The Body Returns to the Network

Human uniqueness after the user

by Sasha Shilina

In early September, AI agents moved deeper into ordinary financial infrastructure. Aave opened an MCP server through which AI assistants can inspect positions, simulate changes and prepare unsigned transactions. Mastercard introduced Agent Connect, allowing agents to exchange product data, assemble carts and participate in authorized purchases. Days later, Visa, Mastercard and Ant International were reported to be developing shared methods for identifying and verifying purchasing agents, while India was preparing an agent registry for UPI.

An agent can now carry a task across several systems: gather information, call services, choose among alternatives, prepare an economic action and preserve state between interactions. Human authorization may still appear somewhere in the chain, although its relationship to each individual action becomes harder to infer from the action itself.

For most of the internet’s history, this ambiguity remained tolerable. An account posted, a wallet signed, a customer bought something, and the presence of a person somewhere behind the interface could usually be assumed. Digital systems were built around representations of people while rarely needing to establish whether ten accounts represented ten people, three people or one. Agentic systems make that approximation increasingly unreliable.

After the user

The user has been one of computing’s most durable abstractions. It joins an account, a person, a source of intention and an interface endpoint under the same name. Software documentation can say that a user clicked, searched, paid or voted because these senses of the word have historically remained close enough to function together.

An agent loosens those connections.

A person may ask one model to research a purchase. The model can call specialized services, retrieve merchant data, compare prices and prepare a transaction. Another person may maintain several persistent agents with different permissions, wallets, histories and tasks. An organization can deploy thousands. Some agents may communicate mostly with other agents and expose only fragments of their activity to the human who configured them.

The population visible to a network now has several possible denominators. Accounts, wallets, agents and people can produce different answers to the apparently simple question of how many participants are present.

As soon as a system distributes voting rights, grants, reputation, free usage, priority access or community privileges, it needs a unit of allocation. If one human can generate a large population of operational accounts and agents at low cost, the account gradually loses its weak historical connection to personhood.

The old figure of the user compressed these relations into one convenient object. The emerging internet has to represent them explicitly.

When information lost its body

N. Katherine Hayles gives this problem a longer history. In How We Became Posthuman, she follows the history through which twentieth-century cybernetics came to treat information as a pattern conceptually separable from the material forms carrying it. Her account begins with the Macy Conferences and extends through artificial life, cybernetic theories of subjectivity and the posthuman imagination. One of its central concerns is the intellectual labor required for information to “lose its body.”

Digital identity developed inside a similar abstraction. A username does not contain a face. An email address does not reveal a body. A public key says nothing about nationality, age or biography. Blockchain systems pushed this principle unusually far because a network could verify control through a cryptographic signature while remaining largely indifferent to the organism holding the key.

That indifference made pseudonymous coordination possible at enormous scale. A person could participate through several wallets and several names; strangers could establish control and authorization without asking an institution to certify a complete civil identity. Cryptography became increasingly precise about keys, signatures and permissions.

Human cardinality stayed outside the claim. A valid signature can establish that a key authorized an action. It cannot establish how many people control the surrounding set of keys. Ten thousand addresses may belong to ten thousand people, to one exchange, to a trading system, or to one person operating a large collection of agents.

Generative AI extends this uncertainty from identity into behavior. Text, conversation, research, navigation and parts of economic decision-making can circulate with decreasing dependence on continuous human attention. The informational separation between person and representation acquires an operational dimension: fragments of agency can leave the person and continue acting elsewhere.

Dividuals with agency

In his 1990 Postscript on the Societies of Control, Gilles Deleuze called the datafied subject a “dividual”: a person decomposed into codes, samples and data that determine access. Identity becomes distributed across machine-readable fragments.

Agentic computing gives those fragments persistence and the capacity to act. Purchasing preferences can operate through one system, professional context through another, wallet permissions through a third. A human body, several cryptographic keys, accounts and models may all participate in the same chain of action.

Crypto had already allowed one person to inhabit many addresses while protocols saw each address independently. Agents make those identities operational: they can maintain them, act through them and accumulate histories of their own. Recovering a unit corresponding to one human life becomes a technical operation. Some systems have little reason to perform it. Others depend on precisely that unit.

Counting humans

James C. Scott called a related institutional operation legibility. In Seeing Like a State, he examines how institutions convert complex populations into forms they can administer. Standardized names, cadastral maps, measurements, registries and censuses produce simplified representations through which a state can count, tax, regulate and plan.

Digital systems perform their own acts of legibility. A blockchain recognizes addresses and transactions. A proof-of-stake system recognizes stake. An advertising platform recognizes profiles and impressions. A social network recognizes accounts and interactions. Each architecture constructs a population through the entities it can see.

Networks can measure activity with extraordinary precision while remaining uncertain about the number of people generating it. Transactions, API requests, wallets and messages are easy to count. Human beings become harder to enumerate once each person can distribute activity across numerous computational actors.

The denominator is a political choice. Corporate governance counts shares. Proof-of-stake governance counts capital. Social platforms count accounts. Electoral systems attempt to count eligible persons. Each unit produces a different distribution of weight.

The body becomes useful again

Human bodies are resistant to cheap replication.

Accounts can proliferate. Wallets can be generated almost indefinitely. Models can be copied and agents instantiated repeatedly. Human biological existence remains attached to a continuous organism with finite time, attention and physical presence.

Biometrics can supply an upper bound on forms of multiplication that accounts and wallets cannot constrain on their own.

This is the territory Humanode has been developing through cryptobiometric verification, where the useful claim concerns the presence of a unique living human. In May 2026, Humanode extended the idea to agentic systems with Agentlink, which allows an agent to be associated with a human who has already passed uniqueness verification through Biomapper. The first integrations applied this relation to services where free access, grants or other benefits would otherwise be easy to farm through disposable wallets.

A service may have no operational reason to learn someone’s legal name, passport number or nationality. It may still need to know that a particular entitlement corresponds to one unique person, or that several agents ultimately descend from the same human principal.

An agent can maintain its own credentials, permissions and reputation, while a human connection can be invoked where eligibility, accountability or allocation depends on personhood. A harder political question follows: how much of the person must become visible in order to establish uniqueness?

Minimal legibility

The return of biometrics to network architecture carries a long and uncomfortable history. Fingerprints, anthropometry, passports, photographs, facial recognition and population registries have repeatedly linked bodily measurement to institutional power. Once bodily characteristics become durable identifiers, data collected in separate contexts can be joined, populations sorted and participation conditioned on visibility.

A proof-of-personhood system therefore has to manage its appetite for information. Scott’s concept can be pushed toward what we might call minimal legibility: giving a system enough knowledge to verify the property relevant to a particular interaction while keeping unrelated characteristics outside its field of vision. For a Sybil-resistant vote, uniqueness may be sufficient. A free allocation may need to know whether the same unique human has already claimed it. An agent service might need a human anchor for rate limits or accountability. None of these operations inherently requires a civil identity profile.

Bodies, however, are messy credentials. Biometric systems encounter false matches, failed recognition, changing physical conditions, accessibility constraints, coercion and questions of revocation. Their databases can become exceptionally sensitive infrastructure. Cross-context linkability can turn a narrow uniqueness proof into a much more expansive identity system.

Cryptography can constrain this visibility. The design question is how much a system should be allowed to know. Proof of personhood lives inside that limit: making a person countable while leaving the rest unreadable.

The political unit

Humanode’s Vortex turns cardinality into governance. The Humanode Foundational Charters anchor the governing order in verified human uniqueness and equal political worth. The architecture developed through Vortex includes proposals, delegation, chambers and other governance mechanisms, with human uniqueness supplying the base membership layer. By August, Vortex Alpha v0.3 had implemented a reporting and Court system with twelve-Governor juries, evidence boundaries, structured findings, appeals and enforcement contracts, still held behind an activation gate pending further review.

Political orders have always had to decide what counts as a member: the household, the property owner, the citizen, the shareholder, the token holder. Each produces a different distribution of political power. Agents make that choice newly difficult when accounts can be populated and operated at machine scale.

Numerical majority becomes harder to interpret as a population of independent human participants. Systems that choose verified human uniqueness as their denominator make an explicit constitutional decision about the entity that receives base political standing.

A governance system could permit agents to conduct research, model proposals, summarize debate, formulate amendments and participate in discussion while keeping membership or final voting weight linked to verified humans. Human agency could coexist with extensive delegated machine activity inside the same political process.

Machines become legible too

The same pressure now applies to machines. Payment companies need ways to establish which AI agent initiated a purchase, whether the agent is legitimate and which permissions it carries. The Visa, Mastercard and Ant International initiative concerns shared methods for agent identification and verification, while India’s proposed registry would make agents visible within the UPI payment environment.

A human can possess a uniqueness proof and authorize several agents. Each agent can maintain its own identity, credentials, transaction history and reputation. Applications can request proof of a human connection where access depends on human cardinality and ignore that connection elsewhere.

What remains difficult is the relation between identities: human to agent, agent to wallet, agent to organization, one agent to another. Personhood becomes one layer among several, carrying properties that computational actors cannot reproduce through simple proliferation.

After the human internet

Much of the cultural argument around AI has focused on imitation, especially whether models can perform activities previously treated as evidence of human intelligence. Agentic systems create a more institutional problem because participation in digital life is increasingly organized around attribution, permission, delegation and standing.

Automated markets may care primarily about authorization and solvency. Communities distributing scarce benefits may need a reliable human denominator. Governance systems may reserve certain rights for people while giving agents extensive operational freedom.

The network ahead contains people, wallets, models, organizations and persistent agents whose boundaries overlap. A person can delegate parts of their agency to several machines; an agent can act through several protocols; an organization can be represented by a changing population of software actors. The simple mapping from one account to one person becomes an increasingly poor description of the social world built on top of computation.

Action can travel farther from the body that initiated it. Institutions still have to decide when embodied human singularity carries political or economic weight.

For Humanode, proof of human sits at this boundary. As observable digital activity becomes increasingly detached from the number of people producing it, a narrow proof that a unique living person exists somewhere in the chain acquires new uses across governance, access and agent accountability.