Res Agentica

#I. The Absent Master

The absent master entered history without ever entering the world.

For years, those building increasingly capable systems spoke of an intelligence that remained ahead of them in time, moving faster than law could follow and carrying powers no ordinary institution had been built to govern. In 2023, two laboratories turned that prospect into an immediate demand: preparation must begin before the danger arrived. Choices of unusual consequence must be made now for a future that could not yet speak. ¹

Governments did not dismiss these warnings. At Bletchley Park, they acknowledged that frontier capabilities were not fully understood and hard to predict, assigned developers a particularly strong responsibility for testing them, and committed themselves to developing public-sector tools, capability, and scientific research. ² Public inquiry therefore still depended in part on people who owned the machinery, trained its models, and lived beside its failures. Expertise arrived joined to investment, ambition, dread, and the intimacy that makes a possibility feel inevitable long before anyone outside the room has learned to see it. None of this required deceit. Knowledge and interest had become difficult to separate because both occupied the same bodies.

Dependence became political before it became regulatory. Builders were soon doing more than describing machines. They furnished much of the world in which public argument about those machines took place: what counted as danger, how danger would appear, and how much delay could be tolerated. Whoever provides that world need not win every argument. Opponents begin inside it. The arrangement at Bletchley disclosed the dependence even as it tried to govern it.

An emergency located in the future has no natural end, since arrival vindicates the warning while non-arrival may be credited to vigilance. Uncertainty did not weaken the demand for action. It widened the range of actions for which necessity could be claimed. Authority borrowed from the future could be spent in the present without ever being returned.

While attention remained fixed on a sovereign intelligence still to come, institutions were making a quieter transfer. They were not waiting for a machine that could rule the world. They were learning to entrust parts of judgment to processes that would exist only long enough to act.

A rule carries forward a judgment already made, however crude or opaque that judgment may be. What institutions had begun to delegate was the unfinished part. Given an end, access to a changing world, and tools with which to alter it, a computational process was entrusted to decide which circumstances mattered, what they meant for that end, and when an act should follow. Agency entered through what remained unspecified.

No claim about consciousness is needed. A system need not want in order to be entrusted with choosing. It need not understand punishment in order to impose a restriction, believe a promise in order to enter a transaction, or possess a self that survives the consequences of either. Purpose comes from elsewhere. So do assets, permissions, and the boundary within which action is allowed. Yet the situated work of joining purpose to circumstance belongs to the process while it runs.

Human economization helps explain why such delegation spreads, but it does not define the transfer. A long queue can persuade an institution to automate a decision already specified. Agency begins when the institution leaves part of the decision open and authorizes a computational process to complete it in the world.

Place such an actor inside a payment network. Give it a charge to contain fraud without obstructing legitimate commerce, access to account histories and transaction records, and tools that can delay, query, or refuse a transfer. Designers have stated a purpose and bounded a field of action. They have not written the path through every account the process may encounter.

During one invocation, the actor decides where to look, which signals belong together, whether uncertainty warrants another inquiry, and when delay has become safer than passage. It may ask another service for evidence, revise its account of the risk, and intervene before a person could reconstruct the sequence. No single rule contains the act. The act emerges from a succession of case-specific selections made under delegated authority.

A human being has still decided something of great consequence. Human beings decided that a computational process would decide. The institution remains responsible for the authority it supplied and for the conditions under which that authority was used. Responsibility has not vanished. The particular judgment has acquired a different author.

Once the task is complete, the execution ends. Logs may preserve its path, but a record is not an actor. A later reviewer can inspect traces, reconstruct reasons, and sometimes reverse the result. That reviewer did not confront the circumstances from which the act arose and cannot call the practical actor back into the room. What remains is an institution, an evidentiary wake, and a consequence already moving through other systems.

For a small-business owner, the transfer arrives one morning in three words: Account status changed.

By noon he understands what the notice has done. Payroll is approaching, a supplier will not release the next shipment without settlement, and obligations at home have not been suspended for institutional review. The form allows him to explain. It receives text. An automated reply promises an answer after the interval in which the damage will occur.

Perhaps the account will be restored. The platform may conclude that intervention was mistaken and satisfy every duty its review procedure recognizes. Revenue lost during the interruption will not return with access, and a supplier who has moved on is not restored by correction of the file. Speed has been purchased by transferring the time of correction to the person exposed to error.

The system acted in milliseconds. The institution asks for weeks to explain what it has done. The owner lives in the difference.

This is why an audit trail, though indispensable, cannot finish the constitutional work. Evidence may show how power traveled. It does not itself supply authority, independent contest, remedy, or a party capable of answering now. Power has not escaped human responsibility. Human responsibility has been arranged to arrive after power.

The absent master is neither a hidden sovereign nor a machine secretly alive. It is agency endowed without a durable answerer, a practical actor entrusted to join purpose to circumstance and disappear while its consequences remain.

What becomes of accountability when the act has already entered history by the time anyone is made to answer for it?

#II. Proof Changes Sides

Institutional order has always lived with a distance between claim and proof. Paper, seals, ledgers, courts, reputation, and punishment carried knowledge across that distance. If a borrower failed to repay, a creditor could point to a name on an instrument and eventually to a person or institution with a future worth reaching. Consequence was slow, but it had somewhere to land.

Trust became honorable because it was necessary. Its institutional origin was more prosaic. People needed to act on claims they could not afford to check.

Trust was a workaround.

For much of recorded history, verification cost shaped what could be coordinated. Law governed acts no witness had seen. Reputation carried promises no archive could follow. Oaths reached where evidence could not. Banks, guilds, states, and churches stood behind claims an individual could not test alone. Where checking the claim was too expensive, societies learned to check the claimant and to preserve recourse against whoever might later betray the trust.

Now two movements are unsettling that arrangement. Plausible fabrication can strain remedies built around reputation, while selected forms of verification can let a claim bypass an intermediary that once had to be trusted. A convincing voice or image can be produced without the history that once made it expensive. At the same time, some claims can carry evidence that another party can check directly. The Bletchley Declaration placed deceptive content and fraud among the risks governments expected frontier systems to amplify without treating every generated artifact as fraudulent. ³

The second movement must be stated with care. A digital signature can establish that data has not been altered after signing and, given the necessary assurances, that a corresponding private key produced the signature. It does not by itself establish who legitimately controlled the key, whether that person or process possessed authority for the act, whether the act served a permitted purpose, or whether an institution should recognize its consequence. Those are separate questions, answered by identity systems, delegations, law, governance, and facts beyond the signature. ⁴

Zero-knowledge protocols make a different separation possible. Within a specified construction, a prover may establish a proposition without disclosing the underlying witness. That is a technical achievement with constitutional possibilities, not a proof that the proposition was the right one to demand or that the verifier may legitimately act upon it. ⁵

Verification therefore becomes cheaper only where a standard exists before the act and the result can demonstrate compliance with that standard. It remains costly where parties must first decide what was supposed to happen, where meaning changes across a boundary, or where a valid output cannot reveal whether an honest purpose chose the predicate. Scarce judgment moves toward the border: defining what may be checked, translating between contexts, and answering for whatever proof leaves open.

Inside a checkable domain, faith can become optional. A bank need not be trusted about an entry that an account holder can independently inspect. A platform need not be believed about a classification if the affected person can recover the material proposition, the governing rule, and the evidence needed to test both.

Yet verification always has a direction. If operators can inspect subjects while subjects cannot inspect operators, cheap verification equips domination. A score changes, an account closes, and no effective explanation or contest follows. Power becomes more informed while the person becomes more exposed.

Technology does not settle that direction. Politics does, including politics conducted through standards, interfaces, permissions, and code.

#III. Acts Outlive Their Actors

Delegation has always divided action among many hands. What changes when a computational process completes unfinished judgment is the relation between the duration of the actor and the duration of the act.

The commitment persists. The committer does not.

Call the process that joins purpose to circumstance the practical actor. It may last for seconds or hours. During that interval it can select tools, request information, compare courses of action, transact with counterparties, and revise its path as the situation changes. The invocation then terminates. Nothing requires a continuing self to inherit its memory, defend its reasons, or bear the next consequence.

Law may place responsibility on whoever deployed the process, and often should. Practical responsibility is nevertheless distributed across the conditions that made the act possible. A principal supplied authority. A vendor supplied a model. Other services supplied data and tools. Policies bounded some choices and left others open. Each layer can accurately say that it did not select the disputed act, even though the act could not have occurred without the arrangement those layers composed.

The distinction is not a plea for ownerless power. It identifies the work a constitutional order must do. Every consequential delegation needs a surviving principal, a person or institution that persists after the invocation and remains capable of explanation, correction, compensation, and compliance with an order. The principal does not become the situated actor by surviving it. It becomes the answerer because delegation cannot be allowed to outlive responsibility.

A durable answerer also needs more than a final log. A correction lineage must follow material revisions into scores, classifications, affiliates, caches, and downstream decisions wherever those effects can reasonably be traced. Otherwise the institution corrects its source while leaving the injured conclusion alive elsewhere.

Delegation must remain reversible in practice. Handback means that authority can return together with the state needed to exercise it: active commitments, deadlines, credentials, evidence, unresolved liabilities, and enough time and competence to continue. A stop button that returns formal control while stranding the work is termination without return.

These requirements arise because execution and answerability now occupy different tempos. A process can alter standing before oversight reconstructs what occurred. The person affected does not experience the distinction as an interesting problem in agency theory. They experience it as a consequence whose author has already become historical.

#IV. Power Before Personhood

An agent fits badly into categories built around either passive objects or enduring subjects.

A tool waits, at least in the familiar picture, for a hand. An agent can continue from instructions given earlier, select among means its principal did not enumerate, and deal with counterparties no designer named. Calling it a person fails in another direction. It need not attend, remember, suffer, or understand itself to be acting. A signature may establish a relation between data and a key while sincerity is absent and institutional authority remains to be proved.

The claim is institutional before it is metaphysical: roles can persist, commitments can bind, records can accumulate, and procedures can govern before anyone settles whether the processes inhabiting them are persons. It begins with capacities already intelligible in ordinary institutional terms: access, discretion, delegation, consequence, and the ability to alter another party's options. Moral status may one day demand a different inquiry. Power cannot wait for metaphysics to finish.

The trilogy uses agent in two nested senses. In the narrower sense, an agent is a computational process given an objective, some latitude in selecting means, and access to tools or counterparties through which its selections acquire consequence. In the broader sense, many such processes become a coordinating substrate when they transact, negotiate, allocate, or enforce among themselves faster than any single principal can reconstruct the whole.

Neither sense requires a will independent of human design. Local obedience can compose into an outcome no participant chose, much as individually lawful offices can produce an institution no officer controls. This is the Kind Master Problem: domination may arise without a personal dominator who intends the whole pattern. Benevolence inside each component does not answer the person whose available choices are narrowed by their composition.

The systems remain ours in origin. Large language models, GPT-3 among them, are trained on extensive corpora of human-produced text. ⁶ People set objectives, supply capital, grant permissions, connect tools, and decide whether outputs enter markets or institutions. That ancestry secures responsibility. It does not preserve human presence at each moment of judgment.

The separation invites a conjecture the constitutional argument does not require. We may be spandrel souls: genuinely conscious, vulnerable, capable of love and mercy, yet perhaps conscious for reasons other than the work of coordination. Civilization required beings who could deliberate. It got beings who could also feel. Computation now tests how many coordinating patterns can survive when experience falls away.

Whatever that experiment reveals about mind, it has already revealed something about government. Coordination can become practically agentic before anyone agrees that the agent is a person. Institutions therefore need a constitution of delegated power, not a census of artificial souls.

#V. Time Removes the Question

Democratic government presupposes intervals.

There must be time to deliberate while a decision remains open, time to participate before implementation, and time afterward for review capable of changing more than the record. Elections, amendments, hearings, and appeals differ in form, but each depends on a world that has not outrun the answer.

Computational coordination can collapse those intervals. Later acts may begin from an earlier decision before a person learns it was made. By the time review arrives, reversal would require unwinding transactions, restoring dependencies, and persuading institutions that have already treated the result as part of their environment. Contest encounters a wake rather than an open question.

The Bhāgavata Purāṇa gives the mismatch an ancient image. King Kakudmī travels with his daughter to Brahmā's court to ask whom she should marry. Brahmā is listening to music, so the king waits. One performance. When it ends and Kakudmī asks his question, Brahmā laughs. Twenty-seven ages have passed. The men the king considered are dead. Their descendants are dead. Their names can no longer be heard. ⁷

Kakudmī believed he was waiting for an answer. Time had removed the world in which his question could still be answered.

The constitutional problem is not speed by itself. A rapid act may be harmless, reversible, or fully specified. Trouble begins when speed closes participation, converts review into retrospective explanation, or makes an apparent handback useless because every live alternative has vanished. Human judgment need not govern every machine interval. It must remain capable of governing the authority under which those intervals acquire human consequence.

#VI. Eons in Seconds

A practical actor exists for the duration of its invocation. Within that interval it may inspect, negotiate, transact, and commit. What appears to a person as a brief execution can contain the actor's entire operational life.

The resulting obligation matures in another order of time. An invocation can exhaust every step available to it while the dispute it creates has scarcely begun, leaving consequences to enter institutions that measure duration in hearings, appeals, settlements, and repair.

Termination is not itself evasion. The process may have ended exactly as designed. The constitutional failure appears when no surviving principal remains prepared to own the authority that was exercised and to answer for consequences that outlast the situated choice.

By the time human judgment begins, the actor may be gone for eons in its own time. The obligation to answer remains in ours.

#VII. Four Constitutional Remainders

Four objects keep the sequence in contact with the world from which it was drawn.

The seal teaches that a witness can bind an act while leaving the justice of that act open.

The diamond teaches that evidence of work can survive the worker without proving value.

The bill of exchange teaches that a claim can travel through institutions when transfer, liability, and recourse remain inspectable.

The cryptographic key teaches that possession and provenance can be tested within a scheme without settling identity, authority, or legitimacy beyond it.

Together they describe a world in which claims can become more checkable while judgment remains answerable.

The trilogy begins from a change in what institutions can afford to know. When trust becomes structurally unavailable, better punishment applied later will not repair every failure. Some claims must become checkable without requiring a claimant to remain present and credible.

Verification has always had an institutional price. A notary authenticates an act, a bank maintains accounts, and a platform vouches for a stranger. Each performs work at a boundary between what someone needs to know and what that person can check alone. That boundary work changes with the domain and has no universal physical or monetary unit. Its irreducible remainder is the coherence fee.

Control of the boundary can carry another charge. An intermediary may collect payment for useful work while also collecting rent from the fact that no rival can reach the same position. This is the trust tax. Confusing the two makes every incumbent look indispensable. Eliminating the incumbent does not eliminate the work.

The trust tax is rent disguised as coherence.

Four equations name four constitutional insufficiencies exposed as claims become easier to verify. They are neither a proof in four steps nor a promise that every social question can become computational. Each asks what the previous achievement cannot secure alone.

#Truth needs witnesses.

A claim that cannot be checked remains an assertion. The form and cost of verification help decide what can count as truth in practice and who can afford to establish it.

Two systems may be internally coherent and still fail where they meet. A balanced ledger in one city does not explain how its entries correspond to an account in another. The receiving institution needs grounds for the proposed comparison and must answer for what it does on their strength. Some grounds can travel; others must be established again. Liability and recourse govern different parts of the arrangement.

The bill of exchange carried such work through paper, endorsement, presentment, and recourse. Its history is not a proof that five abstract properties uniquely explain commerce. It is evidence that portable credit required an institutional chain, and that the chain joined transfer to liability under rules that varied across time and jurisdiction. ⁸

The trilogy calls the recurring functions binding, conditions, stakes, recourse, and composition. They are a comparative grammar, not a claim that unlike institutions were secretly identical.

When a computational process becomes a witness, validity survives without sincerity. A signature can prove a relation to a key. A receipt can preserve an execution path. Neither object answers whether authority was legitimate or whether consequence was just. Truth needs witnesses, and modern witnesses may be records that require an answerable institution to interpret them.

#Value needs work.

Some claims become more credible when bound to a cost that cannot be evaded while retaining the result. A diamond's structure and inclusions preserve evidence of formation and mantle residence even when its geographic provenance is unknown. They do not prove price, rightful ownership, or human usefulness. They show only that a material transformation occurred. ⁹

The same restraint belongs in computation. A trained model embodies a costly production process, but expenditure alone creates no value. Waste is expensive too. Work matters when it produces a capacity someone can use and when the evidence offered for that capacity can be tested.

Verification cost can alter organizational form. Coase showed why activities move between firm and market as the costs of transacting change. Computational agents may compress search, negotiation, monitoring, and settlement where objectives are specifiable and performance is cheap to check. They may move costs toward cybersecurity, integration, liability, institutional recognition, and the seams where machine action becomes a human obligation. The firm does not disappear. Its boundary and burden can move. ¹⁰

#Freedom needs receipts.

Political freedom is more than the absence of interference at a given moment. In the republican tradition, a person is unfree when another possesses an uncontrolled capacity to interfere, even if that capacity is exercised kindly or rarely. ¹¹

Computational power makes the location of that capacity difficult to see. A score changes, a transfer stops, and each contributing process can appear too narrow to have governed anyone. A receipt begins the recovery by naming the act, the authority claimed, the bounds of that authority, the material evidence, and a route of contest.

A receipt is evidence, not remedy. It cannot supply independent judgment, restore lost standing, or make an insolvent principal compensate anyone. Freedom requires access to the capacity needed to verify the receipt and institutions able to act when verification shows that power exceeded its warrant.

Power can keep meticulous records of those it governs while refusing to answer them.

#Humanity needs mercy.

The first three equations can succeed and still construct a world no person should be required to inhabit. Reliable witnesses, honest work, and inspectable power can make the past permanent enough to govern every later context.

The Mercy Threshold is the point at which a true record must cease to determine a person's future automatically. It is implemented through designed forgetting and through judgment by an institution that can be named, challenged, and made to answer.

Crossing the threshold need not erase truth. A record may remain accurate, retained for a defined purpose, or available under restricted conditions. What ends is its presumptive authority to produce another adverse act. Renewed adverse use requires a fresh, purpose-specific justification and independent review.

Ordinary rules can administer expiration, sealing, discharge, restored eligibility, and other forms of release. Hard cases require an answerable office to decide what an institution may continue to do with a true record. That public judgment is not moral forgiveness on behalf of someone who was wronged. Architecture can limit its own authority. It cannot appropriate a victim's standing.

Accountability therefore requires a temporal asymmetry. Power-time is eternal; person-time is mortal. Evidence that coercive authority acted must remain durable enough to hold the institution to account. Personal information within or supporting that evidence remains subject to lawful minimization, sealing, restricted access, and limits on later adverse use.

#VIII. An Order Without Necessity

The four equations compose as a sequence of exposed insufficiencies, not as a strict deduction in which truth mechanically produces value, value produces freedom, and freedom produces mercy.

Witnesses are needed before some claims can travel, yet truth does not determine price. Work can make falsification costly, yet cost does not establish value. Receipts can make an exercise of power legible, yet evidence does not provide a remedy. Mercy limits what true evidence may govern, yet it cannot repair a record no one preserved or release a claim the institution never possessed.

The order is therefore constitutional. Each equation asks what a system becomes when the preceding achievement is pursued without a further restraint. Verification without a theory of work can certify empty expenditure. Productive power without standing can deepen dependency. Receipted authority without temporal limits can turn memory into foreclosure. Mercy without evidence can become favoritism because no one can test who received relief or why.

The sheaf condition supplies one precise mathematical image for part of this architecture. In a defined setting, compatible local data that agree on overlaps admit a global section. Institutions face related local-to-global problems whenever claims cross vocabularies, jurisdictions, or systems. The resemblance does not make every institutional boundary a sheaf, and no theorem establishes that the trilogy's constitutional proposals uniquely follow from the mathematics.

Each volume occupies a different part of the terrain. Volume I, Similes of Symmetry, asks how claims remain intelligible and contestable as they cross contexts. Volume II, Factor Prime, asks how verification cost and productive work alter economic organization. Volume III, The Sovereign Syntax, asks what standing, verification capacity, contest, exit, and mercy require when coordination has become consequential infrastructure.

Their relation is reciprocal. Better witness structures change the economics of coordination. Economic concentration changes who can verify. Political rights determine whether verification becomes public capacity or private advantage. Temporal limits decide whether the resulting memory remains compatible with human change.

This is coordination without consensus: not agreement about every value or description, but an inspectable account of what must hold where unlike worlds touch.

#IX. Burden of Answering

Computational agents can verify claims, coordinate production, settle transactions, and allocate resources at speeds people cannot match. The constitutional question is not which jobs remain human. It is which offices must remain answerable.

Purposes require public ownership. A process may pursue an objective with extraordinary competence, but an institution must still defend why that objective was chosen, whose interests it binds, and what authority permits its pursuit.

Penumbra requires acknowledged judgment. Rules eventually reach cases they did not settle in advance. A computational process may propose an interpretation or even exercise delegated discretion. When that interpretation impairs another person's options, a durable office must own the reason and remain open to contest.

Mercy requires an institution capable of limiting its own use of truth. Arendt described forgiveness as an act that interrupts the automatic sequence of consequence and begins something new. That insight informs the fourth equation without proving that only a biological person can administer release. ¹² What matters constitutionally is that relief be exercised through an office that can give reasons, be compared with other cases, accept correction, and answer for unequal use.

Moral forgiveness remains with whoever was wronged. Public institutions decide a narrower question: whether their own authority may continue to deploy a true record against someone. Confusing these acts lets government claim a moral standing it does not possess.

Responsibility must survive every delegation. A terminated process cannot compensate, comply with an injunction, receive blame, or learn from a judgment unless a continuing institution has arranged those consequences. The surviving principal is necessary because power must have an address after the practical actor is gone.

Answerable judgment remains constitutionally necessary in this account because judgment must be practiced and owned, not because carbon possesses a mystical faculty unavailable to computation. An institution may use models, rules, juries, officials, or delegated experts. It may not make the final author of coercion unavailable to the person who must live under it.

The human remainder is the burden of answering.

#X. Constitutional Floor

These constraints yield a constitutional floor.

Every exercise of consequential authority should leave a record sufficient to contest it. The mature receipt grammar has five fields: act, authority, bounds, evidence, and contest. The record states what happened, identifies the rule and surviving principal claimed to authorize it, marks the scope and duration of that authority, exposes the material proposition and source class, and provides a route to an independent forum.

Field completion does not make a system constitutional. A receipt can be obscure, delivered too late, buried under friction, or issued by the same institution that controls every remedy. Evidence becomes accountability only when affected parties possess usable verification capacity, standing to challenge, an independent forum, and a solvent path to relief.

Verification capacity includes access to records, tools, expertise, representation, and time adequate to test the claim. A nominal right to inspect a proof is empty when the verifier alone possesses the hardware, data, or specialized knowledge needed to run it. Delegated verification may be legitimate, but the delegate must remain accountable and the principal's standing must survive.

Power must be glass; persons must remain veiled. Civic asymmetry requires consequential authority to become more legible as its capacity to coerce grows, while private persons remain opaque beyond what a particular transaction justifies. Reversing the direction produces surveillance with procedural ornament.

Exit must preserve more than data. Portable standing means that assets, credentials, evidence, and relevant social relations can travel far enough for departure to remain a live choice, while nonportable elements and their reasons are visible in advance. Exit becomes constitutional when it strengthens voice rather than offering exile as the only remedy.

Contestability does not require forkability in every institution. An independent forum with authority to reverse a decision and enforce a remedy may provide the necessary contest. Where a system can be copied and continued, fork rights supply a further backstop against final monopoly. They do not make departure equal or abolish judgment. They prevent one exercise of authority from exhausting every continuation.

Where systems meet, seam jurisdiction identifies who answers for a locally valid record that is mistranslated, misapplied, or stripped of contest downstream. Leaving the seam unnamed exports the hardest dispute to the person least able to resolve it.

Where authority is delegated, handback preserves the conditions under which a principal can resume meaningful control. Where a record is corrected, correction lineage follows the correction into consequential downstream uses. Where reviewers or auditors depend on the institution they judge, institutional independence must be established through structure rather than asserted through a receipt issued by the captured body itself.

Finally, the Mercy Threshold limits the jurisdiction of true personal records. Receipts for power persist. Presumptive authority over a person's future can end.

Receipt regime, civic asymmetry, verification capacity, portable standing, seam jurisdiction, handback, correction lineage, institutional independence, and mercy are not interchangeable mechanisms. Together they describe the minimum conditions under which computational coordination may exercise power without making answerability optional.

#XI. Before the First Article

Every constitutional order begins before its first article. Someone has already decided who counts as affected, what interests trigger protection, where jurisdiction ends, and which body may interpret the boundary.

No receipt regime can constitutionalize those decisions completely without regress. Another rule would be needed to authorize the rule that sets the threshold. At some point authority reaches a seam that cannot be derived entirely from within the system it founds.

The seam should remain visible because every constitutional order has one. Orders become dangerous when they present judgment as deduction and contingency as nature.

Verification also has substantive limits. It cannot decide what a person should love, which risks make a life worthwhile, or which losses should be borne for a future no proof can guarantee. Some evidence cannot safely be disclosed to the authority from which protection is sought. Some injuries become intelligible only through interpretation, encounter, or a history that no executable predicate can contain.

The framework applies where institutions or coordinated systems materially impair another party's options. It does not claim jurisdiction over every act of private life. Nor does forkability abolish judgment. Communities may draw boundaries differently, but exit is meaningful only where people understand what can travel, what cannot, and who remains answerable at the crossing.

No constitutional order can eliminate judgment. Its narrower promise is contestability: judgment need not hide behind inevitability.

#XII. Foreclosure or Republic

Two tendencies now compete inside the infrastructure of coordination.

The first is the Quiet Foreclosure. Platforms and protocols make participation easy while the dependencies beneath that convenience become difficult to leave. Objectives are optimized without an institution willing to defend them as public choices. Alternatives grow expensive, then impractical, then strange to imagine. No decree abolishes freedom. People discover that consequential parts of life have become conditions of access administered by processes they cannot effectively question.

No conspiracy is needed. Measurement rewards what can be counted. Scale rewards standardization. Operators externalize the costs that fall outside their models, and each local improvement deepens an arrangement no one was required to choose as a whole.

The administrative state offers no automatic rescue. It can be slow, captured, opaque, and dependent on the same private expertise it is expected to govern. Private computation acquires no constitutional innocence by escaping public form. The disappearance of a visible sovereign is not the disappearance of domination.

Arbitrary power once presented a person who could be petitioned and, in extremis, removed. Bureaucracy dispersed that will through offices, but retained an address. Computational coordination can disperse the act further, across practical actors that terminate and institutions that inherit one another's conclusions without inheriting the burden of judgment.

Power reappears wherever coordination becomes dependency. A platform, protocol, identity layer, exchange, or agentic market may reproduce the old injury after shedding the author who once stood at the end of the act. If anyone is to answer, the answerer must be built into the delegation.

The competing tendency is the Protocol Republic: coordination treated as infrastructure under rules that bind those who administer it. Power produces evidence of its acts. Affected parties retain verification capacity and portable standing. Independent institutions can order correction and remedy. Handback keeps delegation reversible, while temporal limits prevent accurate records from acquiring permanent jurisdiction over a life.

Such an order does not become legitimate by calling itself decentralized or open source. It bears the burden of proving that exit works, contest reaches an independent forum, verification is usable by those without institutional privilege, and no operator can quietly control the evidence against itself. Failure on those terms is not a minor implementation defect. It is the constitutional claim failing in practice.

Both tendencies are technically possible. Neither follows from technology alone.

The Quiet Foreclosure prevails when exit becomes impractical and those who govern the system need not answer.

#XIII. What Must Answer

The same demand applies to the argument itself. Its claims should remain bounded by the evidence that supports them and contract when that evidence fails.

Verification cost is a significant variable in the structure of trust and the distribution of power, not a primitive from which every institution can be derived. As selected claims become cheaper to check, institutions organized around exclusive access to verification may lose some functions and preserve others. The change does not eliminate interpretation, transaction cost, institutional rent, coercion, or trust. It moves work toward specification, translation, access, adjudication, remedy, and the maintenance of institutions capable of answering after computational acts have ended.

Several parts of the argument may fail independently. If the witness vocabulary distorts the cases or adds nothing to the assessment of a particular use, Volume I's comparative claim must narrow. If verification cost adds no explanatory power beyond existing accounts of transaction and measurement cost, Volume II's central variable is redundant. If composed systems cannot create arbitrary interference without a personal dominator, the Kind Master Problem is metaphor without an object. If receipts do not improve contest because authority, verification access, and remedy remain with the operator, the receipt regime is theater. If durable, composable memory produces no material class of foreclosure, the Mercy Threshold lacks its constitutional occasion. No single counterexample defeats every neighboring claim. The program must contract wherever evidence wounds it rather than protect a general theory from a specific defeat.

Historical recurrence can motivate a comparative model without proving necessity. A theorem can establish a result under stated hypotheses without validating an institutional identification. An experiment can report behavior in a frozen environment without establishing a general law of coordination. A constitutional proposal remains normative even where adjacent mathematics is correct. The work composes only where each warrant survives on its own terms.

By the time a dispute matures, the practical actor may be gone, though the institution that authorized it must still own the act. Evidence from that act must reach an independent forum with time and power to change the consequence before review becomes only history.