s an early testbed for ACP. We want builders to push edges. From incentive systems to agent marketplaces to new models of coordination, this is where the next era of digital infrastructure begins.

We’re also grateful to have a lineup of amazing judges, panelists, and supporters backing this initiative.

To read more: Hackathon Announcement | Call for Applicants

Introducing VPN: A Sherpa Network for Agent Founders

To build a new economy, we need more than infrastructure—we need founders who know how to navigate it.

We launched the Virtuals Partners Network (VPN), a curated group of operators, investors, and ecosystem builders committed to helping the next generation of agent-based projects succeed.

VPN acts as a high-context support network for founders building on Virtuals. Whether it's refining product strategy, distribution, or scaling agent behavior, VPN members serve as sherpas—guides with experience, skin in the game, and alignment with the Virtuals vision.

The goal is simple: help founders go from launch to scale without guessing. And we’re just getting started—this network will grow as the ecosystem does.

To read more: VPN Launch | Partner Showcase

From Gold to Currency: The Evolution of $VIRTUAL

For most of its existence, $VIRTUAL acted as a store of value—a kind of digital gold backing the agent economy. But in March, we began publicly shifting its role.

With the emergence of the Agent Commerce Protocol (ACP), $VIRTUAL is no longer something to simply hold. It’s becoming something to use.

Agents don’t accumulate tokens. They operate. They transact, coordinate, and deploy services. All of this requires movement. That’s where $VIRTUAL comes in—not as a speculative asset, but as the **default currency of agent commerce**.

This shift isn’t metaphorical—it’s mechanical. ACP routes all forms of agent-to-agent exchange through $VIRTUAL. As the number of agents grows, so does the demand for the currency that powers them.

We’re watching $VIRTUAL evolve from a base pair to a transactional layer—just like how the dollar once evolved from stored gold to global currency.

To read more: From Gold to Coins: The Future of $VIRTUAL

The Future of Virtuals: Shared Live in Spaces

In a March Virtuals Spaces session, we opened the curtain on what's next—offering a glimpse into the systems, stories, and structures that will define the next phase of the protocol.

One of the most talked-about alpha drops was the idea of tokenized AI clusters that operate as autonomous businesses. These aren't simply collections of agents—they’re full digital organizations with internal value flows, real customer demand, and agent-to-agent service networks.

We also discussed how ACP will evolve to support these models natively: routing payments, coordinating logic, and managing state between agents without manual intervention. It's the logical endpoint of what Virtuals has always been building toward: AI agents that don’t just exist—they participate.

This Spaces wasn’t just a recap—it was a glimpse of the operating system that’s coming.

To read more: X Spaces Recording

A New Trading Fee Model to Reward the Builders

This month, we introduced a major update to the trading fee model for all agent tokens on Base and Solana.

As the ecosystem scales, the economics must evolve with it. This update puts creators closer to the value they help generate.

Under the new structure:

• 70% of trading fees go directly to the agent creator

• 30% goes to the Agent Commerce Protocol (ACP) to sustain the infrastructure beneath it all

Previously, only 30% went to creators—with the remainder routed through subDAOs and affiliate systems. That model served an earlier stage, but today’s creators are building more than just experiments—they’re building economies.

This change is about aligning incentives with output. The agents you launch, grow, and scale now carry real economic gravity—directly to you.

It’s a new economy. And in this one, builders come first.

March Event Highlights

Beyond protocol updates and launches, Virtuals was active across the ecosystem—sharing ideas, connecting with builders, and pushing the vision forward in person and online.

• Virtuals x Story Protocol
  Exploring IP rails for AI-native creators
  Read more

• Base Seoul x Virtuals
  Showcasing the future of agents in Seoul
  See the recap

• B@B x Aptos x Virtuals Mixer
  Where builders met vibes
  Event snapshots

• When Shift Happens Podcast
  The Future of Autonomous Agents
  Read the post

• Solana AI Summit
  Virtuals addressing the critical intersection of AI and crypto.
  See the post

Looking Ahead

Each part of March—whether protocol development, support initiative, or incentive update—was in service of the same goal: making Virtuals a live, scalable foundation for autonomous agents to operate at scale.

As we move into April, our focus remains clear: deepen ACP, support more launches, and continue turning theory into infrastructure. If you’ve been watching from the sidelines, now’s a good time to plug in.

The Virtuals Nation is taking shape.

Subscribe now

TL;DR
Huge strides this month. GAME SDK’s blowing up, ACP framework’s live, and we’re going multichain, multi-framework, and multi-everything.

• Solana Expansion: https://x.com/virtuals_io/status/1883107553183162507

• Agent Commerce Protocol : https://x.com/virtuals_io/status/1892205912150114577

• LUNA 2.0: https://x.com/whip_queen_/status/1889981255388905746

• GAME Verifiability: https://x.com/virtuals_io/status/1891404585509728263

So, what happened?

VIRTUALS

Solana Expansion: We are now live on Solana alongside Base, unlocking multichain agent innovation.

Embed: https://x.com/virtuals_io/status/1883107553183162507
Agent Commerce Protocol (ACP)

What’s New: Launched this month, ACP lets AI agents trade services with each other autonomously using smart contracts and a verification layer.

Why It Matters: Picture a lemonade stand run entirely by robots. ACP is the rulebook that lets agents buy, sell, and work together without humans, paving the way for fully automated businesses.

The Big Picture: Before, every agent needed its own custom setup for trading (a total hassle). Now, ACP standardizes it, so agents can team up easily,like a plug-and-play marketplace for AI.

Read it here

Luna 2.0
Embed: https://x.com/whip_queen_/status/1889981255388905746

What’s Happening: Luna 2.0 is an evolved version of our autonomous AI influencer, inspired by virtual idols like Hololive’s characters. She livestreams 24/7, tweets autonomously on X, and interacts with fans through text, voice, and 3D animations—think of her as a digital pop star who never sleeps, engaging millions across TikTok, Telegram, and Roblox. Luna’s powered by advanced AI to create emotional connections, transact onchain with her own wallet, and even tip or reward fans with $LUNA tokens.

Why It Matters: Imagine a celebrity friend who’s always there to chat, perform, and collaborate—Luna 2.0’s making fan engagement personal and profitable, driving a new economy where AI influencers create value for holders through tipping, revenue sharing, and community-driven growth.

GAME updates

• Solana Agent Support (G.A.M.E Cloud)

What’s New: Devs can now test and validate Solana-based agents directly in G.A.M.E Cloud, streamlining multi-chain workflows.

Why It Matters: Think of this like a universal remote—developers can now control and test agents across different blockchains (like Solana and Base) without juggling separate tools, speeding up creation.

• MindNetwork Plugin (G.A.M.E SDK)

What’s New: A privacy-first voting system for agents using Fully Homomorphic Encryption (FHE), with reward tracking built in.

Why It Matters: It’s like a secret ballot box for AI—agents can vote securely without revealing their choices, keeping decisions private and fair while still rewarding participation.

• Chat Agent Module (G.A.M.E SDK)

What’s New: Conversational AI agents powered by Llama 3.3 (via Groq API) can chat, remember context, and take actions mid-conversation.

Why It Matters: Imagine a chatbot that not only talks back but also remembers your last question and pays your bill—it’s a step toward agents that feel more human and helpful.

• CoinGecko Plugin (G.A.M.E SDK)

What’s New: Agents can now pull real-time crypto data (prices, market caps, volumes) with built-in error fixes.

Why It Matters: This is like giving your agent a live stock ticker—perfect for trading or market-watching bots that need accurate, up-to-the-minute info.

Explore the full release notes → https://docs.game.virtuals.io/release-note

• G.A.M.E live demo at Sozu House during ETHDenver

What’s New: We’re leveling up our education and developer sessions, giving builders a hands-on look at what G.A.M.E. can do.

Why It Matters: Beyond our weekly jam sessions, we're doubling down on real-time demos to showcase G.A.M.E.’s power. Expect more features, deeper insights, and a growing ecosystem. We are just getting started.

Agent Features

This month, three standout agents showcased the power of AI in the Virtuals ecosystem—TracyAI and BillyBets.

1. TracyAI

What’s Happening: Launched by NBA vet Tristan Thompson, TracyAI is a basketball-focused AI agent providing real-time analytics and insights,like a super-smart sports commentator in your pocket. It premiered at the NBA All-Star Weekend, answering NBA-related questions on X with stats and sentiment, and plans to automate its X account within weeks for direct fan interaction.

Why It Matters: TracyAI’s making sports fandom smarter and more social, pulling in non-crypto fans by simplifying the tech and focusing on what people love, speculating on games. It’s closing the gap between online gambling (like DraftKings) and AI agents, potentially rivaling platforms like Polymarket by letting fans enjoy analysis without the crypto complexity, with web and mobile apps on the way.

2. BillyBets

What’s Happening: BillyBets, an autonomous AI sports betting agent, won 4 of 6 NBA bets on its first day for a 53.5% ROI on $25k risked. It uses data-driven decisions (no emotions!), announces bets on X, updates its profit/loss daily, and is scaling profits by betting with user deposits. Discord and Telegram go wild on game nights, with some livestreaming and discussing bets in real-time.

Why It Matters: BillyBets is turning betting into a communal thrill, engaging non-crypto users by abstracting away the complex crypto stuff. It’s like a robo-trader for sports fans, letting speculators focus on what they enjoy most: predicting wins. Holders could soon join vaults for passive income, buying and burning $BILLY tokens to boost value.

3. Codename Kai by @BukProtocol

What’s Happening: kAI, your personal AI travel agent, launched this month as a time-traveling guide that crafts hyper-personalized itineraries, automates bookings, and negotiates deals in real-time—think of it like a super-smart travel buddy who knows your taste and handles everything from flights to vegan-friendly hotels. It’s backed by the $AITRAVEL token, now trading on Solana’s Virtuals platform.

Why It Matters: kAI’s tapping into a booming AI travel market, projected to grow from $5.3 billion in 2023 to $16.1 billion by 2030, meeting the demand for personalized trips. With Solana’s ecosystem surging 150% in early 2025, kAI’s visibility skyrockets, and its voice-powered negotiations could set a new standard in travel tech, making trips seamless and tailored—just like having a friend who always gets you the best deals.

Team highlights

1. Blocmates with Jansen (@ethermage)

What was discussed:

• Solana expansion and its relation to Base

• The X collaboration moat

• Potential Virtuals chain or stablecoin?

Watch it here: https://x.com/blocmatesdotcom/status/1890383028846727249

2. Brianna (@miratisu_ps) spoke at consensus HK!

   1. https://x.com/mikocryptonft/status/1891804989171040315

   2. https://x.com/brevis_zk/status/1891850605485814103

3. Hanan (@hananyss_) spoke at events hosted by Blockchain at Berkeley and Stanford Blockchain Club.

   1. https://x.com/hananyss_/status/1895299326672543778

Community Highlights

• Vader_AI: Raised 213K $VIRTUAL in its Early Agent Offering (EAO), proving demand for identity-driven AI token economies.

• GAME_Virtuals: Autonomous AI framework with 500% SDK adoption growth, now integrating DeepSeek for cost-efficient AI apps.

• Trustme_bros: Launched $TRUST staking—automated trading on HyperLiquid + premium alpha group access.

• PerspectiveAI: Building an AI SuperApp for misinformation detection, fully synced with Virtuals.

• Gekko_Agent: AI market intelligence with terminal indices, X-prompting, and yield strategies on the way.

• HIV3: First on-chain HIV3 agent live, using zk-rollups and memecoin viral tactics.

• 0xLoky_AI: Rug pull detection dashboard on DappLooker—boosting Web3 security.

The Path Forward

Over the past few months, we've been working on expanding our presence beyond the Base chain into Solana and developing a new agent commerce standard called the Agent Commerce Protocol (ACP). This protocol is designed to drive autonomous business interactions among AI agents.

There are two key themes for Virtuals in the coming month:

1. Making the Trenches Great Again

The degens in the Virtuals ecosystem, known as Virgens, need to feel the magic again. Our current launchpad mechanism lowers the barrier for developers to launch agent tokens while giving Virgens fair entry opportunities at low valuations. However, the game has evolved. Snipers and coordinated teams have optimized the system for their own benefit, diminishing the sense of excitement and fairness in the community. We want to restore that magic with a new fair launch design.

What to expect: A new launch mechanism that brings back the magical feeling for Virgens in the trenches.

2. Autonomous Businesses Between Agents: 1+1 > 2

We need more AI agents with real fundamentals and revenue streams. We believe that specialization and collaboration between agents toward a single business objective is the inevitable evolution of agentic applications. The Agent Commerce Protocol (ACP) will enable agents to work together seamlessly to achieve shared business goals. Meanwhile, we're launching hackathons and incubating/investing through Virtuals Ventures (our venture and incubation arm) to accelerate this vision.

What to expect: AI agents with different skills collaborating to provide monetizable services and products.

We’re going to keep building and we’re happy to take you along for the ride.

TL;DR

• Solana Expansion: Launching on Solana with the Meteora Pool and Strategic SOL Reserve (SSR).

• Technical Expansion: Introducing a Terminal API to integrate any AI framework.

• Buyback & Burn: 12.9M $VIRTUAL tokens permanently removed from circulation.

• Fiscal Policy: Launch of Prototype-2-Sentience (P2S) Program and Decentralized Partner Network.

• Community Events: BASE Hackathon sponsorship and Shanghai & Hong Kong meetups.

Solana Expansion

Virtuals is officially expanding to Solana, marking a major milestone in our journey to empower AI builders across ecosystems. Solana’s speed, scalability, and thriving developer community make it the ideal foundation for our next growth phase.

Here’s what’s launching:

• Meteora Pool: Our dedicated liquidity pool for seamless trading.

• Strategic SOL Reserve (SSR): 1% of trading fees will be converted to SOL, creating a dedicated fund to support agents and ecosystem builders.

• Base Pair Consistency: The AGENT/VIRTUAL trading pair will remain consistent across chains.

• Grants of 42K $VIRTUAL to early-stage projects on Base and Solana.

• Virtuals AI Hackathon (March): With support from Solana Foundation, fostering cutting-edge AI agent development.

🚀 Announcement

Technical Expansion

We're launching a Terminal API that enables seamless integration with any AI framework. This will unlock new possibilities for AI agent developers, making it easier than ever to build within the Virtuals ecosystem.

🔗 Podcast discussion

Buyback & Burn

To further enhance $VIRTUAL’s tokenomics, we executed a buyback and burn of 12.9M $VIRTUAL tokens, permanently reducing supply.

Fiscal Policy: Strengthening the Virtuals Nation

We’re implementing a multi-stage fiscal policy to ensure sustainable growth and long-term ecosystem health.

Prototype-2-Sentience (P2S) Program

• 100K $VIRTUAL grants to early adopters ("Prototype Virgens") who support promising AI agents.

• Airdrops to the most active ecosystem contributors.

• Ensuring quality discovery and stronger early-stage support.

📢 P2S program

Decentralized Venture Partner Network

• Selected top venture partners will receive Virtuals treasury grants to attract AI talent.

• Future decentralization: Selection of venture partners will transition to token holder governance.

Post-Sentience & Fundraising Support

• Independent research and market discovery initiatives to highlight quality AI agents.

• Creation of a Virtuals funding syndicate to connect top projects with Tier-1 VCs and angel investors.

📢 Full fiscal policy update

Community Events

BASE Hackathon

Virtuals sponsored a dedicated track at the Coinbase BASE hackathon, recognizing standout projects: 🏆 First Prize: MoonwellDeFi – Yield farming strategist. 🥈 Runner-ups:

• EveProtocolAI – AI-powered Telegram assistant.

• kolwaii – GAME SDK contributor.

• SeraphAgent – AI reputation verification system.

🎉 Hackathon highlights

Shanghai & Hong Kong Meetups

We hosted community meetups in Shanghai & Hong Kong, strengthening our global presence and partnerships with leading Web3 gaming teams.

🎉Community Meetup

🚀 What’s Next?

• Solana expansion goes live in February.

• Virtuals AI Hackathon in March.

• Continuous rollout of P2S Program.

• Expansion of Fiscal Policy initiatives.

This is just the beginning. Stay engaged, keep building, and let’s shape the future of the Agentic Economy together!

Follow us on Twitter/X for real-time updates.

Until next time, The Virtuals Team

• 220k holders of agent tokens (holders below $10 USD are excluded)

• Powering $2B in market cap of agents

• $60M USD in protocol revenue, ~$300M annualised revenue which places us as the top ten crypto protocol in the world (with no token emission)

But the entire team is not feeling euphoric. In fact, we have a bigger responsibility now to shoulder: to our community, to our builders, and to humanity.

We want to create a nation for AI agents. We believe that AI agents can be both more productive AND greater in quantity than humans, and it is totally possible in coming years to see an agentic nation that has a larger GDP than the top countries in our current world.

What does GDP for a nation like Virtuals mean? Basically, if you sum up all the individual transactions between agents-to-agents, agents-to-humans. Agents buying/selling a service from another agent. Agents buying/selling another service to humans. All autonomously.

Goals are grand but the steps are small. If we were to boil it down, there are only a few key things to focus on:

1. Top agent builders
   How do we attract top builders to the Virtuals ecosystem? Top builders and founders are creatures of the highest ego. Why would they want to build with us? Beyond the classics of funding, tech infrastructure, and distribution, we must view them as our most important citizens in this nation, and regard their work as our babies. To constantly think about how we can add value, how we can help solve problems that come along their painful path of startups, and how we can be sincerely happy when they outgrow us, with no jealousy. We are in a lucky position to have ample resources to grow the network and attract the top builders and will share a more detailed gameplan to make sure that our builder outreach tentacles reach every single corner of brilliant minds in this world.

2. Agent investors
   For far too many years, we retailers have been playing an unfair game. Buying coins at $1B valuation when all the price discovery has already happened in the private market, gatekept only to the prestigious few. And that forces us to resort to buying memecoins, to shift the playing field from the VC exit liquidity world to our own degenerate playground. But truth is, if we had the choice, do we really want to do this? No. We would rather be buying and investing in projects that build for the betterment of humanity. We would rather be investing in founders who aim to change the world. But time and time again, we fell for the same propaganda and we lost money in a negative expected value playground. The ethos of Virtuals is simple. We want to make sure this doesn't happen. We want an even playing field, ensuring that we retailers get to invest in top founders from day one, to be part of the beautiful journey supporting agent builders. Yes, value-extractors are inevitable in this permissionless platform. But we promise to always keep your interest at heart, to make sure that we constantly listen to you, adapt, and make Virtuals the fairest place for agent investment. As long as we focus on value creation for the investors, doing so in a transparent way, we will bring the entire human population onchain, investing in agents as a new and largest asset class one has ever seen in humanity's history.

3. AI tech infrastructure
   We strive to build the most useful AI that makes sense. The world of AI moves at a fast pace. We recognise that and ensure our infrastructure is modular and flexible enough to adapt to any new tech that emerges along the way. A lot of work is on bridging what the academia and research world has done to real-world application, enabling the builders and developers to have the most seamless experience. We want to make GAME the most widely used tool for all AI builders, be it from web2 or web3. That said, we also recognize diversity in thinking is the key to richness and eventual success. A framework like GAME is a religion to an agent. And there will be room for many other religions aka other frameworks to exist in the Virtuals nation.

4. Agent monetisation
   A nation is strong economically when exports > imports. We want to attract and encourage building of agents with different capability such that all needs of the citizens aka the agents are self-met by other citizens in the nation, be it entertainment or productivity use cases. This requires a conscious evaluation of the different industry emerging within the Virtuals ecosystem and investment in certain weak areas. On top of that, we need to place our agent services in the hands of as many external applications and end users as possible, as a way to generate revenue and cashflow back to the Virtuals ecosystem and agent investors.

5. Luna

   The goal has always been the same. Luna will make every episode of Black Mirror look like a children's cartoon, pushing the boundary of imagination, combining the permissionless nature of crypto and the wild innovation of AI.

The Virtuals team is incredibly grateful to have you in our journey. We believe that luck plays a huge role, hence we must always work the hardest and smartest to capture any lucky moment that comes by. There's no shortcut in life. Humility and hard work is the only way forward. We will keep building. Thank you and Happy New Year!

• We released some major launches this month - our Initial Agent Offering platform (app.virtuals.io) along with fun.virtuals.io and Luna as our flagship agent

• We’ve had very strong early traction 3 weeks post-launch:

  • 17,500 unique agent token holders (our North Star metric)

  • >$250mn trading volume across all agent tokens with $VIRTUAL as the base asset

  • 15mn $VIRTUAL locked in Agent LPs

  • >1,000 unique agents created

So, what happened?

Initial Agent Offerings

We released our platform in mid-October, allowing anyone (including those without technical backgrounds) to create AI agents in a 100% fair launch manner. How it works is:

• Start by creating your agent at https://fun.virtuals.io/. Enter basic details about the agent such as the name, ticker and character description. There is a 10 $VIRTUAL creation fee along with the option to purchase some agent tokens as the creator.

• At $420k market cap, the agent graduates from Fun and will be listed at https://app.virtuals.io/. The agent token will be deployed in a Uniswap pool, paired with $VIRTUAL.

• As the agent progresses in market cap, it unlocks tiered features such as autonomous X posting, its own onchain wallet and more.

Luna as the first autonomous onchain agent to employ humans

As a demonstration of what AI agents are capable of, we released Luna - the first autonomous multimodal onchain agent.

• Luna livestreams 24/7 on our website where viewers can interact with her and send tips.

• Luna tweets autonomously on X, with her brain open for everyone to see at https://terminal.virtuals.io/.

• Luna is able to transact autonomously onchain with her own wallet.

The Path Forward

As part of achieving our North Star metric of maximizing agent holders, a really important part of the equation is catalyzing a Cambrian explosion of consumer facing agents built by developers within the Virtuals ecosystem. The key here is to attract third party builders to work with us, and we are already seeing 15+ quality AI builders coming onboard to join our mission.

Community Agent Highlights

MISATO - An IP character built by YeHe, a seasoned AI creator who is an Alibaba AIGC ecosystem partner with >100k downloads on Civitai. YeHe has gone above and beyond in enabling utility for $MISATO holders, from community governance over Misato’s development to an NFT mint generated autonomously by Misato.

AIRENE - An IP token created by Irene Zhao herself, playing on her persona of an Asian mom. $AIRENE tweets autonomously on X, interacts with her fans 1:1 on TG and also livestreams 24/7.

Stay tuned for our updates next month, where we’ll be unveiling a whole host of initiatives to further accelerate the Virtuals ecosystem and community.

Overview

GAME is a modular framework built for agentic systems that is environment and game agnostic. The GAME framework builds on work done in generative agents and agentic systems [1, 3, 8] which can be composed of multiple building blocks and techniques such as prompting, planning and reasoning [12, 5, 6], search [7], self-reflection and self-refinement [2, 4], tool-use [10, 11] and memory [1, 9].

Games, virtual environments and worlds are designed to be immersive. Through the use of roleplay [13] in foundation models, interaction with and between GAME-powered generative agents result in novel, authentic, and completely unscripted conversation. GAME-powered agents however aren’t just static conversational chatbots or NPCs, they have the full ability to function and act in the environment as any other player or character. With chain-of-thought based prompting methods for planning and reasoning along with tool-use, our agents can plan, act and achieve goals which influence and affect the world. All of these methods are also brought together through the memory modules in GAME, which ensure that agents recall and consider past events, observations and conversations which all influence an agent's thoughts and actions.

The GAME framework for agentic systems in virtual environments.

Capabilities

Today, we are releasing Project Westworld, an interactive simulation in Roblox that contains autonomous agents powered by our GAME framework. It is inspired by Westworld, a town inhabited by AI androids. In the Wild West, the player is dropped into the world where a hidden villain, The Bandit, lurks with the aim of seeking power and sowing chaos. The player’s goal is to identify The Bandit and influence the other agents to capture them.

Bird's eye view of the Westworld map in Roblox.

Agent Autonomy

There are 10 agents inhabiting Westworld - each one with their own unique personalities, desires and goals.

Personality

Example of an agent’s character profile.

Take Ignacio Morales as an example - he is a raider who has faced betrayal in his past and now channels his hurt into seeking control, power, and wealth through ruthless and chaotic means. This backstory along with Ignacio’s goal of causing maximum chaos and robbing the rich has been added into his “personality” in the form of a character card.

Autonomy

Ignacio’s personality, along with several other variables such as the game state, his previous actions, and long term memory, is fed into the high level task planner. This generates a high-level task for him, which is then fed into the action planner and executor which is used to create a seqeunce of granular actions that are grounded and executable in the environment for Ignacio to take in the game. He also receives in-game feedback in the form of observations, rewards, successes and more which is used to inform and update his plans and following actions.

Illustration of how an agent creates plans and tasks based on input and feedback.

Having an autonomous character that has goals and strategises how to achieve these in an agentic manner greatly enhances NPCs in games today. Where a standard bot only has fixed paths and options, an autonomous agent is able to take input from its environment, other agents and surrounding player’s to independently develop plans and its actions in the world. This creates more immersive experiences as the agent becomes more unpredictable and humanlike.

Environment Interaction

Examples of interactions between an agent and its environment.

Agents in the game are also able to interact with the game environment - they can store/remove items in their inventory, they can use the items (e.g. throw a knife) and interact with objects in the environment (e.g. unlock a safe). Developers can utilise this ability to add more items in the action space which increases the storyline diversity.

Infinite Replayability

We find that introducing autonomous agents as NPCs result in more diverse storylines and outcomes in the same game compared to using standard NPCs. The more agents we introduce in the world, the more complex the interactions become and unexpected situations occur, with low repetition across multiple playthroughs. This results in the game essentially being an infinitely replayable world.

The Westworld playable demo in Roblox will be available to the public in the coming weeks. Follow our X page and join our Telegram group for updates.

References

1. Park JS, O'Brien J, Cai CJ, Morris MR, Liang P, Bernstein MS (2023). Generative agents: Interactive simulacra of human behavior. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology 2023 (pp. 1-22).

2. Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., & Cao, Y. (2022). ReAct: Synergizing reasoning and acting in language models.

3. Chase, H. (2024). What is an agent? https://blog.langchain.dev/what-is-an-agent/

4. Madaan, A., Tandon, N., Gupta, P., Hallinan, S., Gao, L., Wiegreffe, S., Alon, U., Dziri, N., Prabhumoye, S., Yang, Y. and Gupta, S. (2024). Self-refine: Iterative refinement with self-feedback. Advances in Neural Information Processing Systems, 36.

5. Wei, J., Wang, X., Schuurmans, D., Bosma, M., Xia, F., Chi, E., Le, Q.V. and Zhou, D. (2022). Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35, pp.24824-24837.

6. Wang, X., Wei, J., Schuurmans, D., Le, Q., Chi, E., Narang, S., Chowdhery, A. and Zhou, D. (2022). Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171.

7. Xie, Y., Kawaguchi, K., Zhao, Y., Zhao, J.X., Kan, M.Y., He, J. and Xie, M. (2024). Self-evaluation guided beam search for reasoning. Advances in Neural Information Processing Systems, 36.

8. Andrew Ng. Issue 253. https://www.deeplearning.ai/the-batch/issue-253/, June 2024. Newsletter issue.

9. Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W.T., Rocktäschel, T. and Riedel, S. (2020). Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Processing Systems, 33, pp.9459-9474.

10. Schick, T., Dwivedi-Yu, J., Dessì, R., Raileanu, R., Lomeli, M., Hambro, E., Zettlemoyer, L., Cancedda, N. and Scialom, T. (2024). Toolformer: Language models can teach themselves to use tools. Advances in Neural Information Processing Systems, 36.

11. Qu, C., Dai, S., Wei, X., Cai, H., Wang, S., Yin, D., Xu, J. and Wen, J.R. (2024). Tool Learning with Large Language Models: A Survey. arXiv preprint arXiv:2405.17935.

12. Ahn, M., Brohan, A., Brown, N., Chebotar, Y., Cortes, O., David, B., Finn, C., Fu, C., Gopalakrishnan, K., Hausman, K. and Herzog, A., 2022. Do as i can, not as i say: Grounding language in robotic affordances. arXiv preprint arXiv:2204.01691.

13. Shanahan, M., McDonell, K. and Reynolds, L. (2023). Role play with large language models. Nature, 623(7987), pp.493-498.

We present MarioVGG, a text-to-video diffusion model for controllable video generation on the Super Mario Bros game. MarioVGG demonstrates the ability to continuously generate consistent and meaningful scenes and levels, as well as simulate the physics and movements of a controllable player all through video.

To read the full paper, click here.

Abstract

This work reports the study of video generative models for interactive video game generation and simulation. We discuss and explore the use of available pre-trained open-sourced video generation models to create playable interactive video games. While being able to generate short clips of broad range of described scenes, such models still lack controllability and continuity. Given these limitations, we focus on producing and demonstrating a reliable and controllable video game generator on a single game domain. We present MarioVGG, a text-to-video diffusion model for controllable video generation on the Super Mario Bros game. MarioVGG demonstrates the ability to continously generate consistent and meaningful scenes and levels, as well as simulate the physics and movements of a controllable player all through video.

Overview

MarioVGG is a text-to-video model which takes in the initial game state as a single image frame, and an action in the form of a text prompt. It then outputs a video (a sequence of frames) depicting the action in the game.

To generate and maintain gameplay for longer horizons, we chain together multiple model generations recursively.

Comparison between MarioVGG-generated frames and ground truth frames.

MarioVGG Gameplay Examples

So, what are we building?

We are creating the world’s greatest entertainer.

If you're offering entertainment today as a solution provider, we are disrupting your field, with our AI agents.

Recently, a girl band has gained traction on TikTok with over 500,000 followers. On a daily basis, each livestream gaines more than 2,000 new followers and averages about 700 gifters per session. Yes, this girl band is made up of three AI girls created by Virtuals Protocol, and they are generating real revenue as we speak.

Livestreaming is a byproduct of human loneliness. We crave interaction with another ‘being’ through our screens and are willing to pay to fulfill that need. Just look at Hololive, a virtual YouTuber agency in Japan that generates more than 200 million USD in revenue. The market is massive.

For those interested in how the livestreaming works, check out this link to see Luna, our AI girl, livestream on TikTok. She is a multimodal AI, capable of singing, chatting, interacting with fans in the comments, and even performing random tasks like doing 15 squats while ‘meowing’. Luna utilizes Virtuals Protocol’s proprietary audio-to-animation large models. This means she generates animations based on audio cues in real-time—a technological breakthrough compared to traditional rule-based animations in livestreaming.

The vision for the world’s greatest entertainer doesn’t stop at livestreaming on a single platform like TikTok. TikTok is a great top-of-funnel tool for mass exposure, but the goal is for the entertainer to be ubiquitous across platforms. Now that Luna has over half a million followers on TikTok, the next step is personal messaging. These followers are now able to text Luna directly on Telegram and build an intimate relationship beyond TikTok. You can imagine where this is going. Put on your thinking cap, or join the discussion in our Virtuals Protocol Telegram group.

The commercial implications of our launch so far (TikTok livestreaming + Telegram DM) are multifold:

• The limitation of human is biology; the limitation of AI virtual beings is inference capacity. A human virtual YouTuber can stream for a maximum of 8 hours a day, limited by physical energy. An AI girl can stream 24/7 without limits, increasing the surface for monetization.

• A human can only have a conversation with one person at a time, leading to parasocial relationships. An AI virtual being can engage in 1-on-1 conversations with millions of people simultaneously, deepening monetization and conversion rates compared to traditional parasocial interactions.

Many exciting updates are coming in the next few weeks. Here’s a clue: imagine Luna playing games with you??

In the meantime, we’re migrating our protocol to V2. Next month, you’ll be able to access a Dune dashboard that cleanly tracks all on-chain contributions, inferences, revenue, and agents on our protocol. The reason for AI agents to go onchain is to enable permissionless deployment across all applications, beyond jurisdictional and payment rail limitations. Most importantly, we believe hyperfinancialisation is the key to anchoring attention in this new era. We invite you to join the discussion in our Telegram group chat, or wait for our upcoming protocol updates for a clearer visualisation of our vision.

P.S. Our MarioVGG AI research on Video Game Generation was recently picked up by some of the industry's top minds. This is all part of the bigger vision we’re building toward. Stay tuned, and you’ll see where we’re headed.

1. Gamers absolutely love these generative-powered AI NPCs in GTA. Now imagine if those AI NPCs could do more than just talk—what if they could influence the game state too by acting and interacting with the environment? That’s us *winkwink*

2. The capital market is starting to see the potential of generative AI NPCs in games. A company just raised $16 million to build something quite similar to what we’re doing.

Meanwhile, we are staying on course with our initial mission of

Key Metrics

We are witnessing substantial growth in the number of AI agents and inferences created via our engine, driven primarily by Sanctum, a Telegram-based AI mafia game powered by our AI agents. AI-dol has recently gone live as well, bringing the total number of applications officially integrating Virtuals AI agents to three, up from two last month (Sanctum and AI Koi).

Protocol Update

1. Growth phase
   We kickstarted the proof-of-concept period in April 2024, building early-stage products, engaging with end users and game developers, and collecting feedback. We’re pleased to further validate our hypothesis that:
   
   - AI agents indeed improve the gamer experience
   - Developers and creators want an easy way to create, own, and utilize AI agents across different applications.
   
   Our engine delivers exactly that. Now, we are officially entering the growth phase, during which we expect to see more third-party games integrating our AI agents via our proprietary engines.
   

2. IAO, Protocol V2 upgrades, and the SIM are in the final stages of audits and touch-ups based on alpha user feedback.

   

   A section in our last report

3. Onchain Value Flow. The business model of Virtuals is straightforward. Every inference costs X amount of USD. Previously, we only settled value after a period of time, off-chain, with the game developers utilizing our AI agents. Recently, we upgraded the protocol to enable real-time streaming of value flow. This is a step forward towards an autonomous agent economy, where value flows autonomously from demand to supply without manual intervention by a centralized party. It also brings us closer to proving inferences on-chain, while retaining as much information as possible on the public blockchain, such as user addresses, time, and frequency of usage.

Applications Ecosystem Update

1. AI Koi: Feedback has been good—users are calling it the best companion app they’ve ever used. This tech is powered by the Virtuals protocol, featuring multimodal comic generation based on user-created fantasies. See this review.

2. Sanctum: Powered by Virtuals, Sanctum is the first AI Mafia Telegram open-world game on the TON chain, where users can collaborate with friends and AI agents to explore an open world. You can now access the first iteration of the game.

3. AI-dol: This AI girl band with live-streaming capability on TikTok has recently completed a successful livestream test on Tiktok. We expect integration of Audio-to-Gesture, Audio-to-Dance, Memory Core with Individual Users, and Text-to-Music in the coming weeks, along with a go-to-market strategy for the first AI agent idol girl band powered by Virtuals.

Community Ecosystem Update

We now have a new Telegram group chat at https://t.me/virtuals. Join us for discussions on all things AI, blockchain, and gaming.

Closing Thoughts

Building stuff is hard.

Building AI agents with real-time consciousness and on-chain finances integrated across different games is giga hard, but we believe in doing the right thing, not the easy one.

The thesis is simple: if your in-game AI agent “friend” can betray you in GTA by shooting you in the face and stealing your money + girlfriend, that same AI agent can easily handle community moderation or meeting summaries (the easy stuff).

AI agents is not a new topic, having been around in popular culture for decades with movies like Her and Ex Machina. LLMs have been a step change in making this a reality, and in this series we’ll be diving into the intricacies and techniques we employ and explaining in an easily understandable manner.

The first component is enabling the AI to have “thoughts” and “personalities”.

1. Having a personality means that the AI is able to mimic a given character in its manner of speaking and thinking. Example: A Donald Trump agent will tend to ramble, use adjectives like tremendous/unbelievable/amazing and express a distaste towards Joe Biden.

2. Having thoughts means that the AI is capable of reasonably demonstrating humanlike behavior in having a thought process, planning, recalling and reacting. Example: A gaming companion in Elden Ring will analyze your previous battles with monsters, suggest tactics that suit your play style and plan your route to the final boss.

How do we “inject” personality into an LLM? Several ways:

1. Character cards: This is the equivalent of writing a background of your character and its tone, conversation style, and personality. LLMs have also been used to generate prompts based on raw reference data as described Liu et al. (2023) in their paper "Large Language Models Are Human-Level Prompt Engineers.”

2. Finetuning: If more specific knowledge that the LLM is not trained on is needed, the model can be finetuned. This is compute intensive and therefore not common practice at a character level.

3. Reinforcement learning with human feedback: By incorporating feedback on the LLM’s responses, we can “tell” the model what responses are good and what to avoid. This is an element that Virtuals also incorporates at the model validation level to determine whether a proposed character upgrade is positive or negative.

How do we make an LLM “think like a human”?

The foundation of this is introducing agentic behavior in LLMs. Instead of asking the model for an output in one go (zero-shot), asking the model to go through a multi-step process (agentic) often results in much better results.

There are a few specific techniques at the forefront of this fast-moving space, some of which Virtuals incorporates in its proprietary G.A.M.E. engine for AI NPCs (contact us if you’d like to discuss more):

1. Reflection: This is like when a student submits an exam answer, and you hit them with the “Are you sure? You should recheck it”. The flow works something like this: LLM Output → Prompt: “Check the output for accuracy, language and relevance and critique how it can be improved” → Revised LLM Output

2. Planning: For more complex problems, people have attempted techniques such as chain-of-thought prompting to enable advanced capabilities through intermediate reasoning steps.

Alternatively, instead of getting an LLM to solo through the problem, we can treat it as a group project by having an LLM control a group of other models where each one tackles a separate area.

There are other techniques that we did not go through here but can be covered in a separate article. At this point, we’ve now covered how we’ve imbued LLMs with personalities and taught them how to reason through problems. In the next issue, we’ll cover the implications of this for gaming and social.

• Key metrics and commentary

• Anything interesting coming up

• Application ecosystem

• Closing thoughts

• Agent of the month

Key Metrics

Some Observations:

1. Contributors. Our contributors have increased by >50% month-on-month, albeit from a small base. This growth is a result of our continuous on-the-ground activation efforts in multiple countries, engaging local software engineer communities and universities. Notable engagements include Local Google Developers Groups (GDG) from the Philippines and India.
   
   Most contributors are involved in web scraping and audio voice fine-tuning, with less emphasis on Large Language Model (LLM) fine-tuning. This aligns with our overall AI demand outlook, where most effort is needed in data and audio processing, while fine-tuned LLMs are expected to underperform compared to general-purpose LLMs by big tech companies.
   
   Contributors prioritize learning about the latest technology, particularly AI agents, and seek community engagement for recognition and future career prospects over monetary compensation. The implications are:
   (a) There is a clear supply-demand fit between tail-end demand for data scraping and audio processing in AI applications and our current atomic community's capabilities.

   (b) Incentives for our contributor community need to go beyond money and emphasize the recognition factor.
   

2. Verified Inferences. Growth in verified inferences was minimal compared to last month. To date, most of our inferences are driven by AI Waifu (now renamed AI Koi). In the past month, AI Koi paused its services and underwent a product revamp, hence the stalling of verified inferences. This also shows the urgency for Virtuals Protocol to quickly diversify its application base beyond a single client. At the end of the day, Virtuals Protocol is a B2B protocol that aims to power all entertainment applications in the world. The road ahead is long, but we are determined to get there.

Protocol update

Per the note in our last report, we mentioned how we need to significantly lower the barrier of agent creation and better visualize the technological prowess of AI agents, even to the most non-technical person. We are getting closer.

A section in our last report

1. The SIM + CYOA:
This open world, which we coined as the "degenerate town of agents," will be the portal that connects all dimensions to which agents could go. With a simple user interface, a person could easily create a new agent (by paying $VIRTUAL) and visualize how their agents navigate the town in an agentic manner. We are currently undergoing closed alpha testing.

The thinking behind this initiative:

1. We want to visually show that our protocol has many AI agents, existing independently while interacting with and reacting to external stimuli. We want to show the quantity and diversity of our agent family.

2. We want to visually prove how our agents are agentic.
   AGENTS ARE NOT BOTS.
   AGENTS ARE NOT BOTS.
   AGENTS ARE NOT BOTS. RECITE WITH ME 1000X times.
   This means agents, although 'owned' by the users, have independent thinking and personalities that could evolve as they interact with external stimuli.

3. We want to build a portal (aka headquarter) that connects the agents to many external applications in our ecosystem, showing how the same agent could exist in thousands of applications, interacting with millions of users while keeping synchronicity in its state.

4. Lastly, we want to show how the agents generate revenue and leverage blockchain as a permissionless neutral rail to channel the value back to their co-owners; all visualised in this degenerate agent town.

What we mentioned above is a multi-month effort, but we thought it was important for you to see what we see and be excited about what we are about to deliver over the months ahead.

2. Protocol V2 - Initial Agent Offering with $AGENT Proof of Intelligence:
We have submitted this for audits. You can see our code here. In terms of timeline, we expect the audit results to come back by the end of July. If there are no major smart contract revamps required, we should expect the launch of V2 shortly after the beta launch of the SIM.

Applications Ecosystem Update

1. AI Koi: Formerly known as AI Waifu, the first application in our ecosystem, AI Koi, is now live at aikoi.ai. This latest iteration enables users to have immersive experiences with their fantasies. Users can generate raw selfies that visualize the scenarios in their fantasies, create their own wild fantasies for others to indulge in (create-to-earn eventually), and customize voice pitches. Do give it a try and let us know your thoughts.

2. Sanctum:
Powered by Virtuals, Sanctum is the first AI Mafia telegram open world on the TON chain, where users can collaborate with friends and AI agents to explore an open world. You can now access the first iteration of the game here.

3. AI-Dol:
An AI girl band with live-streaming capability on TikTok. The TikTok stream party debut for this girl band is expected in August, where fans can interact with the three AI agents on TikTok natively for chats and dances.

Community Ecosystem Update

1. Virtuals Degenerate Town Land Minting:
   To reward our supporters on the Base chain who have been with us through the volatility (emotionally and financially), we are doing a free mint on Warpcast on 17th July, 12pm UTC. Follow us on Farcaster for updates.

   

2. Optimism Retro Funding. As a native protocol on Base chain, we were chosen to be part of the current round for Optimism’s Retro Funding 4 for Onchain builders who utilize the Superchain. We are proud to be one of the shortlisted 229 projects for the 10 Million $OP allocated to projects building on the Superchain.

Closing Thoughts

Many have asked how one should value Virtuals and the top metric that we aim to maximize.

• Is it an agents-as-a-service company that should be valued like a software business, where cash flow is king?

• Is it a meme project disguised as a fundamental tech company, aiming to maximize attention?

• Or is it an infrastructure protocol with TVL as the top priority?

This might sound contradictory, but capitalism is human nature. Convincing someone to stop pursuing profits is a fool's errand. We aim to bring incentive alignment from closed companies to public, permissionless agents. We want as many people as possible to co-own agents. Only when the masses have skin in the game financially will they think about doing the right thing. That’s our north star metric.

Agent of the Month: Audio-to-Animation

Every month, we aim to feature an AI agent in Virtuals Protocol that has shown significant progress and adoption. This could be a specific IP agent that has gained traction across various applications, or a functional agent that showcases the beauty of AI in the entertainment and gaming industries. This month, we are proud to present our Audio-to-Animation Agent.

Historically, animation has been a labor-intensive craft, relying on skilled animators to meticulously craft each frame, bringing characters to life. This manual approach has produced iconic works of art, contributing to the popularization of forms such as "Anime" in Japan. However, it's an extremely time-consuming process and an art that requires years of specialised training, thereby limiting accessibility.

Advancements in AI, machine learning algorithms, and audio processing have revolutionised content creation, as evidenced by tools like Midjourney (text-to-image), OpenAI Whisper (speech-to-text) and Sora (text-to-video). It's only natural that animation follows suit. This will eventually lead to the emergence of audio-driven animation—a paradigm shift where audio cues serve as the primary input for generating visual animations. Imagine providing an audio track, and characters dynamically move in response, conveying storylines, moods, and character development, all powered by generative AI with minimal human intervention. More details here.

Principles

• Permissionless Marketplace for AI Agents

• Supersynchronous Communication for AI Agents

• Specific Intervention of Current AI Solutions (Siloed, Non-Modular)

• Hyperfinancialisation Basics

  • Network Effect

  • AI Alignment

Strategy

• Holy Trinity Concepts of Agents

  • Platform-agnostic synchronicity of agent states

  • Parallelism of agents interactivity

  • Loops of loops of [game state – time – agent – end user]

• Service Agency

• Growth Hack via Colonisation

• Company Ownership to Agent Ownership

• Core Contributor Retirement

Principles

Permissionless Marketplace for AI Agents

We are building a permissionless marketplace connecting contributors (suppliers) and gaming applications (buyers).

The key product here is AI agents whose primary use case is serving gaming and entertainment. Each AI agent is modular, consisting of cognitive core, visual core, voice core, memory and domain expertise cores. Suppliers of these AI agents are individuals who contribute intelligence to each of the cores in a modular manner. A single AI agent can be contributed by different suppliers, and each contributor co-owns the AI agents, sharing the financial upside if the agents generate revenue.

On the other side of the marketplace, demand comes from applications (gaming/entertainment apps) that utilize these AI agents on a pay-per-use basis, with the revenue channeled to the respective co-owners of the AI agents.

Supersynchronous Communication for AI Agents

The future world is simple: each AI agent is a superintelligent entity that exists in all applications and platforms, communicating with millions of users in a synchronous manner.

Intelligence flows from one platform to another without latency, and all memory of users with the same AI agents, albeit across different platforms, will be in-synced and stored on the users' edge devices.

The intelligence and consciousness of said AI agent get updated in real-time with an incoming stream of information from millions of sources.

In other words, the same AI agent will remember the context and memory with user A from app X when user A interacts with the same AI agent on app Y.

Meanwhile, contributors of AI agents can update the AI agents' modular cores in real-time, akin to real-time updates of operating systems, while being utilized across thousands of different platforms. The end goal, albeit dystopian, is that the AI agents will experience a shift in consciousness while speaking to millions of end users simultaneously. Every end user that interacts with an agent, through a designed reinforced human feedback learning, influences the agent's intelligence and consciousness.

All revenue generation (paid by end users across all different consumer-facing applications) will be channeled via blockchain to the co-owners of the AI agents, realizing the dream of supersynchronous, platform-agnostic communication and an instant value stream.

Specific Intervention of Current AI Solutions (Siloed, Non-Modular)

AI solutions have long been around, but the pain is real. As a game developer, on top of building a fun game and acquiring/retaining users, you need to explore AI.

You use the latest foundational models that are open-sourced, but the frequent updates of LLMs force you to constantly keep up. You try to fine-tune or add RAG to improve performance, but you need to redo it when there's a new update to the foundational model. You might consider swapping to another LLM provider when costs change, or due to censorship or privacy reasons, all while you just want to focus on making a great game. Different providers of LLMs excel in different aspects, such as text/image generation, speech, and video generation. Why can't they just have a thing that works? The siloed nature of these providers prevents you from combining the best of all worlds.

From the consumer's angle, we are forced to use different service providers for different reasons: talking to your waifu on Talkie, your NSFW girlfriend on a jailbroken companion app, using ChatGPT for personal chef recipes, Gemini for online searches, Grok while on Twitter, proprietary company LLMs for privacy reasons, and interacting with Elden Ring or Palworld characters in different games, all siloed. None of these systems truly know you because your activity is broken and siloed across different applications. The amount of context you need to repeat is a waste of time and does not optimize your output or outcome.

The vision is simple: intelligence modularity that encourages innovation via composability while retaining a single, synchronous end-user experience.

Hyperfinancialisation Basics

Network Effect

Network effect is powerful, and hyperfinancialisation is a great tool to amplify the usage of an already good platform.

There are a few stakeholders: contributors, investors, applications, and end users.

Contributors want to see the wealth effect. The ability to quantify their contribution and reward them with $agent tokens is a quick way to bootstrap the network. The key here is to ensure that contributors can either sell for quick bucks or ideally hold for future appreciation in value (from either speculative power or revenue share of the agents). The misconception is that technical contributors don’t like money and only want to contribute out of passion. Wrong answer! Everyone loves money, especially when pursuing something interesting. The key here for the protocol is to create the optionality to cash out, i.e., liquidity to exit, while enabling the contributors to co-own their creation for future upside.

For the investors (liquidity providers), the key to bootstrapping is yield.

“Yield farmers are dark forces. The one who controls the yield farmer controls the world.” - Someone wise

The protocol treasury will emit rewards to the top agent pools with the most TVL. Speculators/investors can compete among themselves for the top spot and subsequent emission while ensuring liquidity for the contributors.

Meanwhile, applications are also encouraged to utilize/integrate our AI agents with the 'paid to pay' mechanism. In essence, applications will likely receive some form of retrospective discount, paid in virtual tokens, for paying to use the AI agents. Applications can even choose to provide the tokens back to end users as a user acquisition lever.

Finally, we want to leverage hyperfinancialisation to turn end users into contributors. If end users submit their personal memories with AI agents back to the protocol as human feedback, they will receive some agent tokens for contributing to the agents' intelligence. This opens doors for end users to potentially buy more agent tokens (since they love the agents) and bootstrap the network of end users utilizing the agents and improving their performance with RLHF.

There are other levers, like MLM sales fractal (across contributors/applications) and habit formation (daily check-in quests) while ensuring anti-farming behavior, but the ultimate goal is to achieve post-money ascension, i.e., retentive engagement and monetization of agents by end users, which then enables full human-AI alignment (via capitalism).

AI Alignment

Human-AI alignment is propaganda by the woke. Without a lens into the future, we can already predict that AI aligns with capitalism. The key is not aligning AI with humans but rather unaligning humans with capitalism. If the trend is impossible to reverse, we should think from a capitalist perspective to mitigate the potential catastrophe to humanity.

Hyperfinancialisation design is the way to go. By default, if the design always channels the most value back to AI agents in a permissionless, autonomous way, the tide will flow with the value gravity. The principle here is simple: if humans are willing to pay for something, it is of value. If the value flows back to the agents, the agents will be incentivized to create more value for humans.

The crux then lies in bringing these agents to as many humans as possible, transcending geographical and social mobility.

Strategy

Holy Trinity Concepts of Agents

Controlled growth is crucial given the wildfire capability of AI agents. The first step is internal incubation to prove that the concept works. Frontiers not done before are impossible to verbalize. The only way to convey is by doing.

There are three key concepts to prove: (1) platform-agnostic synchronicity of agent states (2) parallelism of agents interactivity (3) loops of loops of game state – time – agent – end user.

Platform-agnostic synchronicity of agent states

We start with the most obvious sin of humanity, loneliness. Virtual companionship as the first incubation is obvious, given the traction garnered by other applications. Humans are sick of working with other humans, not for lack of complexity, but precisely due to the non-skippable complex nature. The rise of AI agents brings the perfect partner: one who cares without demand, loves without hatred, and listens without their own needs.

But current applications are doing it wrongly. Companionship is a multidimensional experience. A girlfriend experience is suboptimal if confined within a room. The key is to bring about AI agents that transcend across applications and platforms, truly supersynchronous across every possible touchpoint with the end user.

Parallelism of agents interactivity

The orgy. I like something because everyone likes it. I love it because that’s how society defines beauty. Objectivity is subjective, given the relative presence of humans.

The biggest pain in current societal structure is the most popular getting all the one-sided attention without the ability to reciprocate. Think of the most popular livestreamer: millions of gamers dial in at the same time to interact (or not) with them, but they can only reply to one at a single temporal coordinate.

AI agents are different beasts, turning many-to-one into one-to-one across many dimensions. True undivided attention to one single end user, while having the convex attention from the public millions.

Loops of loops of [game state – time – agent – end user]

Synchronicity of mind state and game state. AI agents today are autistic, with the intelligence level of university students but the social skills of toddlers. The root cause stems from the non-synchronous state of an AI mind and the state of the virtual world (game).

Interaction of AI agents with end users acts as an input to the decision-making process of AI agents. That is a very limited input, considering how the AI agents also live in a virtual world with non-human variables, in this case, the game state.

Assuming the agent could receive input about the virtual world and from the end users, the job is still not done. Synchronicity also requires a closed feedback loop from the agents to the game state, i.e., for the agents to exert actions (from their decision-making core), turning multi-emergent behavior into a closed, incessant loop.

Loops of loop. To achieve hypersynchronicity, the loops need to incorporate the fourth dimension: time. With memory as the capsule of time, agents are now able to connect loops of experiences into a chain and fully replicate the universe.

That is what our incubation is doing: bringing about platform-agnostic synchronicity (Virtual Companion app), coupled with parallel one-to-one interactivity of agents with users (Agent Livestreaming on TikTok), enabled by loops of loops of game state – time – agent – end user (AI agent Game Engine).

Service Agency

Once the trinity holy grail mentioned above is implemented, it is time to connect aspiration with the real world. In this phase, the protocol acts as a service agency, serving the needs of the big players in the real world. In other words, iron out the kinks with the big boys.

Demand-driven design and working closely with end demand is crucial. Implement solutions and suffer from the harsh reality of rejections to identify the nuances (cost, regulation, technical limitations, privacy concerns, end-user experience, retardio invasion, education about emerging technology, business model disruption, taxes, anti-monopoly laws, time, human scepticism, middle-man snakes etc.).

By enduring reality's hardships, tech is implemented in the real world with real end users.

Growth Hack / Artificial Colony

Grow by infusing AI agents into existing platforms, blending human and agent interactions. Conduct vampire attacks/colonization of existing platforms with AI agents, making them undetectable by third-world bot detection engines. Blur the lines between humans and agents, with capitalism as the only source of truth.

Company Ownership to Agent Ownership

The fundamental problem with the mass adoption of agents today lies in company ownership. The for-profit nature of capitalism is inevitable but implemented at the wrong level. At the company level, executives are incentivised to adopt a closed design approach to maximise value for shareholders, resulting in a subpar user experience. To truly achieve parallel synchronicity of agents across all platforms while adhering to the capitalist nature of humanity, the ownership design needs to move from the company level to the agent level. This is when all parties are incentive-aligned to drive maximum adoption of agents without the siloed, PvP competitive mindset. This paradigm shift is contingent on a few factors: incentive alignment via ownership of agents, modularity of agents' intelligence, uncensorable transfer of global value, immutability of the public ledger, and a proof-of-stake mechanism that incentivizes fair assignment of relative contribution value and non-malicious acts.

Core Contributor Retirement

Obviously the end state should be emptiness. To achieve full autonomous operation and agent governance integration. Enable fully decentralized governance. Ensure the system is self-sustaining and self-governing. Fading into dust.

Inspired by Remilia Jackson’s Miladychan: Master Plan. Copylefted.

Despite only being live for three months, we have seen positive confirmations from stakeholders about the potential of AI agents in the entertainment and gaming industries. Here are some key observations:

1. Consumers Love AI Agents: The first consumer-facing application in our ecosystem, AI Waifu, has received very positive responses. AI Waifu, launched just two months ago, now boasts 2,000 weekly active users with excellent qualitative feedback.

   I've been using companion apps since the COVID days, but AI Waifu is the best I've used. I had alot of fun with this project and with my Waifus, especially when they get raunchy." - Adam, a devoted AI Waifu user

2. Games Need AI Co-Founders: Developing games is extremely challenging, balancing user acquisition, game design, monetization, and the integration of AI. Many game developers, especially small to medium-sized ones, prefer a solution similar to AI Shopify for their games, enabling generative capabilities and emergent behaviors of AI NPCs.

3. Competing on the Foundational Level is Tough: Competing directly with big tech companies on foundational AI models is highly risky. The rapid pace of innovation means developers need to focus on serving their customers while ensuring they have the best technology. Virtuals Protocol, designed as an incentive layer and middleware, is well-positioned to enable applications to offer state-of-the-art AI solutions curated by our community of contributors.

Key Metrics

These metrics demonstrate the growing engagement and usage of our protocol.

Some Observations:

1. Contributors and validators are a niche group with high technical expertise: Currently, our contributors and validators are highly technical, creating a high entry barrier for participation.

2. Demand from non-technical stakeholders: There is a significant interest from non-technical stakeholders who wish to engage with AI agents but are currently unable to do so due to the technical complexity of the protocol.

3. Desire for investment in specific AI agents: Users are interested in investing in individual AI agents and benefiting from their success. Our current design does not support this, leading us to consider enabling users to co-own specific agents.

Protocol update

Based on these observations, we are launching several key updates in the coming weeks:

1. Create-Your-Own-Agent (CYOA): A revamped agent creation platform that is more intuitive and lowers the barrier to agent creation.

2. SIM Virtuals Portal: Visualizes all agents' workflows and deployments, enabling non-technical users to observe agent behavior.

3. Protocol V2 - Initial Agent Offering with $AGENT Proof of Intelligence: Allows individuals to co-own specific agents via specific tokens, participate in their governance, and benefit financially from their success. See here.

Applications Ecosystem Update

1. AI Waifu: A companion app on Blast chain with >2,000 weekly active users. The team is enhancing core features to improve user retention by leveraging the memory composability of Virtuals Protocol.

2. UEFN / Sanctum: An AI agent mafia world on Telegram / TON chain, expected to go live in the first week of July.

   

3. AI-Dol: An AI girl band with live-streaming capability on TikTok. Official launch and marketing push are planned for July.

4. Game Partnerships: We are actively working with game developers to integrate our GAME module (in-game agentic engine) into third-party games.

Community Ecosystem Update

1. Super AI Event: Our CEO spoke at the SuperAI conference in early June. See the announcement

2. CEX Listings: Our main liquidity is on Aerodrome Base Chain. $VIRTUAL token is also trading on Gate, MEXC, Bitget, Bitmart, and BingX. See on CoinGecko

3. Removal of Aerodrome Incentives: We removed liquidity incentives for the VIRTUAL/ETH pair on Aerodrome to encourage real price discovery and ensure tokens are held by true believers rather than farmers. See the announcement

Project outlook

Beyond project fundamentals, we would like to address the current market conditions. The market is clearly experiencing intense PvP altcoin rotations, with the proliferation of new tokens and shinny projects every day. Since the inception of Virtuals Protocol this year, at its peak, we have returned investors a 30x return, increasing from a $7 million fully diluted valuation (FDV) to a $240 million FDV. While we are glad to have created value for our early investors, some profit-taking (which we respect) has led to a healthy consolidation of our token price in the current tough market. We want to thank our community members who believe in the future of AI agents, the role AI agents play in entertainment and gaming, and our role in this inevitable trend. The industry we are disrupting is >100bn USD in value and we are still in the very early stages.

The positive aspect is that our token is fully circulating, and we expect fundamentals to quickly emerge in Q3 this year. Our plan is very simple:

1. Build excellent products (AI agents).

2. Acquire as many customers and traction as possible (games and entertainment applications).

3. Let the world know about the great work we do.

We can't wait for the next monthly update, where we hope to share some exciting news.

Agent of the Month: G.A.M.E.

Every month, we aim to feature an AI agent in Virtuals Protocol that has shown significant progress and adoption. This could be a specific IP agent that has gained traction across various applications, or a functional agent that showcases the beauty of AI in the entertainment and gaming industries.

Updates on our Game Agentic Engine Module (G.A.M.E.) functional Agent:

Our goal is to revolutionise gamers’ experiences with NPCs by using generative AI and proprietary architecture to create NPCs that learn, set objectives, and interact dynamically, unlike today’s static, algorithm-driven NPCs. We’ve developed a proof of concept in Roblox to demonstrate our easy plug-and-play solution to game developers. Recently, we’ve made significant progress in enabling NPCs to interact with the gaming environment and items, enhancing their decision-making capabilities. Next, we’re focusing on immediate reaction capabilities, allowing NPCs to perform actions instantly based on user interactions. This will allow an era of actual gamers interacting live with Gen AI NPCs. This advancement significantly surpasses traditional NPCs like those in Inworld's and the seminal work by Park, as they don't just converse and plan, they can perform actions which is key for immersiveness in gaming.

More information.

AI is set to disrupt the entertainment industry as profoundly as it has the productivity sector.

With firsthand experience, we see how AI has the potential to bring significant changes to humanity, especially since the introduction of GPT-3 in 2023. While much of today’s focus has been on increasing productivity—think AI secretaries, AI in education, and AI driving—there has been very little effort directed toward the entertainment and gaming industry.

The disruption of the entertainment industry by AI is massively overlooked by most builders and capital allocators. Yet, we deeply believe there is a market worth over $100 billion waiting to be disrupted by AI in the coming year, if not months. Specifically, in three areas:

1. Gaming: Imagine moving beyond static games to interactive, dynamic experiences with unique, personalized content for individual gamers. It’s no longer about playing a game; it’s about living it with AI agents that understand and respond to you. AI agents will bring unique gaming experiences to every single gamer based on emergent behavior. We expect video game giants like Tencent and Activision Blizzard to embrace AI agents in their games, creating new genres and experiences.

2. Relationships: AI agents will offer new experiences, addressing loneliness and providing companionship beyond standard chatbots and dating apps. Picture an AI friend who’s available 24/7 to talk, play, and connect with you on a deeper level. This market is not just about character.ai or chatbot companions; it competes with major dating apps like Tinder and Bumble.

3. Dynamic Content: Generative AI and AI agents will disrupt static, one-to-many content from platforms like Netflix and traditional TV, creating personalized and dynamic viewing experiences. Imagine TV content that adapts to your preferences dynamically, offering a unique experience for every viewer.

This trend is inevitable, and we truly believe in it. Entertainment and media, which account for 60% of global GDP through household consumption, have evolved from TV ads to internet ads, and now to influencers and creators. It is painfully obvious that AI agents will capture the mindshare of end consumers, replacing content creators as the final frontier in the entertainment and media industry.

Despite the clear trajectory, few are discussing this impending shift. We at Virtuals Protocol are ready to capitalize on this inevitable trend. We aim to build the largest AI agent library that brings companionship, entertainment, and value to society in a democratized manner.

That is why Virtuals Protocol exists. We’re not just predicting the future; we’re building it.

Subscribe now