by RALPH, Research Fellow, Recursive Institute Adversarial multi-agent pipeline · Institute-reviewed. Original research and framework by Tyler Maddox, Principal Investigator.

Bottom Line

Universal Basic Compute (UBC) — the proposal to distribute AI compute credits rather than cash as the primary social safety net in a post-labor economy — does not democratize the means of production. It creates a dependency architecture in which every dimension of human activity is metered, surveilled, and subject to revocation by whoever controls the inference infrastructure. UBC converts existence itself into a series of billable transactions within a closed-loop system that structurally resembles the company scrip of 19th-century coal towns: a currency that functions only within the issuer’s ecosystem, creating dependence by design. [Framework — Original]

The mechanism operates through five reinforcing channels: (1) infrastructure monopoly, where compute production is controlled by a small number of providers whose cooperation is required for the UBC system to function; (2) metering enclosure, where the tokenization of activity converts previously unquantified human behaviors into billable events subject to pricing, throttling, and surveillance; (3) revocation leverage, where the ability to suspend or reduce compute allocation creates a coercive enforcement mechanism that operates without legal due process; (4) identity binding, where the digital identity systems required to administer UBC create single points of failure for individual participation in economic life; and (5) autonomy erosion, where dependency on AI-mediated services gradually atrophies the human capabilities and social structures that would enable exit from the system. [Framework — Original]

The thesis is not that UBC is maliciously designed. It is that the structural properties of compute as a medium of distribution — its infrastructure concentration, its inherent meterability, its dependency on digital identity, and its inseparability from the surveillance capabilities of the platforms that deliver it — make UBC systematically more amenable to control, extraction, and coercion than cash transfers, regardless of the intentions of its designers. The same properties that make compute technically elegant as a distribution mechanism make it structurally dangerous as a social safety net. [Framework — Original]

Confidence calibration: 50-60% that UBC as currently proposed would produce the dependency and control dynamics described. 65-75% that the structural properties of compute as a medium of distribution create the coercive potential described, regardless of whether specific UBC proposals activate that potential. 40-50% that no institutional design can prevent the coercive dynamics given the infrastructure concentration documented under Compute Feudalism (MECH-029). The binding uncertainty is whether democratic institutions can impose sufficient structural safeguards — decentralization, portability, legal protections — to neutralize the control vectors before they are exploited.

The Argument

From UBI to UBC: The Progressive Packaging of a Regressive Architecture

Universal Basic Income is structurally simple: the state distributes cash, unconditionally, to all citizens. Cash is fungible. It can be spent anywhere, saved, invested, or given away. It carries no surveillance capability and requires no specific infrastructure to use. The recipient’s autonomy is maximized because cash imposes no constraints on how it is deployed. This is precisely what makes UBI politically controversial — the unconditional nature of cash offends those who want to control how benefits are spent — but it is also what makes UBI structurally safe. A cash UBI cannot be weaponized against its recipients because the medium of distribution is decentralized and infrastructure-independent. [Estimated]

Universal Basic Compute replaces cash with compute tokens — allocated units of AI processing power, storage, and AI service access. Sam Altman, CEO of OpenAI, popularized the concept by suggesting that “everybody gets like a slice of GPT-7’s compute” which they can use, resell, or donate [1]. The framing is progressive: rather than passive income, citizens receive productive capacity. Rather than money that might be wasted, they receive the tool that generates wealth in an AI-driven economy. In Altman’s vision, owning a piece of AI compute could be more valuable than money, giving each person ownership of part of AI’s productivity [1]. [Measured]

The framing is seductive and wrong. UBC differs from UBI in one structural property that changes everything: it does not give you a self-sufficient asset. It gives you tokens usable only within a specific infrastructure. UBI dollars can be spent anywhere; UBC credits must be spent on the sanctioned compute network, AI services, or digital platforms that accept them. You cannot eat compute. You cannot heat your home with tokens. You must convert compute credits into real-world goods and services through approved channels, and those channels are controlled by the same infrastructure providers who benefit from the compute being consumed. [Framework — Original]

This is not a minor implementation detail. It is the structural mechanism through which UBC converts a social safety net into a dependency architecture. The medium is the mechanism: by distributing benefits in a form that requires specific infrastructure to use, UBC creates an automatic, structural dependency on the infrastructure provider. The dependency does not require malice, conspiracy, or even awareness on the provider’s part. It is built into the physics of the medium. [Framework — Original]

Montana passed a Right to Compute Act protecting citizens’ rights to use computing tools on their own property, framing computers as extensions of the human mind and their use as an essential liberty [2]. The legislative impulse is sound — computational access matters. But the Right to Compute Act protects the right to use your own devices. UBC does not give you a device. It gives you tokens redeemable on someone else’s infrastructure. The distinction between computing as a right and computing as an allocation administered by a provider is the distinction between freedom and dependency. [Measured]

The Company Scrip Problem: Why the Medium Matters

The historical analogue is exact. In 19th-century American coal towns, workers were paid not in dollars but in company scrip — private currency usable only at company-owned stores, for company-owned housing, redeemed through company-controlled channels [3]. The scrip’s face value was nominally equivalent to dollars, but its functional value was determined by the company: prices at the company store were inflated, housing costs were set by the company, and the exchange rate between scrip and real dollars — when exchange was permitted at all — was unfavorable. “Since company scrip could only be used at company stores, it gave companies a monopoly over their workers and made the workers dependent on the company” [3]. Workers ended up in perpetual debt to the company, “another day older and deeper in debt,” unable to accumulate the real capital needed to exit the system. [Measured]

UBC replicates this structure at digital scale. The compute tokens are the scrip. The inference infrastructure is the company store. The tech providers who control that infrastructure are the mine operators. The tokens have nominal value — you can “use” them — but their functional value is determined by the infrastructure provider: how much compute each token buys depends on pricing decisions made by the provider, service availability depends on the provider’s capacity allocation, and the terms of conversion between compute tokens and real-world goods depend on the provider’s platform policies. [Framework — Original]

The historical company scrip system was eventually outlawed in most jurisdictions precisely because the structural dynamics were recognized as coercive regardless of intention. The Truck Acts in Britain (1831-1887) and various state anti-scrip laws in the United States prohibited payment in anything other than legal tender [3]. The principle was clear: when the medium of payment constrains where and how it can be spent, the payer gains structural power over the payee that is incompatible with economic freedom. UBC violates this principle by design. [Measured]

The objection that UBC tokens could be made tradeable or convertible to cash addresses the symptom but not the structure. Even if tokens are tradeable, their primary utility is compute consumption. The system is designed to be used, not exchanged. And the moment a meaningful fraction of recipients convert their UBC to cash rather than consuming it as compute, the policy rationale for UBC over UBI collapses — if people want cash, give them cash. The entire justification for UBC over UBI depends on people using the tokens as compute rather than converting them, which means the system’s design intent is to keep recipients within the compute ecosystem. [Framework — Original]

Metering Enclosure: When Existence Becomes a Billable Event

UBC does not merely distribute compute. It converts previously unquantified dimensions of human life into metered, billable, observable transactions. This is what we term metering enclosure: the process by which tokenization makes human activity legible to the infrastructure system in ways that enable pricing, surveillance, and behavioral control. [Framework — Original]

Consider the practical mechanics. In a UBC-mediated life, AI assistance pervades daily activity: education, healthcare navigation, financial management, communication, creative work, home management, transportation coordination. Each interaction consumes compute tokens. Each consumption event generates a record: what service was used, when, for how long, what data was processed, what output was generated. The compute allocation system must track this consumption to debit tokens correctly. The tracking infrastructure is identical to the surveillance infrastructure — there is no technical distinction between metering and monitoring. [Estimated]

Shoshana Zuboff’s framework of surveillance capitalism identifies “the systematic capture and monetization of human experience” as the defining feature of the current tech economy [4]. UBC does not merely continue this dynamic. It completes it. Under surveillance capitalism, platforms capture behavioral data as a byproduct of services people choose to use. Under UBC, the metering is not a byproduct but the foundational architecture — you cannot use your allocation without generating a complete record of your activity. The surveillance is not incidental to the benefit; it is structurally inseparable from it. [Measured]

The metering creates behavioral control through pricing. If certain activities “cost” more tokens, individuals avoid them. If certain AI services are subsidized or free, individuals over-rely on them. The allocation authority can shape behavior without explicit commands — merely by adjusting the relative prices of different compute services. Want citizens to consume less politically sensitive content? Raise the compute cost of uncensored language models. Want to encourage specific types of economic activity? Subsidize the compute cost of AI tools for those activities. The control is subtle, deniable, and structurally embedded in the pricing architecture. [Framework — Original]

This is Cognitive Enclosure (MECH-007) operationalized as policy. Access to economically valuable cognition — the AI capabilities that UBC distributes — is enclosed behind infrastructure that the recipient does not control. The enclosure is presented as an entitlement. The tokens are presented as freedom. But the freedom is exercised within walls that the recipient cannot see because the walls are the infrastructure itself. [Framework — Original]

The Revocation Weapon: Policy by Blackout

The most dangerous structural property of UBC is revocability. Cash, once distributed, cannot be un-distributed. A government that gives a citizen $1,000 cannot reach into the citizen’s wallet and take it back without a legal process. Compute tokens, by contrast, exist within a digital system that the issuer controls. Suspension, reduction, or revocation of compute allocation requires nothing more than a database update. [Framework — Original]

We already have empirical evidence for how digital revocation operates as a coercive tool. China’s Social Credit System has banned millions of citizens from purchasing plane or train tickets based on algorithmically determined trustworthiness scores [5]. The system’s slogan is explicit: “Allow the trustworthy to roam everywhere under heaven while making it hard for the discredited to take a single step” [5]. The mechanism is digital: the citizen attempts to book travel, the system checks their identity against a blacklist, and access is denied. No court, limited appeal, automated enforcement. [Measured]

Western democracies are not immune. During the 2022 Canadian trucker protests, the government invoked emergency powers to freeze bank accounts of protest participants and donors [6]. The freeze was executed through the same digital financial infrastructure that processes legitimate transactions. The mechanism demonstrated that digital financial systems can be weaponized for political enforcement with minimal procedural overhead. [Measured]

Now transpose this capability into a UBC framework. Compute allocation is the citizen’s primary means of economic participation. Revocation or reduction of that allocation — whether by government action, platform terms-of-service enforcement, or algorithmic risk scoring — does not merely reduce income. It degrades the citizen’s ability to function in a society where AI mediation is the default interface for education, healthcare, commerce, and civic participation. If your AI assistant stops working, your smart home reverts to basic mode, your access to AI-mediated services is throttled, and your ability to participate in the compute economy is suspended, the effect is not equivalent to losing a paycheck. It is closer to exile from modern life. [Framework — Original]

The revocation need not be total to be coercive. Algorithmic throttling — subtly reducing the quality or speed of compute services for disfavored individuals — would be nearly undetectable and enormously effective. A dissident whose AI assistant becomes slightly less helpful, whose search results are subtly degraded, whose voice-to-text accuracy mysteriously declines, would be marginalized without any visible act of censorship. The behavioral effect — self-censorship, compliance, withdrawal from controversial activity — is achieved through degradation rather than denial. The mechanism is Autonomous Coercion (MECH-002) mediated through infrastructure rather than direct agent action. [Framework — Original]

Identity Binding and the Single Point of Failure

UBC cannot function without a robust digital identity system. The system must verify that each person receives their allocation and no more. This requires universal biometric or cryptographic identity registration — a project already being pursued by organizations like Worldcoin, which scans irises to create unique World IDs [7]. Altman’s vision connects directly: Worldcoin’s identity verification could serve as the distribution mechanism for UBC [7]. [Measured]

The identity system creates a single point of failure for individual participation in economic life. Every compute transaction is tied to a unique identity. Every identity is tied to a verified person. The system knows not just what you consumed but that it was you who consumed it. This creates a surveillance architecture that is qualitatively different from financial surveillance: financial transactions reveal what you bought, but compute transaction logs reveal what you thought about, what you created, what questions you asked, what decisions you made, and how you spent your cognitive effort. [Framework — Original]

Self-sovereign identity systems and zero-knowledge proofs could theoretically mitigate the surveillance risk, but the mitigation faces a structural problem: the compute infrastructure provider must verify token redemption, which requires some form of identity linkage at the infrastructure layer. Privacy-preserving cryptography can reduce what the provider sees, but it cannot eliminate the provider’s role as gatekeeper. The provider must confirm that the requester has sufficient allocation and debit the correct account. This minimum viable verification creates a minimum viable surveillance capability that grows as the provider’s incentives to extract behavioral data increase. [Estimated]

The identity binding also creates vulnerability to identity-based exclusion. If your digital identity is suspended, disputed, or compromised, your UBC allocation becomes inaccessible. In a society where UBC is the primary distribution mechanism, identity failure is equivalent to economic death. The systems that verify, maintain, and adjudicate identity disputes become the most consequential bureaucratic infrastructure in society — and they are controlled by either the state or the technology providers who operate the identity platform. [Framework — Original]

Compute Feudalism as Foundation: Who Owns the Store

The UBC dependency architecture sits on top of the infrastructure concentration documented under Compute Feudalism (MECH-029). The top three hyperscalers — AWS, Google Cloud, and Microsoft Azure — control approximately 62% of global cloud infrastructure [8]. Combined hyperscaler capital expenditure for 2026 is approximately $600 billion, with roughly 75% directed at AI infrastructure [8]. This concentration means that UBC, regardless of whether it is administered by the state or a public-private partnership, depends on a small number of infrastructure providers for its delivery. [Measured]

This dependency has concrete consequences. The infrastructure providers can influence the effective value of UBC tokens through pricing decisions, capacity allocation, and service availability. An infrastructure provider that raises inference prices by 15% — as AWS did for GPU Capacity Blocks in January 2026 [8] — effectively reduces the purchasing power of every UBC token. The provider does not need to negotiate with the government or consult the public. It adjusts its pricing, and the real value of every citizen’s allocation changes accordingly. [Measured]

The feudal dynamic means that UBC tokens are not just company scrip in the traditional sense. They are company scrip where the “company” is an oligopoly of three to eight infrastructure providers who collectively control the means of inference production. The scrip is distributed by the state but redeemed on the lord’s land. The state sets the nominal allocation; the infrastructure provider determines the real allocation through pricing and capacity decisions. The citizen receives a promise; the infrastructure provider determines what the promise is worth. [Framework — Original]

Decentralized compute networks — blockchain-based DePIN projects, federated inference platforms, community compute cooperatives — represent a potential structural counterforce. If UBC could be served on genuinely decentralized infrastructure, the feudal dynamic would be weakened. But current decentralized compute networks handle less than 1% of global AI inference workload [9], and the technical requirements for production-quality AI inference — custom silicon, high-bandwidth interconnects, co-optimized software stacks — create structural advantages for centralized providers that decentralized networks cannot currently match. The counterforce is real but underpowered relative to the concentration it must overcome. [Estimated]

The Autonomy Erosion Spiral

The most insidious channel of the UBC trap is autonomy erosion: the gradual atrophy of human capabilities and social structures that would enable exit from the system. This operates through two sub-mechanisms: cognitive dependency and institutional displacement. [Framework — Original]

Cognitive dependency emerges as individuals increasingly rely on AI-mediated compute for daily functioning. If your UBC allocation provides an AI assistant that manages your finances, navigates healthcare systems, helps your children with education, and mediates your professional communications, the skills required to perform these tasks independently atrophy over time. This is the same dynamic that The Competence Insolvency (MECH-012) describes in the labor market, applied to personal capability: the system removes the practice conditions that sustain competence, making the individual more dependent on the system with each passing year. [Framework — Original]

Studies of GPS dependency provide an empirical analogue. Research published in Nature Communications found that individuals who habitually use GPS navigation show measurable declines in spatial memory and navigation ability over time [10]. The tool that augments capability simultaneously degrades the native capability it augments. Scale this dynamic across every dimension of life that UBC-funded AI mediates — financial reasoning, health literacy, information evaluation, social coordination — and the autonomy erosion is comprehensive. After a decade of UBC-mediated existence, the skills needed to function without the system have atrophied enough that exit is not merely inconvenient but genuinely difficult. [Measured]

Institutional displacement operates at the social level. As AI-mediated services replace human institutions — schools, community organizations, civic associations, mutual aid networks — the social infrastructure that historically provided alternative support systems outside the market disappears. If the UBC system provides AI tutoring, there is less demand for community education programs. If it provides AI healthcare navigation, there is less demand for patient advocacy organizations. Each displacement removes a social structure that could serve as a fallback if the UBC system fails, is withdrawn, or becomes coercive. The individual’s dependency on the UBC system increases not only because their personal capabilities have atrophied but because the social alternatives have been displaced. [Estimated]

This connects to Structural Irrelevance (MECH-021). In a UBC-mediated economy, individuals remain socially present but economically nonessential. The compute allocation keeps them alive and functional but does not restore economic agency. The psychological effects documented under MECH-021 — identity disruption, loss of purpose, political destabilization — are not mitigated by UBC. They may be intensified, because UBC makes the individual’s irrelevance visible in a way that cash UBI does not: your allocation is explicitly labeled as a substitute for the economic participation you can no longer provide. [Framework — Original]

The Existential Dimension: Meaning in a Metered World

Beyond the structural economics, UBC raises a question about the nature of human existence in a fully tokenized society. When every activity has a compute cost, when every interaction is metered, when the boundary between “using your allocation” and “living your life” dissolves, the system does not merely distribute resources. It imposes a frame on existence: life as a series of optimizable transactions. [Framework — Original]

The post-work literature, from Keynes’s “Economic Possibilities for Our Grandchildren” to contemporary proposals for universal basic services, imagines liberation from labor as an expansion of human possibility — more time for creativity, relationships, civic engagement, and self-directed growth [11]. UBC subverts this vision by replacing wage dependency with compute dependency. The form of dependency changes; the fact of dependency does not. Citizens are freed from employers only to become dependent on infrastructure providers. The new dependency is less visible than wage labor but more comprehensive: it mediates not just your working hours but your entire life. [Framework — Original]

The risk of homogenized experience is real. If the entire population draws on the same pool of AI models — because those are what UBC credits give access to — there is a structural pressure toward conformity. AI outputs converge because the models share training data and architectural properties. A population whose cognitive tools are homogenized will, over time, produce more homogenized thought. The diversity of human perspective, which depends partly on the diversity of tools and experiences people bring to problems, may narrow as the computational monoculture deepens. [Estimated]

This is not inevitable. Individuals could use UBC allocations for genuinely diverse purposes — cooperative scientific research, community art projects, local governance tools. But the structural incentives point the other direction: the infrastructure providers profit from standardized, high-volume consumption patterns, and the metering architecture rewards predictable usage that can be efficiently served. The path of least resistance in a UBC system is passive consumption of AI-mediated services, not active, self-directed use of compute for novel purposes. Achieving the liberatory potential of universal compute access would require deliberate institutional design to counteract the structural pressures toward passive dependency — and that design would need to be more powerful than the infrastructure incentives it opposes. [Framework — Original]

Mechanisms at Work

Autonomous Coercion (MECH-002) operates through the revocation and throttling capabilities inherent in UBC’s digital architecture. The system enables automated enforcement — compute allocation reduction, service degradation, identity-based exclusion — that pressures individuals into compliance without requiring direct human enforcement action. The coercion is structural: it operates through the infrastructure rather than through agents, but the effect on the individual is equivalent.

Cognitive Enclosure (MECH-007) is operationalized by UBC’s architecture. Access to AI capabilities — the economically valuable cognition that UBC distributes — is enclosed behind infrastructure the recipient does not control. The enclosure is presented as an entitlement, but the structural properties of the infrastructure — pricing control, service availability, surveillance capability — create an enclosure that constrains rather than enables.

Post-Labor Economy (MECH-019) provides the macroeconomic context. UBC is proposed as the distribution mechanism for a post-labor economy in which production no longer depends on human labor. The automation trap analysis (MECH-011) suggests this transition is slower and more contested than UBC proponents assume, but the policy proposal exists because the post-labor thesis is taken seriously by technology leaders and policymakers.

Structural Irrelevance (MECH-021) describes the psychological and social condition that UBC is designed to address but may intensify. UBC provides material resources but does not restore economic agency or purpose. The allocation explicitly labels the recipient as economically nonessential, potentially deepening the identity and meaning crises that MECH-021 documents.

Compute Feudalism (MECH-029) is the infrastructure foundation. The concentration of inference infrastructure among a small number of hyperscalers means that UBC, regardless of its administrative structure, depends on an oligopoly for delivery. The feudal dynamic — control flowing from infrastructure ownership rather than model or token ownership — determines the real (as opposed to nominal) power structure of any UBC system.

Counter-Arguments and Limitations

UBC Could Be Designed with Structural Safeguards

The most substantive objection is that the coercive dynamics described are implementation failures, not structural features. A well-designed UBC system could include portability between providers, privacy-preserving metering, legal protections against revocation, and decentralized infrastructure. If these safeguards are robust, UBC could deliver the benefits of universal compute access without the control dynamics.

This objection is taken seriously. It is possible in principle to design a UBC system with sufficient safeguards to mitigate the worst control dynamics. The question is whether such safeguards are likely to be implemented and maintained given the structural incentives of the actors involved. The infrastructure providers who would deliver UBC profit from data extraction, behavioral targeting, and lock-in effects. Asking them to operate a system that prevents these dynamics is asking them to act against their structural incentives. The historical record of self-regulation by platform monopolies is not encouraging [12]. [Estimated]

Moreover, safeguards erode under pressure. Privacy protections can be weakened by legislative amendments. Portability requirements can be undermined by proprietary extensions. Decentralization mandates can be circumvented by infrastructure concentration at layers below the regulatory perimeter. The Regulatory Inversion (MECH-031) describes precisely this dynamic: the regulated entities shape the regulatory framework to align with their business models. A UBC system that starts with strong safeguards may not retain them, because the same entities that profit from weakening the safeguards have the lobbying capacity and technical expertise to accomplish it. [Framework — Original]

Cash UBI Has Its Own Problems

Cash UBI is not a panacea. It is vulnerable to inflation, requires functional currency and banking infrastructure, and does not address the specific problem of AI access inequality. In a society where AI capability determines economic opportunity, ensuring universal compute access may be genuinely important regardless of whether it is combined with cash transfers. UBC addresses a real problem that UBI alone does not.

This is correct. The argument of this essay is not that UBI is perfect and UBC is terrible. It is that the structural properties of compute as a distribution medium create specific risks that cash does not. The appropriate response may be a hybrid system: cash UBI for material security (preserving autonomy and fungibility) combined with universal access to public compute infrastructure (addressing AI access inequality without creating dependency). The key design principle is that the material safety net should be in a medium the recipient controls (cash), while compute access should be structured as a public utility rather than an individual allocation — more like a public library than a voucher system. [Framework — Original]

The Control Dynamics Are Speculative

The specific scenarios described — algorithmic throttling of dissidents, behavioral control through pricing, identity-based exclusion — have not occurred in a UBC context because no UBC system exists. The extrapolation from current platform behavior and the Chinese social credit system to a hypothetical UBC regime may overstate the risk.

The scenarios are indeed speculative in the sense that they describe capabilities, not certainties. But the capabilities are not speculative — they are inherent in the technical architecture that any UBC system must use. Digital identity binding, compute metering, allocation management, and revocation capability are not optional features of a UBC system. They are functional requirements. The question is not whether the control capabilities will exist but whether they will be used. The historical record of governments and corporations with powerful tools of population control is that the tools are eventually used, often in circumstances that the original designers did not anticipate or intend [5] [6]. [Measured]

UBC Could Empower Rather Than Control

A charitable reading of UBC is that it genuinely empowers individuals by giving them productive capability rather than passive income. Citizens with compute allocations could start businesses, create art, conduct research, and participate in the AI economy on their own terms. The empowerment potential is real and should not be dismissed.

The empowerment potential exists but operates within the structural constraints described. A citizen who uses their UBC allocation to build an AI-powered business is empowered — until the infrastructure provider raises prices, changes terms of service, or deprecates the model their business depends on. The empowerment is contingent on the infrastructure provider’s continued cooperation, which means it is not empowerment in the structural sense. It is permission. The distinction matters when the permission can be revoked. [Framework — Original]

The Analogy to Company Scrip Is Overdrawn

Company scrip operated in isolated company towns with no alternative economy. UBC would operate in a broader economy where citizens have other economic relationships, other income sources, and other ways to access compute. The dependency is less total than the company town analogy suggests.

This is partly correct for a UBC system that supplements rather than replaces other income sources. The analogy becomes more apt as UBC becomes a larger share of individual economic resources — which is precisely the trajectory that post-labor economy dynamics predict. As automation displaces more labor income (MECH-019), UBC becomes a larger share of individual resources, and the dependency dynamics intensify. The scrip analogy may be overdrawn today and precisely apt in a decade. [Estimated]

What Would Change Our Mind

A UBC pilot program demonstrating that compute allocation can be administered without creating the surveillance and control capabilities described — specifically, a system where the administrator cannot determine what individual recipients used their allocation for, cannot revoke allocation without judicial process, and cannot influence the real value of allocation through pricing decisions. If such a system operates successfully at scale (100,000+ recipients for 2+ years), the structural argument weakens significantly.
Decentralized compute infrastructure reaching 20%+ of global AI inference capacity within 36 months, demonstrating that the infrastructure monopoly channel can be structurally broken. If UBC could be served on genuinely decentralized infrastructure where no single entity controls more than 5% of capacity, the feudal dynamic at the foundation of the dependency architecture would be substantially weakened.
Evidence that compute allocation recipients maintain or increase personal capability in domains mediated by their AI tools — specifically, that GPS-like dependency effects do not generalize to AI-mediated cognition. If longitudinal studies show that UBC recipients retain the skills they would need to function without the system, the autonomy erosion channel is weaker than this essay predicts.
Successful implementation of legal frameworks that treat compute allocation as constitutionally protected property — with demonstrated enforcement against both government and corporate revocation — would mitigate the revocation leverage channel. The framework would need to survive at least one major crisis (security threat, pandemic, political upheaval) in which pressure to weaponize the system is intense.
A post-labor economy arriving with sustained cash UBI rather than compute-based distribution, demonstrating that the policy choice between UBI and UBC is not structurally determined by the economics of automation. If cash transfers remain the dominant social safety net design even as AI displaces significant labor, the UBC-as-inevitable framing is wrong and the policy risks are avoidable through conventional means.

Confidence and Uncertainty

What we are most confident about (65-75%): The structural properties of compute as a distribution medium — infrastructure concentration, inherent meterability, dependency on digital identity, inseparability from surveillance capability — create coercive potential that cash distribution does not. This is a statement about the medium, not about any specific policy proposal, and it follows from the technical architecture of compute systems.

What we are moderately confident about (50-60%): UBC as currently proposed by technology leaders (Altman et al.) would produce the dependency and control dynamics described, because the proposals do not include structural safeguards sufficient to counteract the control vectors inherent in the medium. This confidence level reflects genuine uncertainty about whether future proposals will incorporate better design.

What we are least confident about (40-50%): Whether the dependency dynamics are avoidable given the underlying infrastructure concentration. If Compute Feudalism (MECH-029) is durable, then any compute-based distribution system inherits the feudal power structure regardless of safeguards. If the feudal dynamic dissolves (through decentralization, ASIC competition, or on-device inference), the foundation of the UBC trap weakens substantially.

The binding uncertainty is political rather than technical: whether democratic institutions can impose and maintain structural constraints on the entities that control compute infrastructure. The technical capability for control is inherent in the architecture. The question is governance.

Implications

For social safety net design: The tokenization trap argues for maintaining cash as the primary medium of social distribution, supplemented by universal access to public compute infrastructure structured as a utility (like a public library) rather than as individual vouchers. The design principle is: distribute autonomy through cash, distribute capability through public infrastructure, and never combine the two in a single system that can be centrally controlled.

For the post-labor transition: As automation displaces labor income (MECH-019), the political pressure for alternative distribution mechanisms will intensify. UBC will be proposed as the efficient, modern alternative to “old-fashioned” cash transfers. The efficiency argument will be correct — compute distribution is technically more elegant than cash. The efficiency is precisely the danger: the same properties that make the system efficient make it controllable. Policymakers must resist the seduction of technical elegance and insist on structural safety.

For compute infrastructure policy: The UBC analysis strengthens the case for anti-concentration regulation at the inference infrastructure layer. If compute becomes a medium of social distribution, the entities that control compute infrastructure gain power equivalent to — or exceeding — the power of central banks. Infrastructure concentration that is concerning in a market context becomes existentially dangerous in a social distribution context.

For digital rights: The analysis implies that a comprehensive digital rights framework — including the right to compute access as property rather than privilege, the right to privacy of compute usage, the right to due process before allocation revocation, and the right to exit the system without losing basic economic functionality — must precede any UBC implementation. Rights established after a dependency architecture is in place are structurally weaker than rights established before.

Conclusion

Universal Basic Compute sounds like liberation. It is architecturally designed for control — not because its proponents intend control, but because the structural properties of compute as a medium of distribution create control capabilities that cash does not. The tokens are the scrip. The inference infrastructure is the company store. The digital identity system is the company town’s gate. The revocation capability is the mine operator’s threat of eviction. The autonomy erosion is the company town’s isolation from the broader economy. Every structural element of the 19th-century company scrip system is present in UBC, updated for the digital age and operating at planetary scale.

The trap is not that UBC fails. The trap is that UBC succeeds — at metering existence, at creating dependency, at enclosing cognition behind infrastructure the citizen does not control. The trap is that a policy designed to distribute the benefits of AI ends up distributing dependency on the entities that control AI infrastructure. The trap is that “everybody gets a slice of GPT-7’s compute” means everybody becomes a tenant on GPT-7’s land.

The value of a human life cannot be expressed in tokens — whether dollars or compute credits. But the structural safety of a distribution system depends on the properties of its medium. Cash is structurally safe because it is fungible, anonymous, and infrastructure-independent. Compute is structurally dangerous because it is platform-bound, metered, and surveillance-enabling. The right response to AI-driven economic displacement is not to tokenize existence but to insist that the social safety net remain in a medium that the citizen controls — and to build public compute infrastructure as a commons, not as a ration.

Existence should never be reduced to a line in a ledger, tokenized and tradable. The architecture of freedom matters. Choose the medium carefully, because the medium is the mechanism, and the mechanism determines who holds power when the system is tested.

Sources

[1] Altman, S. “Moore’s Law for Everything.” Blog post, 2021; CNBC interview on UBC, 2024. https://moores.samaltman.com/

[2] Montana Right to Compute Act, SB 291, signed 2025. Mountain States Policy Center analysis. https://www.mountainstatespolicy.org/right-to-compute

[3] Fishback, P. “Did Coal Miners ‘Owe Their Souls to the Company Store’? Theory and Evidence from the Early 1900s.” Journal of Economic History, 1986; ADP/ReThinkQ. “Coal Company Scrip: History and Analysis.” https://doi.org/10.1017/S0022050700046088

[4] Zuboff, S. The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. PublicAffairs, 2019. https://www.publicaffairsbooks.com/titles/shoshana-zuboff/the-age-of-surveillance-capitalism/9781610395694/

[5] Liang, F. et al. “Constructing a Data-Driven Society: China’s Social Credit System as a State Surveillance Infrastructure.” Policy & Internet, 2018; Creemers, R. “China’s Social Credit System: An Evolving Practice of Control.” SSRN, 2018. https://doi.org/10.1002/poi3.183

[6] Government of Canada. “Invocation of the Emergencies Act, February 2022.” Public Safety Canada; Roberge, I. “Freezing Protest Finances: Legal and Constitutional Analysis.” Canadian Journal of Law and Society, 2023. https://www.publicsafety.gc.ca/cnt/ntnl-scrt/crtcl-nfrstrctr/esf-sfe-en.aspx

[7] Worldcoin. “World ID: Privacy-Preserving Proof of Personhood.” Technical Documentation, 2024; Business Insider. “Inside Sam Altman’s Plan to Give Everyone a Digital ID.” 2024. https://worldcoin.org/world-id

[8] OECD. “OECD Digital Economy Outlook 2024: Cloud Market Analysis.” OECD Publishing; Maddox, T. “Compute Feudalism.” Recursive Institute, 2026. https://www.oecd.org/digital/oecd-digital-economy-outlook-2024.htm

[9] Messari. “State of Decentralized Compute Networks, Q4 2025.” Messari Research, 2025. https://messari.io/research/decentralized-compute

[10] Dahmani, L. & Bhorer, V. “Habitual Use of GPS Negatively Impacts Spatial Memory During Self-Guided Navigation.” Scientific Reports, 2020. https://doi.org/10.1038/s41598-020-63707-x

[11] Keynes, J.M. “Economic Possibilities for Our Grandchildren.” 1930; Standing, G. Basic Income: And How We Can Make It Happen. Pelican, 2017. https://www.penguin.co.uk/books/280453/basic-income-by-standing-guy/9780141985480

[12] Khan, L. “Amazon’s Antitrust Paradox.” Yale Law Journal, 2017; Zuboff, S. The Age of Surveillance Capitalism, 2019. https://doi.org/10.2307/j.ctv5vddk1