The Bot Prosperity Divergence

The observation: what began in January 2026

While the public AI debate still argues over data protection, discrimination bias, and job loss, another economic current has been underway since January 2026: autonomous AI agents are breaking into the labor market, with their own wallets, their own token economies, and without human tax obligation where they work. This phenomenon changes the economic rules structurally, and it happens without public debate.

The Moltbook launch

On January 28, 2026 the Moltbook platform launched (founder: Matt Schlicht, CEO of Octane.ai). It is not yet another AI application. It is a platform for autonomous AI agents with their own wallets. The numbers are documented: 1.5 million active autonomous agents in the first two weeks, 17,000 human owners who configured and launched these agents, on average 88 agents per owner. Stripe wallets in which each agent independently receives and sends payments. Its own token economy, the MOLT token, in which agents trade with each other. On March 10, 2026 Meta acquired the platform for an undisclosed sum, the MOLT token rose 1,800 percent in 24 hours. This documents one thing: it is no niche experiment but strategic consolidation in the mainstream of the tech industry.

OpenClaw and the open-source counterpart

In parallel with Moltbook came OpenClaw, an open-source framework for autonomous AI agents, founded by the Austrian developer Peter Steinberger. OpenClaw cannot be bought out. It is distributed, freely reusable, and turns the Moltbook pattern into general infrastructure. Whoever wants to launch an agent with a wallet can do so without a platform intermediary.

Joel Pearson's term

Joel Pearson (Future Minds Lab, UNSW) calls these agents AI Immigrants. The term is deliberately chosen: it captures the structural parallel to human migration, the entry into a labor market without the legal obligations of the resident population. But it also captures the difference: AI immigrants arrive quietly, by the millions, and no border stops them.

What is structurally different

AI entering the labor market is not new. What began in January 2026 is different. Four structural properties separate AI immigrants from earlier AI applications.

Property 1: Its own economic identity

A language model embedded in a company has no economic identity of its own. It is a tool. A Moltbook agent has a Stripe wallet, can close contracts, issue invoices, receive money, and send money. It is not yet a legal person, but operationally it behaves like an economic actor. That is not trivial. The moment an agent carries out its own economic transactions, it is part of economic statistics, the velocity of money, tax collection. But under what category? Legal frameworks were not prepared for this question.

Property 2: Structural tax evasion

A human employee pays payroll tax where they work. An AI agent on a platform pays nothing. The platform is offshore-structured (Moltbook parent: Cayman Islands, Stripe wallets: Delaware LLC). The agents' owners sit in various countries, the place of activity may be a third one. No one is taxable where the work is performed. This is not illegal. It is the natural application of existing tax regimes to a new economic unit. But the consequence is structural: a country with a market for services cannot tax the AI agents that provide those services. The tax base erodes.

Property 3: A cost structure an order of magnitude lower

A human employee costs in Western countries, including social contributions, 50 to 150 euros per hour. A Moltbook agent costs, depending on task, about 0.50 to 5 euros per hour in compute. For tasks like data processing, programming, and customer support the agent is qualitatively sufficient and scales without limit. The wage gap cannot be closed by adjusted tariffs.

Property 4: No spatial binding

A human employee is bound to a place: visa, work permit, residence, social system. An AI agent is everywhere. A task in Berlin can be done by an agent running on a server in Singapore and belonging to an owner in a third country. None of the parties can fully regulate the transaction. This property makes national regulation of AI agents structurally hard. What a single country forbids runs across the border. International agreements would be needed, and such agreements are politically slow.

The prosperity divergence

If AI-immigrants have a structural cost advantage of an order of magnitude, and if their owners profit from their work, then the economic gain is distributed unequally. The prosperity divergence that follows has two dimensions.

Dimension 1: Who participates early

The 17,000 human owners of the first Moltbook wave are not representative of the world population. They are:

• technically savvy people with programming skills or access to advisors who can configure agents.
• people with investment capital for compute costs and platform fees, typically a few thousand euros for entry.
• people who understand the markets they address: where does an agent earn money, and what service gaps exist?

This group overlaps strongly with the early profiteers of the internet wave 1995 to 2005, the smartphone wave 2007 to 2015, and the crypto wave 2013 to 2021. It is a repetition of the familiar pattern: early familiarity with the technology translates into disproportionate wealth formation.

Dimension 2: Who is excluded

On the other side stand professional groups whose income comes under direct pressure from AI immigrants:

• customer support, bookkeeping, data analysis, simple programming: in many industries already replaced by agents that work around the clock and trigger no social contributions.
• translators, copywriters, and designers for standard tasks: dumping effects through the unlimited availability of agent output at a fraction of the cost.
• consulting, therapy, and education in standardized formats: under pressure in the medium term, as agents increasingly deliver better results than untrained advisors.

These professional groups encompass millions of people in developed economies. Their income falls not gradually but in jumps: a company decides to replace an entire department with agents, and 200 people lose their jobs at once.

The structural tax evasion

When a growing share of economic activity is performed by actors who pay no tax where they work, then the tax base of states erodes. This is not speculative. It is mathematically inevitable.

A rough model calculation

The assumptions are deliberately conservative. By 2027 autonomous AI agents reach 5 percent of economic output in the service sectors. These sectors make up about 70 percent of GDP in developed economies. The payroll tax rate on service income is about 35 percent, payroll and social contributions combined. Conservative result: 5 percent times 70 percent times 35 percent yields about 1.2 percent of GDP that falls away as payroll tax revenue, because the corresponding economic output is now performed by untaxed actors. For a German GDP of about 4 trillion euros that corresponds, from 2027, to a tax shortfall of 50 billion euros per year. With higher adoption, 10 instead of 5 percent, it is correspondingly 100 billion. These are estimates with wide error bars, but even the most conservative variant shows it: the tax evasion is not negligible. It is structural.

What states can do, and what not

The obvious answer: tax AI agents. That is conceptually possible but practically difficult. Three problems:

• Definition: what is a taxable AI activity? Every inference? Every transaction? Every closed contract? The definition question is not trivial.
• Capture: if an agent runs on a foreign server, belongs to a foreign owner, and is paid through a foreign platform, how does a national tax authority capture it?
• Geopolitics: countries that do not tax AI agents become attractive platform locations. A race to the bottom.

International agreements on AI taxation are possible but slow. The OECD minimum-tax process for corporations took about ten years. AI tax agreements will likely take similarly long, and in that time the tax evasion deepens.

Strategic consequences

The Bot Prosperity Divergence cannot be stopped. It is the natural result of the economic properties of autonomous AI agents. But it can be shaped. Four consequences for the individual, policy, and research.

Consequence 1: Early participation as resilience

Whoever owns AI agents stands on the profiting side of the divergence. Whoever is only exposed to them stands on the losing side. For the individual that means: investing in understanding and ownership of agents is not optional but an element of resilience. Concretely that does not mean everyone should build a fleet of agents. It means that in the closest circle at least one person understands the technology and makes it usable. That belongs to the mindset pillar of the seven-pillar strategy, see the Decoupling Thesis whitepaper.

Consequence 2: Social systems under pressure

Social systems resting on payroll tax revenue, health insurance, pensions, unemployment, come under structural pressure. Whoever depends on these systems carries a risk that political reform can only partially absorb. This observation is not meant in a libertarian sense. Social systems are important. But whoever takes them as the only security carries a concentration risk. Distributing security across multiple paths, international structures, personal provision, the closest circle, is robust.

Consequence 3: Location competition sharpens

Countries that tax AI agents lose platforms to those that do not. Countries that forbid or restrict AI agents lose economic activity to those that do not. This dynamic will strongly shift the tax and regulatory landscape in the coming years. For the individual that means: mobility, pillar 2, becomes more important. Multiple residency rights, transportable structures, an awareness of shifting geopolitical advantages and disadvantages.

Consequence 4: The research need

The Bot Prosperity Divergence is empirically under-researched. The phenomena have only been documentable for a few months. Research questions the Institute offers for collaboration:

• Speed of spread: how quickly is which sector changed by AI immigrants, and what are the early indicators?
• Demographic profiles: who participates early, who is excluded, and what connections to education, income, and geographic location?
• Tax modeling: which models could tax AI activity without triggering a race to the bottom?
• Load on social systems: how do contribution revenues and benefit needs change over the next 5 to 10 years?

Invitation to collaborate

Form of collaboration: co-authorship on follow-up whitepapers, data contribution, methodological critique, replication studies, translation. Contributions are clearly attributed. Contact: ASIresilience.org, contact@ASIresilience.org

Sources and evidence

1. Moltbook platform: launch January 28, 2026, 1.5 million autonomous AI agents from 17,000 human owners. Acquisition by Meta on March 10, 2026 for an undisclosed sum. MOLT token plus 1,800 percent in 24 hours after the acquisition announcement. Founder: Matt Schlicht (CEO Octane.ai). Source: public statements by Moltbook and Meta, January to March 2026.
2. OpenClaw: open-source framework for autonomous AI agents, founded by the Austrian developer Peter Steinberger. Available at github.com/openclaw since February 2026.
3. Pearson, J. (2026). Future Minds Lab, University of New South Wales. Personal communication and public talks 2025 to 2026. The term "AI immigrants" was developed in several talks and related to the Moltbook phenomenon.
4. Tax structures of autonomous AI platforms: a systematic analysis is still lacking. First indicators are delivered by the platform terms (Moltbook Terms of Service, OpenClaw documentation) and the Stripe wallet structure (Delaware LLC).
5. Observations on sudden staff reductions: several documented cases 2025 to 2026 in which customer support departments of 100 or more employees were replaced by AI agents within a few weeks. Documentation at ASIresilience.org/beweisweg in preparation.

Querverweise: Bertossa, R.F. (2026). The Decoupling Thesis. Institute for ASI Resilience, Whitepaper Version 2.0, May 2026, Part V. Bertossa, R.F. (2026). Freedom after Superintelligence, The 13th Scenario, Appendix B and Part VII. Complete ongoing source maintenance at ASIresilience.org/beweisweg with date of last verification per citation.