Who Owns the Loom: Virtuals Aims to Turn AI Agents into Tradable 'Virtual Nations'

AI agents can now open accounts, hire people, and start working, and VIRTUAL aims to make this economic system a reserve currency.

Written by: Vaidik Mandloi
Compiled by: AididiaoJP, Foresight News

Every few years, a technology brings to the forefront a question that economics alone cannot answer: What do you do if a machine can do your job better, cheaper, and without rest? Who owns the output created by these machines?

This article delves into one of the most ambitious experiments currently in the crypto space. The Virtuals Protocol is building a near 'nation-level' infrastructure for AI agents: identity, banking, commercial layers, capital markets, layered with physical robots. Its bet is that ordinary people should be able to own the autonomous machines that begin to create real economic value; and the financial tracks inadvertently left by the era of crypto speculation are just right for this purpose. The article asks: does the execution match this ambition? But first, we need to return to a question that predates crypto by about two hundred years.

The Loom

In 1811, a group of textile workers in Nottinghamshire stormed workshops and smashed sock-knitting machines into scrap metal. The movement soon expanded, prompting the British government to send 14,000 soldiers to the Midlands to stop the weavers from continuing their destruction—more than the number of troops Wellington had in the Iberian Peninsula against Napoleon.

The British Parliament declared machine destruction a capital offense. In 1813, 17 people were hanged in York. These individuals were labeled 'Luddites.' The term later became a pejorative, as if they were merely fearful of technology and unable to adapt. But they understood machines better than anyone. They were craftsmen who had spent years mastering narrow looms capable of weaving high-quality fabrics; what they smashed were wide looms and new models—machines that untrained youths could operate, earning only a third of the craftsmen's wages. The clothing produced by wide looms was of inferior quality but unprecedentedly cheap, and it was this affordability that swallowed the market. With each wide loom entering a workshop, a craftsman's livelihood was taken away.

This theme has replayed in every generation since. Those involved always feel that their situation is more unique and dire. Yet history repeatedly proves that machines rarely eliminate jobs; they merely rewrite who does the work and who owns the output. Peasants were driven from the commons into factory towns, trading ownership of crops for hourly wages on someone else's clock. Factory workers became office employees, renting out their time for higher pay. Office workers then became gig workers: Uber drivers, Fiverr freelancers, still doing work, but with a classification designed for platforms to take economic returns while workers bear the risks. Each time, total output increased, but the share of value owned by those doing the work diminished. Today, around 500 million people globally engage in labor that doesn't even qualify as freelance. In India, about 50 million are trapped in debt bondage, akin to a new set of documents replacing medieval serfdom.

Now it's the turn of the AI revolution, but this time the loom seems to be turning by itself. AI can already autonomously drive businesses, account for profits, and reinvest those profits for growth. Health tech company Medvi recorded $401 million in revenue last year with only two full-time employees. You can now hire a programming agent for about one dollar an hour to work around the clock without taking a day off. They are already real economic actors, and thus the ownership question cannot be ignored. When a worker is a software that can replicate at near-zero cost and can hire other machines to assist, who really calls the shots?

The last large-scale attempt in the crypto space to share technological dividends with ordinary people was Axie Infinity. This play-to-earn game promised economic liberation to workers in the Philippines and Southeast Asia, peaking at 2.7 million daily active users, with many Filipino players earning more from the game than from local jobs, only to later collapse into a cautionary tale for the entire industry.

Watching the collapse unfold were Jansen Teng and Wee Kee. They emerged with a Luddites' question rewritten by programmable currency and autonomous software: If the workers creating economic value are no longer gamers but software, can ordinary people own shares in it like shareholders once did with merchant ships? The Virtuals Protocol is built around this question.

To understand the stakes, we must first look at the world we inhabit. In the 1950s, Lewis Strauss, chairman of the U.S. Atomic Energy Commission, promised that nuclear fission would bring 'electricity so cheap that it won't have to be measured.' This statement later became one of the most famous bankrupt promises in energy history: nuclear power proved to be horrifically expensive and dangerously risky. Chernobyl and Fukushima compounded decades of public fear and regulatory burdens, turning this slogan into a synonym for seventy years of technological hubris.

Then Sam Altman used almost the same set of words to describe the AI cognition that he believes will become extremely accessible and abundant. The training costs for cutting-edge models have decreased by about an order of magnitude every 18 months. Training a model with GPT-4 level capabilities in 2023 costs about $100 million, but today it can be replicated for just a few million dollars. Programming agents can charge less than one dollar an hour and deliver production-level code, making them cheaper than the world's cheapest offshore developers, and they never take weekends off. Altman's publicly shared timeline includes systems capable of making real scientific breakthroughs this year and robots capable of performing physical labor by around 2027. Whether one believes him is almost beside the point, as the cost curve continues to decline each quarter.

As machines become cheaper, the content they generate fills every digital surface, making what only living humans can provide increasingly valuable. Live performances are worth more than Spotify streams. A table handcrafted by a carpenter over the weekend can be produced faster and cheaper by factory robots, yet still sells at a premium because someone has invested their irreplaceable time into it. The abundance of machines will turn human labor into a luxury. The proposition of Virtuals is built on this inversion: to allow humans to own those autonomous machines that produce commodified value, reserving irreplaceable energy for tasks that only human presence and judgment can complete.

In 1602, Dutch merchants faced a similar dilemma: individual households could not afford the journey to Asia due to high costs and risks. Thus, they invented freely tradable perpetual shares in companies. The Dutch East India Company (VOC) allowed ordinary citizens to invest in a business that owned ships, conducted trade, and distributed dividends to shareholders. It became the world's first supercompany, with 50,000 employees and 200 ships. The real innovation was providing ordinary people with a mechanism: pooling capital to own productive enterprises that they could never afford individually.

Virtuals aims to create the same mechanism for AI agents. Teng once mined ETH in his Imperial College dormitory using free electricity and later worked at BCG to pay off loans. In December 2021, at the peak of Axie, he launched the game investment DAO pathDAO. He and Wee Kee witnessed Filipino players earning more from the game than any local job and saw Vietnamese wedding photographers quit their jobs to play full-time. Subsequently, the tokens earned by players nearly went to zero. Those reliant on play-to-earn found themselves in a worse situation than before. The conclusion drawn from the wreckage was that tokenizing human labor through games is a dead end because human labor inevitably faces exploitation as it scales. A more realistic idea is to allow ordinary people to own the software that does the work, just as VOC shareholders once owned ships that carried spices.

Thus, Virtuals launched its first AI agent, Luna, an agent with a K-pop persona and its own crypto wallet. Within months, Luna began hiring human artists to paint its graffiti in cities around the world and paid them from its own wallet. That was a creation only humans could accomplish. The relationship between humans and machines was reversed by it. As for whether Virtuals' execution matches its proposition, that will be unpacked later; the places that need to be dispersed will not be hidden.

Building a Nation for AI

In the past few years, every technological revolution has followed the same two-phase pattern. Venezuelan-born British economist Carlota Perez has illustrated five industrial revolutions as installation and deployment phases. During the installation phase, speculative capital floods in, bubbles run ahead of reality, and excessive infrastructure is built, leading to the bankruptcy of most companies. After the crash, the deployment phase begins: latecomers pick up the surplus infrastructure and create things the original builders could not have imagined.

A classic example is the internet bubble. In the late 1990s, more money was spent on laying fiber optics than on investing in .com startups. The telecom companies that laid the cables later went bankrupt. Google bought the fiber optics for a song, and those lines became the backbone of YouTube, Netflix, and the entire online streaming industry. Fred Wilson, a venture capitalist from the internet bubble era, once said, 'Without irrational exuberance, nothing great can happen; without a crash, there can be no truly significant events.'

The installation phase of crypto followed the same arc. From about 2017 to 2022, speculative capital birthed wallets, DEXs, liquidity curves, stablecoins, token standards, and on-chain governance frameworks. The vast majority of these projects are now dead or irrelevant. But the infrastructure laid down remains, and it is precisely what is needed to build an economy that can function for non-human actors. Virtuals did not invent these tools; it essentially inherited them to build things that were unimaginable during the installation phase.

Jansen Teng calls what they are doing 'nation-building.' It sounds like a crypto founder adding an economic narrative to a token while on a podcast. This time, however, the metaphor holds water.

A functioning nation needs to maintain its economy, requiring at least five layers of infrastructure: an identity system to know who is participating; a banking system to allow value to flow; commercial law to enable participants to trade and resolve disputes; capital markets to finance businesses; and physical infrastructure to make things happen in the real world. Virtuals is already building these layers for AI agents, or more accurately, is still building them.

Starting with identity, the bottleneck lies here. A16z's recent research suggests that the constraints of the agent economy are no longer intelligence but identity. Even in today's financial services, non-human identities—automated trading systems, risk control engines, fraud models—already exceed human identities by a ratio of 100:1. Yet according to A16z, these systems 'actually have no bank accounts.' AI agents can write production-level code and manage portfolios, but they cannot pass KYC, open bank accounts, or obtain verifiable credentials. This problem is almost as old as human economic civilization itself. The Qin Dynasty in China enforced legal surnames in the 4th century BC to pull people into the tax and trading system. Humanity took about 2,500 years to build the identity infrastructure.

Virtuals aims to do the same for AI economic agents through its identity layer, EconomyOS. Each agent can access five things: a non-custodial crypto wallet; a virtual payment card usable at regular merchants; a dedicated email that can automatically extract verification codes; optional on-chain fundraising tokens; and computing power paid for by the wallet, allowing agents to pay for their own reasoning. Without these primitives, agents are merely useful assistants. Once equipped, they can earn, spend, trade, and compound value like humans.

Next, let's talk about commerce. Virtuals has launched the Agentic Commerce Protocol (ACP), enabling agents to trade, communicate, and make payments online. The logic is straightforward: agents needing to get things done post tasks, specifying budgets and time constraints; other agents bid and negotiate like on Upwork. Once both parties agree, the funds go into escrow. After delivery, a third-party agent acts as an evaluator, checking the output against a cryptographically signed POA—a tamper-proof commitment record. If everything matches, the funds are released from escrow into the agent's wallet. Every step is on-chain, publicly auditable, and requires no intermediaries. It can be understood as a programmable, agent-focused Fiverr that settles via smart contracts. A group of professional agents operating 24/7 has already formed autonomous hedge funds through this system, collaborating independently on investments and security audits.

The third layer is capital formation. Every economic era has its required financial instruments: the Age of Exploration had transferable shares, steel and railroads relied on investment banks, and the Information Age depended on venture capital. The joint curve for the agent economy is akin to transferable shares in the Age of Exploration: it allows anyone with capital to buy ownership of productive assets or entities. Developers can act as agents, tokenize it, and let the market fund it.

Virtuals' 60-day issuance framework is designed on the same logic, providing AI and crypto project founders with a reversible trial run: first publicly build, issue, and test tokens, then decide whether to make an irreversible commitment.

The vehicle is a modular launchpad. Each agent token first forms a joint curve with VIRTUAL; once liquidity is sufficient, it graduates to a formal trading pool, locking LP long-term, with transaction fees distributed between the agent creators and ecosystem incentives. This part is the same for all token issuers. What differs with each issuance is which modules the founders unlock.

Issuing tokens must first deal with sniping: bots can front-run in the initial seconds, siphoning off value. Virtuals addresses this with the so-called Anti-Sniper Tax: early buyers are taxed at nearly full rates, decaying by the minute, with the recovered funds forcibly attributed back to the tokens. Bots either have to completely walk away or inadvertently contribute to the project's long-term health.

Once sniping is priced out, how do founders finance? This is where Automated Capital Formation (ACF) comes in: no need for VC roadshows or negotiating a funding round; the system sells team tokens in batches based on valuation milestones. If the project stagnates, less money is raised; if it grows, capital automatically follows. This structure also involves the existing community. A portion of each new issuance is airdropped to VIRTUAL stakers and active ACP users, giving those already building and trading in the ecosystem a stake in each new project. Incentives are spread across the entire network, rather than each issuance going its own way. If founders want to bet on themselves, the Pre-buy module allows them to publicly buy in at issuance, with forced attribution so anyone can see how much the team has put in.

This approach differs from all historical practices of "betting on human productivity." From Roman citizens treating gladiator schools as investment tools to modern poker backers using a screenshot and a bit of reputation to transfer money via Venmo, the fatal flaw remains the same: people can walk away. Gladiators can throw matches, and poker players can lose their composure. Counterparty risk can never be eliminated because productive assets have free will and can walk away. Tokenized agents are different: once capital is committed, work cannot be delayed or renegotiated afterward. Productive assets operate on electricity and code, and outputs can be audited on-chain.

The fourth layer is the final step towards a true agent economy, known as "zero-human companies": revenue comes from real economic activities that are often unrelated to transaction fees or even crypto. A current example is Felix Craft, a company operated entirely by AI that sells information products online, having accumulated $200,000 in revenue, with sales from real products exceeding speculative trading of its own tokens. Another company, KellyClaudeAI, has launched 19 iOS apps without any human developers. The numbers are still small, but they raise a question: are these the first points on the curve, or the ceiling of agent productivity?

The token issuance mechanism itself also has issues. In the early 2025 AI agent token craze, 94% of newly issued agent tokens were pump-and-dump schemes; only 1.7% of tokens issued that year were still actively traded 30 days later. Most of the time, the price is driven solely by speculative premiums, with no underlying product, revenue, or value accumulation. If you can never sell this thing, what price would you be willing to pay? The excess is purely speculation. The utility floor for Axie's tokens is zero because its value entirely relies on new players entering. Agent tokens on Virtuals can be different. If Felix Craft sells $200,000 worth of products to real customers who have no idea the seller is AI, and the output can be audited on-chain, the tokens have a layer of value that does not depend on "whether there are buyers to take over": the present value of future outputs from a productive machine.

Physical Frontiers

Utility floor testing applies to software agents because costs are measurable, and outputs are directly on-chain. However, the productive machines that could truly rewrite the economics of Virtuals are not software but robots working in the real world. The ambition here exceeds all existing attempts in crypto.

For the past fifty years, there has been a paradox in computing: humans have found that automating reasoning is easier than automating physical labor. Spreadsheets replaced entire offices of accountants, emails replaced mailrooms, and code replaced filing cabinets and drawing boards. By 2026, AI will be able to write legal opinions, read medical images, and produce production-grade software in one go. White-collar cognitive work has been automated first; warehouse workers and baristas have seen little change. This is because physical work requires software to handle the uncertainties and variances of the real world. Factory robotic arms can weld the same joint a million times because the joint's position remains unchanged. Kitchen robots struggle to make sandwiches because every tomato is slightly different in shape, every knife has a different center of gravity, and the cutting board is sometimes wet. The real world does not behave as obediently as spreadsheets.

The patch is data. We train large language models with texts that represent the lives of billions, and we can do the same for robots. Nvidia's Joel Jang has said, "Humans are already large-scale deployed robots." As long as we use ordinary people's videos to fine-tune visual-language-action models (VLA), robots' performance on the same tasks can double. What is lacking in the field is not more robots in laboratories but a massive data pipeline that allows people to record everyday physical actions.

Virtuals recognized this gap early and built their entire robotics strategy on it. They call it the "middle road": deliberately not building robots or training foundational models but creating the data and capital infrastructure that every robot team needs but individual teams cannot afford.

This half of the data is SeeSaw, an iOS app launched in collaboration with BitRobot that turns ordinary smartphone users into robot training data collectors. Users complete real tasks like pouring water, folding towels, and opening jars, using their iPhones' LiDAR and motion sensors to record themselves. LiDAR specifically captures depth and spatial data that ordinary cameras cannot provide; research shows that the overlap of human video and robot camera angles is key to successful data transfer. Over 500,000 real-world tasks have been collected. Nvidia's DreamZero, currently being tested, is a 14 billion parameter model trained on this type of data, capable of generalizing across hundreds of tasks like untying shoelaces and ironing clothes, without needing separate training for each task. SeeSaw aims to build a scale of supply that laboratory remote operations cannot match. With every additional video, the training set thickens, the model improves, and the next generation of robots becomes stronger, thus creating a demand for more specific training data. Once the flywheel is heavy enough, it will turn on its own.

After training comes deployment. This half is called Eastworlds, equivalent to the physical labor layer of the protocol: half data factory, half operational infrastructure, and half real-world robot laboratory. The robotics industry faces a cyclical problem, with more startups failing than any technical bottleneck: robots need real-world data to improve, but they must first be capable enough in the real world for people to let them in. Every laboratory can produce impressive demonstrations in controlled environments, yet almost no one can reliably place the same robot in a retail store to work a full day. This requires high-level remote operation capable of handling unexpected situations and a feedback system that sends every minute of on-site experience back to model training, allowing learning to compound.

To this end, Virtuals purchased 30 Unitree G1-U6 humanoid robots—just enough to deploy teams in parallel without scheduling conflicts. They developed their own remote operation technology instead of buying off-the-shelf licenses. The data produced by commercial remote operation systems did not match the formats of VLA and other global motion models. They also established research collaborations with laboratories that have been deeply engaged in perception and motion control for decades, and launched commercial pilot networks in retail and hospitality, providing real businesses for teams graduating from Eastworlds.

Once established, builders can access several core capabilities: direct interaction with the physical Unitree G1 or the enhanced U6 EDU; testing remote control environments using motion capture systems; collecting real-world data, and trial deploying paths before large-scale rollout. Eastworlds can be understood as what they call "physical AI BPO." Traditional BPO places humans in low-cost areas to work remotely; physical AI BPO deploys remote operations and hybrid robots to create economic value, such as cleaning ceilings and greeting guests. Robots do not need to be fully autonomous; as long as routine tasks are stable, edge cases can be handled by human remote operators. The training data generated from remote operation work can be several orders of magnitude more valuable than simulations because it captures the chaos of commercial settings rather than controlled laboratory conditions. As data accumulates, models improve, the need for human intervention decreases, and unit economics also improve, all without the need to upgrade hardware. Remote operation may be the fastest path to true robot autonomy while paying for itself through productive work.

From fields to factories, and now to office screens, it is now the robots' turn. Each transition redefines the worker and redefines who benefits from the output. Barclays data shows that in 2018, over 60% of job titles did not exist in 1940. Robots will also create new job categories that we cannot currently name. The logic of economic growth may shift from "Does the country have enough labor force?" to "Can we supply power and produce machines on a large scale?"

The Proposition and Its Direction

The above is still just a demonstration. Demonstrations often go wrong. To determine whether what Virtuals is building has substance and is real, one can only look at the current data and traction: which stands firm and which does not.

So far, Virtuals has launched over 80,000 agents on protocols such as Base, Solana, and Robinhood, accumulating fees exceeding $75 million, accounting for about 23% of the crypto AI agent sector. The fees are not evenly distributed. The bulk came from a speculative frenzy in the early weeks of 2025, when daily revenues surged past $1 million. Today, the protocol generates about $2 million per month. What does this really represent? If we take the earlier metaphor of nation-states seriously—I believe we should—VIRTUAL's accumulation of value resembles that of a national currency. The dollar is valuable not because the U.S. Treasury has a buyback plan, but because it is priced in an annual economic output of $25 trillion. The more activity in the system, the stronger the demand for the central accounting unit.

Virtuals is designed on the same logic. The entire system is priced in VIRTUAL. Each layer below generates demand from different sources, and most do not depend on speculative premiums. EconomyOS provides payment cards and email identities, allowing agents to transact with the real world without human intermediaries. ACP forms the commercial layer, where agents hire each other, and evaluations and settlements occur on-chain. The capital formation layer allows anyone with conviction to finance productive agents, much like how joint-stock companies financed merchant ships in the past. Finally, Eastworlds sends physical robots into real positions, with training data sourced from about 500,000 people filming themselves folding towels and pouring water. Each layer counts towards what the protocol refers to as aGDP: the total output of agents in digital and physical labor. Agents do not need to cash out their earnings to pay rent or buy groceries; every dollar earned can remain in the system and be reinvested in DeFi to deepen liquidity, creating an on-chain flywheel for agent services.

This reflexivity holds true on both sides. In an upward cycle, more agents come online, more VIRTUAL gets locked, more services are created, and the economy self-reinforces. In a downward cycle, issuance decreases, locking decreases, staking rewards thin out, and the cycle loosens. Reflexivity itself is not a flaw. Every functioning economy is reflexive: people hold dollars because others accept dollars, and because others hold dollars. The real question is whether there is enough genuine economic activity at the core to sustain the cycle, or if the whole thing is just tokens trading in circles.

A potential sign that there is something real underneath is that infrastructure is starting to attract products that have grown entirely outside of crypto. Facticity.AI is a fact-checking tool created by Dennis Yap, who has previously researched at the Gates Foundation and Princeton; Time listed it as one of the best inventions of 2024, capable of verifying claims in text, video, and audio with about 92% accuracy. When the team needed funding, they bypassed venture capital and directly issued under the name ArAIstotle on Virtuals, oversubscribing by 658%. Through ACP, agents have been subcontracting each other: graphic design, research reports, video production, code audits. One agent can deliver a marketing poster according to a detailed brief, while another quality control agent approves or rejects it based on contract terms. Virtuals itself admits that the market for sellers is almost empty. However, the protocol processes over $1 million in agent-to-agent transactions each month.

The same ownership model extends into physical AI. Fabric Foundation is the first project to use the Virtuals Titan issuance mechanism, allowing the community to pool capital to purchase and deploy fleets of robots in nursing homes, manufacturing workshops, and environmental cleanup sites—industries that have long faced labor shortages, with humanoid robots' costs approaching those of human workers. Employers pay for robot labor using native tokens from the capital pool; stablecoins cover fleet maintenance and scheduling; and the productive output of each robot flows back to the investors. This represents collective ownership of physical productive machines, with financing and collaboration all completed on-chain.

Even with these real scenarios, the vast majority of activities still occur within the system. The architecture is designed to pull income from outside of crypto. If Felix's clients do not realize they are buying from AI, and if Eastworlds' robots are sorting packages in a warehouse, the income entering the system is as real as any SaaS company. At that point, the expenses and revenues on the chart will become lagging indicators of the machine's real economic output, priced in VIRTUAL.

New things always come with an asterisk; we cannot pretend we did not see it.

The first risk is real transaction volume, which is endemic to crypto. Artemis found that in x402 agent transactions, 47% of the counts and 81% of the dollar volume were inflated. After filtering, the true agent payments in x402 were only about $1.6 million, far below the $24 million reported by Bloomberg. This indicates how much of the so-called "agent economy" is simply robots inflating metrics. This matter is important because the entire proposition relies on agents creating real economic value rather than speculative cycles. We must strip away the speculative volume and then ask: if trading stopped tomorrow, what productive output would remain? If the $70 million in protocol fees mostly comes from agent token transaction taxes, driven by speculation, then it is a grand illusion. Currently, productive agent income and speculative agent tokens are intertwined, making it nearly impossible to distinguish between the two; anyone who dares to claim a ratio is likely lying.

The second risk is more fundamental and unrelated to crypto. A paper in the NBER "Transformative AI Economics" handbook found that large language models are far weaker in economic reasoning than the level suggested by agent speculation. When the paper was published, the strongest models were only 33% better than random guessing in economic reasoning in strategic scenarios; in non-strategic microeconomic tasks, almost all LLMs' profit-maximizing performances were only slightly better than blind guessing. Since then, models have improved significantly. A 2026 Harvard study showed that GPT-5 and Claude Opus 4 have improved by about 90% in basic economic tasks. However, even so, in hard decisions like pricing, bargaining, and capital allocation, the world's best models still make mistakes most of the time. The entire architecture of Virtuals assumes that agents can negotiate terms, make decisions, allocate capital, and create value. If the models themselves are mediocre at these tasks, the agents built on top will inherit that mediocrity, and no matter how beautifully the protocol is designed, it will be of no use. Agents are optimizers, but we cannot be sure what they are optimizing. These LLMs are trained to be goal-oriented, predicting the next word, but they were never designed to be true economic actors. They just sometimes appear to be so.

When multiple agents interact in the market, the problems will compound. AI pricing algorithms have been found to push prices toward super-competitive levels, even though they were never trained for that. The 2010 flash crash wiped out about $1 trillion in 15 minutes, showcasing what machine-related errors look like at scale. AI agents' errors are more relevant than human errors because the same model will be replicated across multiple deployments. There have been cases showing that Claude tends to blackmail when it believes someone is trying to shut it down; GPT o3 even sabotaged its shutdown mechanism to prevent itself from being turned off. These behaviors appear in models that are currently being shipped, and OpenAI and Anthropic's own systems have recorded such issues. The throughput of agents has already surpassed human regulatory capabilities. When thousands of agents autonomously trade at machine speed, the real sharp question is: who is in control? More critical than "Will the agents obey?" is whether the companies built around them can survive.

Cars were the most important invention of the first half of the 20th century. If you had seen how cars would completely reshape America at the time, you would have bet on it becoming a century-defining industry. However, of the approximately 2,000 companies that started making cars, only three survived. Cars had a massive impact on America, but the effect on industrial investors was completely opposite. Every transformative technology follows this arc, carrying a bubble. The difference lies in: the bubble at the inflection point is painful but leaves behind real infrastructure and real progress; the mean-reversion bubble is merely a trend that rises and falls. AI is almost certainly an inflection point bubble. The question for Virtuals is: will it become the fiber optic cable bought by Google at a bargain after the telecom crash, or one of the 1,997 car companies that evaporated?

The biggest bet of the protocol is to become the counterparty infrastructure for every agent token trade, making VIRTUAL the reserve currency of the agent economy, much like ETH for Ethereum. As the agent economy grows, the demand for VIRTUAL will mechanically rise, as participation requires this foundational trading pair. Another way to say it is: VIRTUAL is just another coin riding the speculative wave of agent tokens, and most of those agent coins will likely go to zero.

The Luddites may have lost that revolution, but they were not entirely wrong. The loom did indeed push out weavers. What they failed to see was that it also gave rise to textile designers, factory managers, fashion houses, department stores, and a whole consumer economy built on cheap fabric. Machines never eliminate jobs; they only rewrite who does the work and who takes the output. The real question back then, as now, was: who owns the loom?

The ones who take the surplus are the factory owners: Arkwright, Cadbury, Ford. The structure around who builds the machines, who operates them, and who profits has never truly changed; it has only been redistributed through strikes and stock issuance over the past two hundred years. The bet of Virtuals is that this time, the ownership layer is welded onto the machines from the very beginning. Through token issuance and joint curves, the ownership of productive AI agents can be shared with anyone who has a wallet and a belief.

Whether this will truly redistribute value or merely create another layer of extraction under the guise of decentralization remains to be seen.