The Constitutional Governance of Formalisation got-v1-3c-7 | Canonical paper | Issue 1 https://goto-go.com/machine/documents/got-v1-3c-7/ GOT-V1-3c-7 The Constitutional Governance of Formalisation Observational Revision, Competitive Forks, and Stewardship Without Ownership Abstract Formalisation begins as an answer to ungrounded authority. It becomes a governance problem when its own operational Soft Horizons enter institutional life as objects of authority. A named method can be cited, adopted, taught, implemented, certified, automated, litigated, marketed, resisted, simplified, or invoked as proof of legitimacy. A discipline designed to prevent unsupported claim-strength can then acquire unsupported strength of its own. The word method may become a badge. A completed form may become a defence. A release may become doctrine. A certificate may become closure. A stewarded version may become a trust signal. A machine-readable implementation may become the practical standard even where it cannot carry the residual the method requires. These forms are not errors by default. They are necessary operational closures and governance trust carriers. They reduce repeated audit by letting users, institutions, courts, regulators, vendors, and machine systems rely on prior formalisation-work rather than reconstructing the method locally each time. They become capture when reliance on the carrier is treated as warrant, when the carrier is mistaken for the relation-preserving process, or when the trust earned by one governance form is spent in another without carrying its residual. The danger is not only rejection. Rejection leaves the discipline outside the institution. Adoption can be more dangerous because adoption brings the discipline into the very systems it was designed to audit. A regulator may prefer administrability. A court may prefer procedural fit. A government may prefer delivery. A corporation may prefer certification. A platform may prefer scale. A machine vendor may prefer encodability. A profession may prefer protected discretion. A public campaign may prefer morally desired outcomes. Each pressure may be reasonable from inside its own domain. None is sufficient to establish epistemic authority. The governance problem is therefore not how to make formalisation popular, institutionally convenient, or easy to implement. The problem is how formalisation remains corrigible without becoming capturable. It must remain open to correction, but not open to alteration by preference. It must remain usable, but not soften its discriminator burdens for adoption. It must be publicly inspectable, but not publicly amendable in the epistemic sense. It must support institutional and machine-mediated use where appropriate, but not let compliance, certification, schema fit, or implementation become relation-preservation. The governing rule is therefore stricter than self-consistency: The method may change only through demonstrated relation-preservation. Its present principles govern procedure and technical formalisation, but those proposed principles remain answerable where demonstrated relational failure reaches them. A proposed revision must show that the current formulation fails to preserve relation in a defined class of cases, and that the proposed revision preserves relation more strongly under finite conditions. It must identify the affected relation, the locus of failure, the availability condition under which the failure appears, the closure demand the current formulation cannot satisfy, the comparator cases, the residual reduced, the residual introduced, and the conditions under which the revision itself would fail. It must also state whether the change concerns adaptive application, the current technical formalisation, or a proposed constitutional element, because the burden of correction rises with the level being altered. This paper distinguishes proposed constitutional elements, the current technical formalisation through which those elements are made auditable, and the adaptive applications through which domains use them. It separately distinguishes epistemic authority, custodial authority, and operational authority. Epistemic authority belongs only to demonstrated relation-preservation under finite-observer conditions. Custodial authority belongs to stewardship of releases, archives, dissent records, amendment histories, residual registers, and fork comparisons. Operational authority belongs to institutions applying the method in particular domains. Custodial and operational authority are necessary governance Soft Horizons and legitimate trust carriers. Capture begins when either pretends to be epistemic authority. Formalisation must therefore be governed without ownership. No originating author, founding text, foundation, institution, profession, court, regulator, public body, vendor, machine system, or political majority owns the conditions by which claims become evaluable. They may propose, preserve provenance, inspect, challenge, apply, contest, evidence, fork, and correct. They may not remove relation-preserving burdens by authorship, status, vote, convenience, or adoption pressure. Nor does stable inclusion in the current technical formalisation make a technical structure constitutionally permanent. Governance succeeds only where formalisation remains strong enough to resist manipulation and corrigible enough to learn. 1. Introduction: when formalisation becomes authority Formalisation is introduced to solve one problem and immediately creates another. The first problem is ungrounded authority. Language becomes load-bearing. Terms drift across contexts. Categories acquire procedural force. Records become warrant by habit. Institutions convert under-specified words into decisions. Machines scale inherited abstractions through data, architecture, metrics, classifications, and outputs. Formalisation answers this by requiring consequential claims to carry their grounding conditions. The present technical formalisation makes relation, locus, availability condition, closure demand, abstraction, evidence path, record path, carrier path, comparator, residual, threshold, safeguard, and update recoverable. These structures are the current means by which the burden is made auditable; their presence prevents claim-strength and reliance-strength from migrating into the easier carriers of authority: a term, form, release, certificate, stewarded version, schema, institutional process, or machine-readable implementation. The second problem appears once formalisation itself becomes usable. A formal discipline does not remain a private idea. If it has value, it enters institutional life because it reduces the work of repeated audit. It gives courts, regulators, boards, professions, vendors, public bodies, platforms, and machine systems a way to rely on prior closure-work. It becomes a named method, a release, a template, a certificate, a stewarded version, a schema, or a compliance process. It can be cited in reports, invoked in litigation, adopted in policy, translated into training, embedded in review processes, requested in procurement, referenced by courts, marketed by vendors, and partially encoded into machine systems. At that point, formalisation becomes load-bearing. The danger is structural. A method designed to prevent words from carrying authority beyond their grounding can itself become a word carrying authority beyond its grounding. An institution may say that a decision was method-compliant while the affected relation remains hidden. A vendor may say that its tool implements the method while reducing residual to a field. A regulator may say that an organisation has satisfied the method while treating completion as adequacy. A court may treat method-use as evidence of procedural reasonableness. A public body may invoke the method to reassure constituents. A foundation may release versions whose status becomes more important than the argument by which they were justified. The same authority failure returns at a higher level. The method becomes a label. The label becomes a defence. The defence becomes an institutional habit. The institutional habit becomes a trust signal. The trust signal becomes a source of legitimacy. Once legitimacy attaches, actors have incentives to preserve the symbol while weakening the burden. They will not always do this cynically. Often they will do it because the full burden is difficult. They may want administrability, judicial clarity, regulatory adoption, procurement compatibility, public confidence, machine-readability, professional fit, political acceptability, commercial viability, or urgent moral action. These pressures matter for action. They do not determine epistemic strength. Governance is needed because adoption does not protect a method from capture. Adoption is one of the main ways capture occurs. A discipline can be rejected, ignored, opposed, or misunderstood. These failures are visible. The method remains outside the system, and the conflict is explicit. Adoption is more ambiguous. The institution appears to accept the discipline. It adopts its vocabulary. It creates a procedure. It asks for forms. It allocates duties and answerability. It records compliance. It produces training. It announces alignment. Yet the adopted form may no longer impose the burden that gave the discipline force. Adoption is neither validation nor failure. It is exposure: the moment an operational closure begins to travel inside institutional life. This is why governance cannot be reduced to implementation. Implementation asks how a discipline enters practice by creating usable operational Soft Horizons. Governance asks how those horizons remain answerable once they are in use. A method can be implemented badly because a user lacks competence. It can be implemented narrowly because a domain has special constraints. It can be implemented incompletely because resources are limited. Those are ordinary practical problems. Capture is deeper. Capture occurs when practical pressures redefine the burdens of the method while preserving its name, status, or legitimacy function. The governance of formalisation must therefore obey the same discipline that formalisation imposes on claims. It must not claim authority by office, prestige, public consensus, institutional adoption, implementation scale, commercial uptake, legal mention, certification, machine-readability, or moral urgency. It must expose the relation its own governance claims to preserve. It must state what counts as failure, what counts as improvement, what residual remains, what comparator cases test the revision, what affected loci reveal, and what would force future correction. If a governance process cannot meet that burden, it becomes another source of ungrounded authority. The guiding question is not who owns the method. No one owns it in the epistemic sense. A body may steward releases. An institution may apply it. A court may refer to it. A regulator may require a form of it. A community may contest it. A vendor may encode aspects of it. A researcher may fork it. But none of these positions makes a formulation stronger by status. Strength must be earned by relation-preservation under finite conditions. The proper question is therefore: under what conditions may formalisation be revised, extended, simplified, implemented, or rejected without becoming a captured instrument of authority? The answer begins with demonstrated relation-preservation: The method must not change by preference, but neither may its present standard protect itself merely because it is inherited. The governing burden is not the method’s name, steward, release status, adoption history, institutional authority, popularity, clarity, implementation success, or originating authorship. It is the recoverable preservation of relation under stated locus, availability condition, closure demand, comparator pressure, residual visibility, threshold, safeguard, and update. Section 4 distinguishes the proposed constitutional elements that give this burden continuity from the current technical structures and adaptive applications through which it is expressed. This rule is not a refusal of change. A method that cannot change cannot learn. It will fail wherever application reveals hidden residual, missing locus, weak comparator, poor abstraction, unworkable threshold, absent safeguard, or inadequate update discipline. It will preserve its form while losing relation. That is rigidity, not reliability. But a method that changes by preference cannot be trusted. If every institution, public majority, profession, regulator, vendor, foundation, implementation community, or machine system may revise the burdens according to convenience, adoption, political pressure, moral urgency, or desired outcome, then the method no longer disciplines authority. It becomes another language by which authority negotiates its own comfort. Governance must preserve the tension between stability and corrigibility. Stability is necessary because a method must be reliable enough to resist manipulation. Its proposed constitutional burdens cannot shift merely because an institution finds them costly or inconvenient. The current technical formalisation expresses those burdens through relation, locus, availability condition, closure demand, boundary, interface, scale, horizon, abstraction, evidence path, comparator, residual, threshold, safeguard, and update. These structures remain stable requirements of the present formalisation; they may not be discarded by preference, but they may be altered, combined, or replaced where a proposed revision demonstrates at least equal or stronger preservation of the constitutional work they perform. Stability therefore protects burden without turning current technical form into permanent constitution. Corrigibility is necessary because no finite formulation can anticipate every domain, failure mode, institutional pressure, machine-mediated distortion, affected-locus challenge, or future comparator case. A formulation may need revision because application reveals that it fails to preserve relation in a defined class of cases. Corrigibility prevents the operational Soft Horizon from becoming doctrine. The method must be able to learn from its own use. The governance task is to allow correction without allowing capture. That task is constitutional in the strict sense used here. It concerns the conditions under which a formal discipline may alter itself without losing the burden that makes it what it is. It is not a general theory of political constitution. It is not a replacement for democratic lawmaking, judicial authority, professional judgement, institutional mandate, ethical deliberation, public value, or practical administration. Those domains retain their own functions. The concern here is narrower: how changes to a discipline of relation-preserving audit are justified. The answer cannot be ownership. Ownership would reproduce the authority problem. If the originating author or founding text owns the method, priority may become epistemic authority. If a foundation owns it, the foundation’s status may become the source of validity. If a profession owns it, professional convenience may become epistemic criterion. If a regulator owns it, administrability may displace relation-preservation. If a vendor owns it, marketability may shape the burden. If public vote owns it, preference may become warrant. If machine implementation owns it, encodability may become truth. The answer is stewardship without ownership. Stewardship maintains conditions for correction. It preserves records. It publishes revisions. It exposes dissent. It compares forks. It records failure cases. It distinguishes accepted changes from rejected changes. It protects terms from drift. It keeps amendments recoverable. Stewardship is a custodial Soft Horizon: it stabilises the correction path without owning truth. It does not make the method true. It does not make claims strong. It does not confer epistemic authority by institutional position. Its legitimacy lies in keeping the correction path recoverable. The aim is not to protect a text from change. The aim is to protect correction from capture. 2. The governance problem: formalisation becomes load-bearing Formalisation begins as an answer to ungrounded language. It becomes a governance problem when the formalisation itself enters the world as language and becomes a Soft Horizon of its own. A named method can be cited. A standard can be adopted. A template can be completed. A release can be referenced. A steward can be invoked. A schema can be implemented. A certificate can be issued. A training programme can be sold. A decision can be described as method-compliant. A court can mention the method as evidence of due process. A regulator can require use of the method. A corporation can advertise alignment. A platform can encode audit fields. A government can claim that a policy decision has passed epistemic review. Each form closes the wider discipline into something usable. At that point, the method has become a load-bearing term and a trust carrier. This is not accidental misuse. It is the predictable consequence of usefulness. A discipline that helps reconstruct claim paths, expose residual, separate record from warrant, distinguish threshold from closure, and preserve update conditions will be attractive to institutions because it saves work. Institutions need defensible procedures. They need reviewable records. They need language for acting under uncertainty. They need methods that can survive litigation, regulation, public scrutiny, internal accountability, and machine mediation. A formal discipline appears to offer those things by allowing reliance on prior formalisation-work rather than full local reconstruction in every case. That attraction creates danger because institutional usefulness and epistemic strength are not the same relation. An institution may value the method because it makes decisions defensible. That may be operationally useful; it does not mean the decision preserves the affected relation. A vendor may value the method because it creates a market for compliance tools. That may support uptake; it does not mean the tool preserves residual. A regulator may value the method because it offers a uniform standard. That may aid oversight; it does not mean the standard carries the right abstraction. A court may value the method because it makes reasoning visible. That may improve review; it does not mean the reasoning is adequate. A government may value the method because it gives public assurance. That may matter politically; it does not mean the assurance is warranted. The governance problem begins where these values attach themselves to the method’s name. The phrase “the method was followed” can become a substitute for showing that relation was preserved. The phrase “method-compliant” can become a shield against affected-locus challenge. The phrase “audited” can become a way to close scrutiny. The phrase “aligned with the method” can become marketing. The phrase “released by the steward” can become appeal to authority. The phrase “machine-readable implementation” can become practical closure because what the schema can encode becomes what the institution can see. In each case, formalisation suffers the same failures it diagnoses. It can become compressed language. The term method carries implied authority greater than the visible conditions of its use. It can become semantically drifting language. It may mean one thing in conceptual statement, another in legal deployment, another in institutional compliance, another in software implementation, another in public rhetoric, and another in commercial certification. It can become circular ungrounded dependency. A decision is trusted because it followed the method. The method is trusted because a steward released it. The steward is trusted because institutions adopt it. Institutions adopt it because it gives decisions legitimacy. Legitimacy then circulates without returning to the relation under review. This is the governance analogue of ungrounded load-bearing language. The method’s vocabulary may remain present while its burden disappears. Relation may become a field. Residual may become a checkbox. Comparator discipline may become a list of alternatives no one expects to challenge the verdict. Update may become a formal clause without an operational trigger. Affected-locus visibility may become consultation theatre. Threshold may become a label for a decision already taken. Safeguard may become a policy phrase. Audit may become documentation. The form survives. The discipline weakens. This danger is intensified by the normal pressures of institutional life. Institutions prefer repeatability. Courts prefer administrable tests. Regulators prefer consistent criteria. Vendors prefer certifiable products. Machine systems prefer field structures. Governments prefer public assurance. Professionals prefer discretion protected by familiar categories. Public campaigns prefer methods that support urgent moral conclusions. Each of these pressures may be understandable. None is the same as relation-preservation. A governance structure that treats these pressures as merely external will fail. They are not external. They enter the method through adoption. They shape how the method is interpreted, simplified, taught, implemented, funded, cited, and defended. They determine which parts become visible in practice and which parts are treated as too difficult. They create incentives to preserve the language of discipline while reducing the cost of discipline. The method must therefore be governed against symbolic success. Symbolic success occurs when the method becomes valuable as a sign of legitimacy. It may be cited because citation reassures. It may be completed because completion demonstrates diligence. It may be adopted because adoption signals modern governance. It may be encoded because encoding signals scalability. It may be certified because certification signals diligence. These signals may have practical value, but they become dangerous if they substitute for the relation-preserving work the method requires.1 A method can fail by being ignored. It can also fail by becoming too useful as a symbol. The second failure is subtler. Ignored methods lack influence. Symbolically successful methods have influence but lose discipline. They become part of the environment of authority. Actors learn to speak them. Consultants learn to package them. Institutions learn to pass them. Software systems learn to approximate them. Courts learn to cite them. Public bodies learn to invoke them. Over time, the question shifts from whether relation was preserved to whether the recognised form was used. The governance problem is to prevent that shift. Formalisation must remain answerable to the relation it was introduced to preserve. A named method does not create epistemic strength. A release does not create epistemic strength. A certificate does not create epistemic strength. A completed template does not create epistemic strength. A schema does not create epistemic strength. Institutional adoption does not create epistemic strength. Public approval does not create epistemic strength. Machine execution does not create epistemic strength. Only demonstrated relation-preservation under finite conditions creates epistemic strength. This does not mean the surrounding forms are useless. They may be necessary. Releases help stabilise reference. Templates can reduce omission. Certificates can attest that a process was performed. Schemas can make some records interoperable. Institutions can assign audit duties. Courts can require reasons. Regulators can enforce procedural visibility. Machine systems can assist monitoring and comparison. These forms are operational Soft Horizons of the method and practical reliance carriers for users who cannot reopen the full discipline at every point of use. They may support the method, but they do not replace it. Every supporting form has the same danger: it may become more practically authoritative than the discipline it was meant to support. A template can become the method because it is what users touch. A certification can become the method because it is what institutions need. A schema can become the method because it is what machines can execute. A release can become the method because it is what stewards publish. A court reference can become the method because it is what lawyers cite. A vendor tool can become the method because it is what procurement buys. The governance task is not to abolish these forms. It is to keep them in their proper authority layer. A release stabilises reference. It does not guarantee truth. A template supports use. It does not guarantee relation-preservation. A certificate attests process. It does not convert completion into warrant. A schema enables implementation. It does not define adequacy. A steward maintains correction paths. It does not own epistemic authority. An institution applies the method. It does not validate the method by applying it. A machine system may implement parts of the discipline. It does not become the discipline because it can execute fields. This distinction is the first condition of constitutional governance. The method must enter the world through operational forms, but it must not be captured by them. A finite system needs Soft Horizons in order to act. Governance does not demand that formalisation remain pure by refusing usability. It demands that the usable closures remain visibly provisional, indexed, and corrigible. Formalisation becomes load-bearing when others begin to rely on its name, release, forms, credentials, tools, or compliance signals. Governance begins when that reliance must be audited. 3. Adoption as capture risk Adoption is exposure, not validation. This distinction should govern every institutional use of formalisation. When a regulator, court, public body, corporation, platform, vendor, profession, research community, or political movement adopts a method, the method gains reach. It gains operational life. It becomes easier to cite, teach, fund, procure, encode, audit, litigate, and require. But none of that proves that the adopted form preserves the relation the method exists to preserve. Adoption is the point at which the method begins to answer not only to its own conceptual burden, but to the incentives of the adopting system. This is why adoption is the primary capture risk. Rejection is easier to diagnose. A system refuses the method, misunderstands it, or ignores it. The distance between method and institution remains visible. Adoption is more dangerous because the institution speaks the method’s language while potentially weakening the conditions that make the method useful. The adopting institution may keep the vocabulary and lose the burden. It may keep relation as a field but not require relation-preservation. It may keep residual as a checkbox but not allow residual to block stronger claims. It may keep comparator as a heading but not require meaningful alternatives. It may keep update as a clause but not create an operational trigger. It may keep affected-locus language but not give affected loci a challenge path. It may keep safeguard language but not constrain power. It may keep certification while narrowing inspection to form. It may keep method-compliance while treating compliance as warrant. The method appears to have entered the institution. In fact, the institution may have absorbed only its visible shell. Regulatory adoption gives one form of capture risk. A regulator may prefer versions of formalisation that are administrable across many organisations. That preference is understandable. Regulators need repeatability, enforceable criteria, comparability, and manageable review. But administrability is not relation-preservation. If the regulator simplifies the method so that audits become inspectable only as completed forms, certification marks, or uniform fields, it may make oversight easier while weakening the discipline. The captured method then supports regulatory visibility without preserving the affected relation.2 Judicial adoption gives another form. A court may prefer formulations that fit existing legal categories, burdens, standards, and procedural structures. That too is understandable. Courts cannot operate without administrable concepts. But procedural fit is not epistemic adequacy. If a court treats method-use as evidence of reasonableness without asking whether the method preserved relation in the case, then the method becomes procedural comfort. It closes challenge by saying the recognised discipline was invoked. Governmental adoption creates delivery pressure. A public body may want formalisation because it helps justify decisions under uncertainty. It may want a method that can be trained, rolled out, monitored, audited, and reported. It may also want reassurance. Under pressure to deliver, it may simplify residual, narrow affected-locus challenge, weaken comparator burden, or convert threshold action into policy closure. Delivery may be necessary. It is not the same as relation-preservation. Corporate adoption creates certification pressure. A corporation may prefer a version that can be packaged, certified, sold, or passed through procurement. It may want visible compliance, brand value, audit-ready documentation, and reduced liability. These may produce real improvements in record-keeping and review. But they may also shift attention from relation-preservation to defensibility. The organisation asks: can we show that we followed the recognised method? The stronger question is: did the claim path preserve the relation under the closure demand being made? Platform adoption creates scale pressure. A platform may need methods that work across millions of cases, automated workflows, changing contexts, multilingual environments, and rapid decisions. It may therefore prefer fields that can be standardised and thresholds that can be automated. But scale is not adequacy. A method that cannot carry residual, affected-locus challenge, context, and update through scale may create an appearance of disciplined governance while expanding a reified Soft Horizon. Machine-vendor adoption creates encodability pressure. A vendor may prefer a version of formalisation that can be represented in software before its conceptual burdens are settled. Relation becomes a field. Residual becomes a status. Comparator becomes a menu. Safeguard becomes a flag. Update becomes a scheduled event. Such encoding may be useful where the domain has been sufficiently formalised. It becomes capture where machine-readability is treated as proof that the relation-preserving burden has been met. Professional adoption creates discretion pressure. A profession may welcome formalisation where it protects professional judgement from crude external metrics. It may also resist parts of the method that expose discretion, comparator weakness, inherited categories, or record-path failure. The adopted method may then preserve professional status while softening the audit burden. The profession says it has a disciplined method. The method no longer disciplines the profession’s authority. Public campaign adoption creates moral urgency pressure. A movement may adopt formalisation because it supports a desired outcome, exposes institutional failure, or strengthens demands for accountability. That may be legitimate and important. But moral urgency does not own epistemic burden. A desired outcome may justify action threshold; it does not remove the need for residual, comparator, safeguard, and update. If the method is revised to guarantee preferred outcomes, it ceases to discipline authority and becomes another instrument of it. These capture routes differ, but their structure is the same. An adopting system preserves what helps it act and weakens what makes it answerable. This is why adoption cannot be treated as validation. Wide adoption may show that a method is useful, legible, teachable, or institutionally attractive. It may show that the method meets a practical need. It may reveal that the method has reached a viable operational Soft Horizon. It may even generate valuable evidence about how the method performs in domains. But adoption itself is not epistemic authority. It must be audited like any other trust carrier. The question is not whether the method has been adopted. The question is what happened to the method’s burdens during adoption. Did relation remain visible, or did it become a field? Did locus remain specific, or did institutional vantage replace affected-locus exposure? Did availability condition remain distinct from warrant, or did record presence become proof? Did closure demand remain explicit, or did action become certainty? Did abstraction remain exposed, or did categories become natural? Did evidence path remain recoverable, or did procedural status replace evidence? Did comparator discipline survive, or did alternatives become theatre? Did residual retain blocking force, or did it become caveat language? Did threshold remain threshold, or become proof? Did safeguard constrain, or merely reassure? Did update become triggerable, or remain aspirational? Adoption should therefore produce an adoption audit. An adopting system should state what version of the method it is using, what domain translation it has made, and at what level the change occurs. It should distinguish adaptive application from alteration of the current technical formalisation and from any claimed challenge to a proposed constitutional element. It should state what burden it has preserved, what burden it has modified, what residual the adoption introduces, what failure cases it anticipates, what affected loci can challenge, what comparator cases will test the implementation, what update condition applies, and what cannot be inferred merely from the fact of adoption. This does not make adoption impossible. It makes adoption honest. A method cannot remain effective if it refuses all institutional translation. Domains differ. Law, medicine, education, platform governance, safety engineering, public administration, science, procurement, certification, and machine deployment each require different operational forms. Adaptive application is necessary. But adaptive application must not become burden reduction by another name. The rule is strict: Adoption may translate the method into domain form. It may not remove the conditions that make the method auditable. Ordinary domain translation therefore occurs at the level of adaptive application. Where a domain cannot preserve a present technical burden without altering the technical formalisation, the change enters the amendment process. Where the difficulty reaches one of the proposed constitutional elements, it enters constitutional challenge and carries the higher burden developed below. Translation must not become an unmarked route between these levels. Where translation is needed, the translation must preserve the burden. A legal version may express comparator through precedent, burden of proof, and adversarial challenge. A medical version may express comparator through differential diagnosis, evidence grading, and follow-up. A machine-governance version may express comparator through benchmarks, deployment testing, error analysis, and affected-locus review. A public-administration version may express comparator through policy alternatives, evidence streams, stakeholder challenge, and retrospective audit. The form may differ. The burden remains. Adoption becomes capture when the adopting system treats local convenience as epistemic improvement. A regulator’s need for inspection becomes the method. A court’s need for procedural fit becomes the method. A vendor’s need for certification becomes the method. A machine system’s need for schema becomes the method. A public campaign’s need for outcome becomes the method. The method then changes not by relation-preservation, but by institutional preference. That is why adoption must remain under governance. The method may enter institutions. It may be used, taught, translated, implemented, cited, and contested. But every adoption is itself a claim: that this operational form preserves the method’s relation-preserving burden for this domain, under these constraints, with this residual, and under this update condition. That claim must be auditable. Adoption is not the end of governance. It is where governance begins. 4. Public inspectability without public epistemic ownership Formalisation must be publicly inspectable. A method designed to discipline authority cannot be hidden behind authority. If its terms, burdens, amendment history, residual register, release structure, rejected revisions, dissent records, fork comparisons, and domain translations are inaccessible, then the method becomes the very thing it was built to prevent: a closed carrier of unexamined claim-strength. Public inspectability is therefore not optional. It is one of the conditions by which formalisation remains answerable. Affected loci, institutions, experts, courts, regulators, researchers, implementers, and the wider public must be able to see how the method claims strength, what burdens it imposes, what changes have been made, what residual remains, what dissent exists, what forks have been proposed, and what failures have forced revision. A closed method cannot reliably correct itself because it cannot be adequately challenged.3 But public inspectability is not public epistemic ownership. This distinction is decisive. The public may inspect the method. It may challenge it. It may evidence failure. It may expose affected-locus distortion. It may identify domain cases the current formulation cannot carry. It may support forks. It may test implementations. It may reject institutional misuse. It may show where a release, certificate, schema, steward, or compliance process has become a trust carrier without preserving the underlying burden. It may force the method to answer. The public may not make the method true by preferring it. It may not make a revision stronger by voting for it, and it may not remove a proposed constitutional burden or a current technical burden merely because that burden is difficult, unpopular, inconvenient, politically costly, commercially unattractive, or emotionally unsatisfying. But public, affected-locus, scientific, technical, or implementation evidence may expose a failure that enters the amendment path. Preference does not revise the method; demonstrated relation-preserving challenge can. The public can force the method to answer. The public cannot force the method to flatter. This distinction protects against two opposite failures. The first failure is technocratic closure. A closed expert, custodial, professional, institutional, or technical body controls the method, excludes challenge, treats dissent as misunderstanding, and uses specialised authority to prevent affected-locus correction. The method becomes insulated. It may remain internally coherent while failing the relations it claims to preserve. Its vocabulary becomes stable, but its correction path narrows. Its releases become orderly, but its residual disappears. Its stewards become trusted, but their trust is no longer answerable enough. Technocratic closure is especially dangerous because it can look disciplined. It uses correct terms. It maintains records. It cites expertise. It distinguishes versions. It may even preserve a formal amendment process. But if affected-locus evidence cannot reopen the formulation, if external comparator cases cannot challenge its abstraction, if domain failures are treated as implementation errors by default, or if custodial authority decides what counts as valid criticism by status alone, then the method has become authority-protective rather than relation-preserving. The second failure is populist alteration. A public, stakeholder group, institution, profession, political movement, campaign, or implementation community may use participation to alter the method toward desired outcomes, easier adoption, moral satisfaction, administrative convenience, institutional fit, commercial viability, or public reassurance. The method becomes responsive in the wrong way. It changes because people want it to change, not because the proposed change preserves relation more strongly. Populist alteration is also dangerous because it can look accountable. It may speak in the language of inclusion, participation, democratic legitimacy, public value, moral urgency, lived experience, stakeholder voice, or practical need. These may all matter. Affected-locus evidence is often indispensable. Public challenge can expose failures that experts, institutions, or stewards miss. But participation is not warrant by itself. Voice can expose relation; it cannot abolish the burden of preserving relation. The governance discipline rejects both failures. It requires exposure without surrendering epistemic burden. It requires challenge without preference capture. It requires participation without treating participation as proof. It requires expertise without technocratic ownership. It requires public accountability without public possession of epistemic strength. The governing sentence is: The public may inspect, challenge, evidence, fork, and test formalisation; it may not vote away the burdens that make claims evaluable. Nor are those burdens protected merely because the originating formulation named them. Preference cannot revise them, but demonstrated relational failure can. That sentence does not make the method anti-democratic. It prevents democracy, expertise, institutional mandate, stewardship, public pressure, and machine implementation from being misdescribed as epistemic closure. Political authority may decide action. Legal authority may decide consequence. Public authority may define institutional mandate. Affected-locus evidence may expose failure. Expert judgement may improve comparison. Implementation experience may reveal limits. None of these may remove the conditions by which claims become auditable. A public body may decide that action is required under urgency. That is a political or administrative judgement. It does not remove the need to distinguish threshold from closure. A court may decide legal consequence. That does not remove the need to distinguish procedure from relation-preservation. A regulator may define a compliance form. That does not remove the need to distinguish form from adequacy. A community may expose harm. That does not remove the need to specify the relation, locus, evidence path, residual, and update condition. A machine system may make implementation possible. That does not remove the need to show that the schema carries what the method requires. Public inspectability is the condition of correction. Public epistemic sovereignty over the burdens of audit would be capture by preference. Formalisation must therefore be public enough to be answerable and stable enough not to become preference-driven. This is the constitutional balance. The method cannot be locked away by stewards. It also cannot be rewritten by demand. It must remain exposed to the world, but changed only by relation-preservation. 5. Stable burdens, proposed constitutional elements, and adaptive application A formal discipline requires stable burdens and adaptive application. The burdens must be stable because they are the conditions under which the discipline continues to constrain claim-strength rather than merely accompany it. If a burden can be removed whenever a domain finds it difficult, expensive, inconvenient, politically awkward, technically resistant, or hard to encode, then formalisation has not adapted. It has weakened. A claim that does not preserve relation is not made stronger by administrative need. A category that hides abstraction is not made safer by adoption. A record that loses provenance is not made warrant by institutional repetition. A decision that hides residual is not made closed by urgency. A machine output that cannot preserve update is not made adequate by scale. But stability creates its own danger if the current formalisation is mistaken for the permanent form of the method. No finite formulation can know in advance that its own technical architecture is the best possible expression of the burden it is trying to preserve. A method that treats every term, field, discriminator, path, comparator, or procedural structure in its first mature release as constitutionally fixed may remain internally consistent while becoming progressively less able to learn. It can preserve the form through which a principle was first expressed after that form has ceased to preserve the relation as well as an available alternative. The governance problem therefore requires a distinction that the language of stable burdens alone does not make sufficiently clear. There is a difference between the principles that appear necessary if the discipline is to remain recognisably directed toward the same methodological purpose, the technical structures through which the present formalisation expresses those principles, and the operational forms through which particular domains make those structures usable. These levels overlap, but they are not identical. At the deepest level are the elements that the present work proposes as constitutional to the method. Constitutional is used here in the same restricted sense already established by this paper. It does not mean constitutional by legal enactment, founding authority, authorship, institutional declaration, or public vote. It means the principles that presently appear necessary if future revision is to remain continuous with the discipline of relation-preserving audit rather than merely inherit its name. The next level is technical formalisation. TEM presently gives those principles a particular architecture. It names relations, loci, availability conditions, closure demands, horizons, abstractions, evidence paths, record paths, carrier paths, comparators, residuals, reliance limits, thresholds, safeguards, updates, and other structures because those distinctions make the burden auditable. These are not casual vocabulary choices. They perform work. They prevent known forms of collapse. They make claims reconstructable where less disciplined language allows claim-strength to migrate into familiar nouns, trusted records, completed procedures, or institutional status. Some relations under audit impose additional recovery burdens without becoming universal elements of the formal packet. Responsibility is one such relation. Where a load-bearing claim concerns responsibility, accountability, liability, duty, authorship, decision ownership, or causal attribution, formalisation must recover enough of the claim path to distinguish Responsibility Provenance from Responsibility Warrant. Responsibility Provenance concerns the recoverable consequential branch relations attributable to loci through distinction-conditioned organisation. Such relations may include realised transformation, responsibility-bearing non-application, retained or maintained conditions, and material conditioning of another locus’s Counterfactual Inclusion Field. Responsibility Warrant concerns the social, moral, legal, organisational, procedural, or institutional allocation of answerability, duty, burden, liability, supervision, stewardship, or decision ownership. The distinction is indexed to the relation being audited. Where application is material to the attribution, recovery must be sufficient to identify the relevant Counterfactual Exclusion. Where non-application, maintenance, retention, or cross-locus field-conditioning is material, recovery must instead be sufficient to identify the comparator-stable difference made to the relevant Counterfactual Inclusion Field. Responsibility Provenance is therefore not a universal formalisation field, and the particular technical form through which responsibility is made recoverable does not become constitutional merely by inclusion in the present formalisation. The constitutional burden is narrower: role, office, authority, procedure, or allocated answerability must not carry causal-provenance strength beyond the recoverable relation. Where that substitution occurs, governance has allowed an operational or custodial allocation to acquire an epistemic and attributional force it has not earned. The third level is adaptive application. A court, laboratory, hospital, regulator, public body, safety process, machine-mediated system, or research community cannot be required to reproduce one identical documentary form whenever it carries the same epistemic burden. Domains have different evidence relations, different affected loci, different closure demands, different temporal pressures, different failure modes, and different consequences of error. Operational form must therefore vary.4 The mistake is to allow movement between these levels to occur without being marked. If every current technical structure is treated as constitutional, the discipline becomes technically sacrosanct while continuing to call itself corrigible. If every constitutional burden is treated as merely one current implementation choice, adoption can remove the difficult parts while preserving the method’s name. If every domain is required to implement the technical architecture in one standard form, the method becomes bureaucratic rather than relational. The first constitutional task is therefore to ask what the present formalisation is actually trying to preserve. This paper can propose an answer. It cannot settle that answer by declaring it. The originating author does not acquire epistemic authority over the constitutional identity of the method merely by having formulated it first. Priority permits the first proposal to be made and its reasoning to be exposed. It does not grant permanent ownership of the distinction between principle and implementation. If constitutional governance is coherent, the proposed constitutional elements must themselves remain answerable to the same discipline of relation-preservation, comparison, residual, and correction that they impose upon later formulations. What follows is therefore not presented as a finished constitution of TEM. It is a proposal concerning the elements that appear, from inside the present formalisation, to carry its constitutional burden. The proposal begins with the relation between claim-strength and what can actually be recovered. A claim must not carry more strength than the recoverable relation warrants. This is not merely one field within the present method. It is the governing reason the fields exist. Precision does not create claim-strength. Formal language does not create it. Publication does not create it. Institutional status does not create it. Certification does not create it. Repetition does not create it. Machine execution does not create it. These things may contribute real evidential, comparative, reconstructive, or reliability relations. Where they do, their contribution should count. But none may carry epistemic weight simply because it has become a trusted carrier. Claim-strength remains attached to the relation the claim purports to preserve. This immediately requires finite access to remain relationally indexed. A finite locus never encounters a target without conditions of encounter. Access occurs somewhere, for something, through a boundary or interface, at some operative scale, under some abstraction, over some horizon, with some availability condition, and under some closure demand. The present method makes those conditions explicit because unindexed access allows a local achievement to travel as if it were unconditioned possession. The constitutional burden is not that every future formalisation must preserve the exact present syntax of the Indexed relation. A better formalism may combine fields, separate them differently, or expose dependencies that the present architecture handles poorly. What cannot simply disappear is the requirement that finite access remain relationally located. A claim cannot silently detach itself from the conditions through which the relation became available and retain the same strength. This is why closure demand also matters at the level of principle. There is no single state called closure that serves every epistemic task. Recognition is not explanation. Classification is not causal demonstration. Prediction is not proof. Proof is not necessarily sufficient reliance for high-consequence action. Legal settlement is not scientific settlement. Administrative completion is not epistemic closure. An institution may need to act before a scientific question is settled. A scientific question may be strongly settled for one purpose while remaining inadequate for another. Unless the demanded closure remains visible, evidence earned for one question can migrate into another without further work. A successful prediction can become causal warrant. A recognised pattern can become a classification. A classification can become an identity. A procedure can become adequacy. Action can become retrospective proof. The terminology through which closure demands are expressed may change. The requirement that the demand itself remain explicit cannot disappear without weakening the discipline. The same constitutional separation applies to abstraction. Finite access is necessarily selective. Observation does not reproduce a target whole. It stabilises some relation through a cut, scale, registration condition, model, measurement, category, or record. Formalisation then introduces further abstractions because the relation must travel through language, records, institutions, procedures, and machines if it is to become usable beyond the immediate locus of observation. These abstractions are real. A record is a real organised state. A mathematical model is a real informational structure. A category instantiated in an institution may alter real trajectories. A machine representation may produce material consequence. A legal classification may reorganise the possibilities available to a person. The fact that an abstraction is an abstraction does not make it fictitious. But the reality of the abstraction does not make it identical with the ontological target it compresses. The constitutional burden is therefore not to eliminate abstraction. That would be incoherent for a finite observer-system. It is to preserve the distinction between the relation represented and the representation through which that relation becomes available, retained, carried, compared, or acted upon. A measurement must not become the phenomenon merely because it is precise. A record must not become the event merely because it persists. A category must not become the person merely because the institution requires a category. A model must not become the target merely because it predicts successfully. A completed method-form must not become relation-preservation merely because the form was correctly completed. This distinction becomes especially important where claims depend upon informational discrimination. TEM does not begin from the assumption that every physical difference is already an operative distinction for every locus. Nor does it treat every lawful state change, every externally describable variation, or every coincident downstream outcome as evidence that information has performed the explanatory work attributed to it. Across the wider architecture, physically supported difference becomes consequential only through relations by which distinctions become available, discriminable, registered, retained, applied, propagated, or otherwise capable of altering downstream possibility. The present technical formalisation contains particular ways of disciplining that transition. They matter because without a discriminator, informational language can expand until every lawful change can be redescribed as processing and every successful outcome can be narrated retrospectively as the result of an operative distinction. The constitutional burden is more general than any one existing discriminator. Where a consequential claim depends upon informational processing, control, observation, classification, inference, signalling, memory, or applied distinction, the method must preserve some adequate means of distinguishing the operative informational relation from background state change, external interpretation, or mere outcome coincidence. The present technical machinery is therefore not sacrosanct merely because it is technically developed. A future discriminator that preserves the relation more strongly should be capable of replacing it. What cannot simply be removed is the discriminator burden itself. The same reasoning governs informational and evidential pathways. A relation may remain real even where the path by which it reached a present observer has become difficult or impossible to reconstruct. Reality does not owe the observer perfect provenance. Physical processes may erase prior states. Information may pass through repeated coarse-graining. High-noise, low-signal transformations may preserve enough relation to affect an outcome while destroying much of the path by which that influence could later be reconstructed. Aggregation may preserve statistical structure while losing individual history. Translation may preserve operational use while losing original detail. Institutional systems may compress complex histories into categories. Machine systems may transform information through many stages whose collective path becomes opaque. Records may be incomplete. Carriers may disappear. Earlier loci may no longer be accessible. A discipline of recoverability cannot answer this by pretending that unrecoverable relation did not occur. Failure to reconstruct a pathway is not positive exclusion of the pathway. What changes is the strength of the claim available to the present locus. Where the path can be recovered strongly enough for the closure demand, the relation may support correspondingly strong claim-strength. Where recovery is partial, the unrecovered structure must remain residual and the claim must narrow. Where the path required by the requested audit has degraded beyond meaningful reconstruction, non-auditability may be the proper output. The principle is therefore not universal traceability. It is proportional recoverability. The stronger the claim being made about a consequential relation, the greater the burden to recover the informational, evidential, record, or carrier path on which that strength depends. Where recovery fails, the relation is not erased. The observer’s entitlement to describe it strongly is reduced. This is one reason comparator discipline remains central. A formulation cannot gain strong support merely by surviving observations that all relevant alternatives would also survive. Agreement may reflect genuine confirmation, but only where the agreement discriminates. Repetition may strengthen a formulation, but repeated outputs sharing the same hidden dependency cannot automatically be counted as independent confirmation. Internal coherence may matter, but coherence inside an abstraction does not establish adequacy outside the abstraction that defines it. Strong claims therefore require exposure to forms of pressure capable of changing them. The present method expresses this through comparators, alternative formulations, affected-locus challenge, counterfactual cases, interventions, prior failure cases, forks, excluded branch-space, and other structures of disciplined contrast. A scientific comparator will not always resemble a legal comparator. An experimental ablation will not always resemble an institutional challenge path. The form may vary considerably. What must not disappear is the possibility of discrimination. A comparator that cannot alter the claim is theatre. An alternative that is listed but excluded from having evidential force is theatre. A challenge route that records dissent but cannot reopen the relation is theatre. Strong claim-strength requires the possibility of losing strength. This leads directly to residual. Residual is not the embarrassing remainder left after the important positive result has been stated. It is part of the epistemic structure of the result itself. It preserves what the current formulation has not settled, what that unresolved relation prevents the formulation from claiming, and what future observation, access, comparison, failure, or reconstruction could force update. A strong formulation may carry very little residual relevant to a particular action or closure demand. Nothing in this discipline requires practical paralysis merely because ultimate closure remains unavailable. But the formulation does not become stronger by deleting the remainder. A method that records only what survived while allowing failed comparisons, inaccessible pathways, unresolved countercases, affected-locus discrepancies, model mismatch, scale limits, excluded alternatives, and update conditions to disappear becomes less corrigible as it accumulates authority. Residual is therefore more than a reporting convention. It is one of the means by which a finite method remains visibly finite. The same discipline must extend beyond claim-strength into reliance. Finite observer-systems cannot reconstruct every relation every time they act. They must rely on records, experts, institutions, measurements, procedures, releases, models, and machines. Trust carriers are necessary precisely because repeated full audit is impossible. Governance cannot remove this requirement without making the method unusable. But reliance-strength and claim-strength are not the same relation. A highly supported claim may still warrant cautious reliance where the consequence of error is extreme or irreversible. A weaker claim may justify temporary action where delay itself creates greater exposure. A provisional classification may support investigation without supporting final attribution. An emergency threshold may justify intervention while leaving the underlying explanatory relation open. The current formalisation addresses these conditions through reliance paths, reliance limits, thresholds, safeguards, monitoring, stop conditions, and update. Future technical work may improve that structure. The constitutional burden is that action cannot be permitted to rewrite the epistemic status of the relation merely because action became necessary. Reliance must remain bounded by what the underlying claim has earned, by the consequences attached to reliance, and by the conditions under which that reliance remains acceptable. These principles also determine the relationship between method and procedure. Formalisation necessarily produces procedure. Once the burdens of a method become stable enough to travel, they become templates, audit sequences, schemas, review practices, institutional duties, release structures, software fields, and compliance processes. This is useful. Without such operational Soft Horizons, every use would require reconstruction from first principles. But procedure is downstream of the reason the procedure exists. A completed process cannot validate itself. The ordinary direction of the discipline is therefore from principle to technical formalisation and from technical formalisation to application. The principle identifies the relation-preserving burden. The technical architecture makes that burden auditable. The operational procedure makes that architecture usable under domain constraints. The result does not normally reverse that sequence merely because the result is inconvenient, surprising, popular, profitable, urgent, or institutionally preferred. If an outcome conflicts with a principle, the first question is not how to change the principle so that the result can stand. The first question is whether the claim, evidence, abstraction, comparator, implementation, threshold, or procedure failed to preserve the relation. But principle cannot become a final refuge from observation. A pattern of result-level failure may expose a poor implementation. If correction at the operational level does not resolve it, the failure may expose weakness in the technical formalisation. If serious comparison repeatedly shows that the technical architecture itself generates avoidable residual, suppresses relevant distinction, blocks recovery, or distorts the relation it is intended to preserve, then the challenge may reach the proposed constitutional principle beneath it. Principles govern procedure. Procedure does not validate principle. Relation remains capable of correcting both. This is the point at which constitutional governance must refuse self-immunisation. The proposed constitutional elements are not protected from revision simply because they have been called constitutional. That would make the word constitution a new authority carrier. A principle that systematically prevents relation-preservation in a defined class of cases must be open to challenge. A proposed replacement must carry a high burden because changing principle has wider consequences than changing implementation, but the correction path cannot be constitutionally closed. This is also why corrigibility must include the possibility that a better relation-preserving formulation displaces an inherited one. Priority is not epistemic authority. Authorship is not epistemic authority. Release status is not epistemic authority. Institutional adoption is not epistemic authority. A familiar technical vocabulary is not epistemic authority. A long history of successful use may itself become evidence relevant to reliability, but history does not remove the possibility of a later formulation preserving the relevant relation more strongly. The present architecture is entitled to resist weak challenge. Corrigibility does not mean instability. A proposed change that merely simplifies implementation, improves marketability, fits an institution more comfortably, increases machine-readability, changes terminology, satisfies political preference, or produces a desired result has not demonstrated epistemic improvement. But where another formulation exposes a failure the current architecture hides, carries a relation the current structure cannot recover, reduces important residual without creating a larger unresolved failure, survives relevant comparator pressure, and preserves the constitutional burden more strongly, governance must permit correction. That correction may replace an operational form. It may replace a technical mechanism. In sufficiently serious cases, it may force reconsideration of a proposed constitutional element itself. This is not a weakness of the constitutional proposal. It is one of the proposed elements. No component of the method becomes sacrosanct outside consistent application of its governing principles and continued accounting for residual. The constitutional proposal identifies what the current formalisation believes it must preserve. It does not grant those beliefs immunity from the world they are intended to describe. This distinction allows the existing stable burdens to be located more precisely. The present technical formalisation requires relation, locus, availability condition, closure demand, boundary, interface, scale, horizon, abstraction, evidence path, record path, carrier path, comparator, residual, reliance limit, threshold, safeguard, and update because these structures currently make its constitutional burden recoverable. They remain stable burdens of the present method. Their reclassification as technical does not make them optional. Relation prevents assertion from becoming self-grounding. Locus prevents the status of one observer, institution, model, user, or affected subject from being silently transferred to another. Availability condition prevents existence, describability, access, registration, and closure from collapsing into one state. Closure demand prevents evidence earned for one kind of settlement from being spent as another. Boundary and interface keep finite access mediated rather than imagined as possession. Scale and horizon prevent a locally stable relation from travelling outside the range in which it was established. Abstraction exposes the cut through which a target became usable. Evidence path keeps evidence attached to the relation that makes it evidence. Record path prevents retention from becoming warrant merely by persistence. Carrier path keeps inherited reliance attached to the route through which prior closure-work travelled. Comparator creates discriminatory pressure. Residual prevents overclosure. Reliance limit constrains what a trust carrier permits. Threshold allows action without turning action into certainty. Safeguard limits exposure under residual. Update keeps the closure revisable. These burdens have reasons. A future formalisation may discover that some can be combined, that others require further subdivision, that a present term poorly captures the relation it was designed to preserve, or that an entirely different architecture performs the same constitutional function more strongly. That possibility is not permission to delete the burden by relabelling it. A proposed technical replacement must show what relation the current structure preserves, where that structure fails, what the replacement improves, what residual is reduced, what residual is introduced, and under what conditions the replacement would itself require correction. The technical architecture is corrigible. Corrigibility is not burden-free substitution. Adaptive application begins one level further downstream. No finite technical formalisation can anticipate every domain, case, institution, technology, legal setting, scientific context, medical urgency, governance pressure, or machine-mediated distortion. Even where the constitutional and technical burdens remain stable, the operational form through which they are carried must respond to the relation being audited. Comparator discipline in a scientific model will not look identical to comparator discipline in a legal decision. A clinical safeguard will not always resemble a safeguard in long-term public administration. An update condition in formal mathematics will not necessarily resemble an update condition in medicine, safety engineering, platform governance, or machine deployment. A carrier path through scientific citation differs from one carried through legal precedent, regulatory guidance, institutional memory, or software infrastructure. The form must therefore remain adaptive because the relation varies. In law, the relevant burdens may be expressed through issue, evidence, burden, standard, reasons, precedent, adversarial challenge, remedy, and appeal. In medicine, they may travel through symptom, diagnosis, differential, test, intervention, consent, follow-up, monitoring, and safety review. In science, they may be expressed through hypothesis, instrument, measurement, model, comparator, residual, replication, and revision. In public administration, they may travel through category, eligibility, policy purpose, record, evidence, discretion, threshold, challenge, appeal, and review. In machine-mediated systems, they may be expressed through data provenance, label definition, objective, model, evaluation, threshold, deployment conditions, monitoring, affected-locus challenge, human review, and update. These are translations, not templates. A legal judgment need not look like a machine model card. A clinical record need not look like a scientific paper. A laboratory methods section need not look like a regulator’s certificate. A platform decision process need not look like a constitutional court. Domain forms may differ because the relevant loci, closure demands, evidence paths, scales, consequences, safeguards, and update conditions differ. The form differs. The burden remains. Adaptive application becomes illegitimate only where translation removes the relation-preserving work it claims to carry. A legal version that makes affected-locus evidence procedurally incapable of reopening the claim may preserve legal form while weakening relational access. A machine-readable implementation that removes residual because residual is difficult to encode may preserve schema compatibility while losing closure discipline. A public-administration process that records comparator alternatives but does not permit them to alter the decision has preserved a heading rather than discriminatory pressure. A certification regime that treats completion as adequacy has converted operational status into epistemic authority. A scientific implementation that preserves only the selected formulation while allowing material countercases and unresolved failure to disappear has increased presentational clarity at the cost of corrigibility. A teaching version that preserves the vocabulary while removing the reasons the burdens exist may preserve recognition while losing the method. The rule remains: Translate the burden; do not dilute it. The distinction developed here makes that rule more precise. At the proposed constitutional level, the discipline must preserve the principles by which finite claims remain answerable to relation: claim-strength cannot exceed recoverability; finite access must remain relationally indexed; closure demand must remain explicit; abstraction must not collapse into target; consequential informational distinction must be distinguished from background change where the claim depends upon it; pathway recovery must be proportionate to claim-strength; strong claims must remain exposed to relevant discriminatory pressure; residual must remain represented; reliance must remain bounded; procedure must remain subordinate to principle; and principle itself must remain answerable to demonstrated relation. At the technical level, the present architecture remains stable because it is the current formal means through which those principles become auditable. It may not be discarded by convenience, preference, institutional pressure, or novelty. It may be changed where another formulation demonstrates stronger preservation of the constitutional burden. At the operational level, application remains adaptive because domains must translate technical burden into forms appropriate to their own relations, affected loci, consequences, and closure demands. Adaptation is legitimate while the burden survives the translation. This three-level distinction is intended to preserve the tension that constitutional governance requires. Stability without corrigibility becomes doctrine. Corrigibility without stable burden becomes drift. Technical detail without constitutional purpose becomes ritual. Constitutional principle without technical exposure becomes aspiration. Adaptive application without either becomes branding. The method must remain resistant enough that its burdens cannot be negotiated away by convenience, and open enough that neither its founding text nor its present technical architecture becomes the authority it was designed to discipline. The proposed constitutional elements do not settle that problem. They make the problem inspectable. That is the most this paper can legitimately claim for them. 6. Epistemic, custodial, and operational authority Governance requires authority distinctions. Without them, every useful carrier tries to become the ground. A release begins to look like truth. A steward begins to look like owner. A certificate begins to look like warrant. A template begins to look like adequacy. An institutional process begins to look like relation-preservation. A machine implementation begins to look like the method. A public vote begins to look like epistemic settlement. The paper therefore separates epistemic authority, custodial authority, and operational authority. This section applies the account of Epistemic Authority introduced in On the Origin of Trust. Custodial and operational authority are applications of Structural Authority within this governance account. Epistemic authority belongs only to demonstrated relation-preservation under finite-observer conditions. A formulation is stronger where it preserves relation more strongly for a stated locus, under a stated availability condition, under a stated closure demand, through exposed abstraction, evidence path, comparator pressure, residual visibility, threshold, safeguard, and update. Nothing else substitutes for this. Not release status. Not adoption. Not certification. Not legal reference. Not institutional approval. Not expert prestige. Not public demand. Not machine execution. Epistemic authority is not ownership. It does not attach to a person, office, foundation, profession, regulator, court, public body, vendor, or machine system by virtue of identity or position. Those actors may contribute evidence, challenge, application, correction, stewardship, or implementation. They may expose failures the current formulation cannot carry. But the strength of a formulation comes from the relation-preserving work shown, not from the actor’s position. Originating authorship, a founding text, and canonical release status do not create a fourth authority class. The originating author contributes the first formulation, its reasoning, and its provenance. Once exposed to scrutiny, the constitutional standing of that proposal and the technical standing of its mechanisms depend on the same demonstrated relation-preservation required of later amendments. Custodial authority belongs to stewardship. It concerns releases, archives, amendment histories, residual registers, dissent records, fork comparisons, provenance, terminology control, support-document discipline, and public recoverability. It also preserves the distinction between proposed constitutional elements, current technical formalisation, and operational or domain implementations, so that stable custody does not silently promote a technical component into constitutional status through age, repetition, or release practice. Custodial authority is necessary because a method without stable custody cannot be used with a recoverable correction path. Users need to know which release is current, what changed, why it changed, what was rejected, what residual remains, what forks exist, and where the correction record is held. Custodial authority is therefore a legitimate governance Soft Horizon. It stabilises memory. It reduces repeated audit. It lets users rely on a release without reconstructing the entire amendment history each time. It lets institutions cite a version, compare implementation, review changes, and track dissent. But custodial authority does not make a formulation true. A steward may maintain a release; it cannot make the release epistemically stronger by maintaining it. A steward may reject a proposed amendment; the rejection is legitimate only if the reasons preserve relation better than the proposed change. A steward may publish a version; publication does not become proof. Operational authority belongs to institutions applying the method in particular domains. A court may decide how a method-informed claim enters legal reasoning. A regulator may require organisations to expose claim paths. A hospital may adapt the method to clinical record and intervention. A platform may adapt it to moderation, automation, appeal, and monitoring. A corporation may use it in risk governance. A public body may use it in administrative decision-making. A machine system may implement fields, prompts, logs, thresholds, and monitoring structures. Operational authority is also necessary. A method cannot make all domain decisions from the centre. It is not a substitute for law, medicine, science, public administration, engineering, ethics, or machine-governance practice. Domain institutions must apply the method under their own mandates, standards, evidence practices, legal constraints, and operational conditions. But operational authority does not validate the method by using it. Nor does it validate a decision merely by saying the method was used. Capture begins when custodial or operational authority pretends to be epistemic authority. Custodial capture occurs when the stewarded release becomes doctrine. The archive becomes canon in the wrong sense. The release number becomes the argument. The steward’s acceptance becomes truth. Dissent is treated as disorder. Forks are treated as disloyalty. Amendment is treated as threat. The custodial body no longer preserves correction; it owns the appearance of correction. Operational capture occurs when institutional use becomes warrant. A regulator says the method was followed and treats that as adequacy. A court treats method-use as reasonableness without relation audit. A corporation treats certification as evidence that its system is safe. A platform treats field completion as affected-locus accountability. A vendor treats implementation as correctness. A public body treats method-alignment as legitimacy. In each case, operational authority has turned use into ground. The separation of authority layers prevents these migrations. Epistemic authority answers the question: does the formulation preserve relation under the conditions claimed? Custodial authority answers the question: has the formulation, revision, dissent, residual, release, and fork record been preserved and made recoverable? Operational authority answers the question: how is the method applied in this domain, by this institution, under this mandate, with these consequences? Each authority is legitimate in its own place. Each becomes dangerous when it claims another’s function. Authority and responsibility must therefore remain distinct indexed relations. Custodial and operational authority may carry Responsibility Warrants: stewardship, supervision, answerability, duty, liability, or decision ownership may be allocated because a governance system requires a locus able to act. None of those allocations establishes Responsibility Provenance for the consequential relation under review. Conversely, Responsibility Provenance may attach to a realised transformation, a responsibility-bearing non-application, a retained or maintained condition, or material field-conditioning attributable to distinction-conditioned organisation even where no office, mandate, or institutional warrant names it. Epistemic authority concerns the strength of a formulation; it does not assign causal provenance. Governance must therefore neither infer provenance from authority nor treat an allocated warrant as having displaced provenance elsewhere. A steward should not decide domain consequence by custody alone. A court should not decide method strength by procedural fit alone. A regulator should not decide epistemic adequacy by compliance alone. A vendor should not decide relation-preservation by tool implementation alone. A public majority should not decide audit burden by preference alone. A machine system should not decide adequacy by executable schema alone. This authority separation also protects institutions from impossible burdens. A court need not become the owner of the method in order to use it. A regulator need not decide final epistemic doctrine in order to require recoverable claim paths. A steward need not apply the method in every domain in order to maintain releases. A vendor need not settle the method’s conceptual burden in order to implement support tools, provided the tool does not claim to replace that burden. The distinction lets each actor do necessary work without overclaiming. A healthy governance structure therefore keeps authority layered and cross-auditable. Epistemic claims remain answerable to relation. Custodial decisions remain answerable to amendment record, dissent, residual, and public inspection. Operational applications remain answerable to domain evidence, affected-locus challenge, comparator pressure, safeguard, and update. No layer becomes self-validating.5 The method must be strong enough to resist capture by any one layer. It must also be open enough for each layer to expose failures the others cannot see. 7. Amendment by demonstrated relation-preservation Amendment is governed by demonstrated relation-preservation. This is the constitutional centre of the paper. Procedure remains subordinate to principle, while the proposed principles themselves remain answerable to demonstrated relation. Amendment is therefore neither self-validation nor alteration by preference. It forbids change justified only by convenience, status, popularity, adoption pressure, certification need, schema fit, moral desire, professional comfort, institutional defensibility, or originating authority. A method that cannot change becomes rigid. A method that changes for the wrong reasons becomes capturable. The first task is to locate the level at which the failure actually occurs. Some failures are operational: the domain application is poor while the current technical formalisation and proposed constitutional burden remain intact. Some are technical: a present mechanism, distinction, path, comparator, or representational form fails while the constitutional burden it serves can be preserved more strongly by another formalisation. The deepest class is constitutional challenge: the claim that a proposed constitutional element itself systematically blocks, distorts, or loses relation. These possibilities must not be collapsed, because correction at one level does not automatically justify revision at the next. A proposed amendment must begin by identifying the failure class. It is not enough to say that the current formulation is difficult, unpopular, expensive, slow, hard to teach, hard to certify, hard to encode, hard to apply, or inconvenient to a domain. These may be implementation pressures. They may reveal real problems. But by themselves they do not show that the method fails. The amendment must state what kind of case the current formulation fails to preserve. Is there an affected-locus relation it cannot carry? A domain case it misclassifies? A residual it hides? A comparator it cannot survive? A machine-mediated distortion it cannot expose? A legal translation that breaks a burden? A release structure that prevents correction? The amendment must then specify the relation at stake. What relation is the current formulation supposed to preserve, and how does it fail? The failure must be located in relation, not merely in user preference. A revision is not stronger because it reads better. It is stronger if the current formulation loses the relation and the proposed formulation preserves it more effectively. The amendment must specify the locus of failure. For whom, or for what system, does the failure appear? Does it affect the claimant, the affected locus, the institution, the steward, the machine system, the court, the regulator, the public, the model user, or the domain practitioner? A revision that improves institutional usability while worsening affected-locus exposure is not a straightforward improvement. A revision that improves machine-readability while weakening residual carriage is not stronger merely because it is easier to execute. The amendment must specify the availability condition under which the failure appears. Did the relation enter as complaint, record, observation, measurement, model output, legal claim, certification process, implementation record, affected-locus challenge, fork, or retrospective audit? The path matters because failures can occur at access, abstraction, record, comparator, threshold, safeguard, update, or carrier level. A revision that mislocates the failure may solve the wrong problem. The amendment must specify the closure demand the current formulation cannot satisfy. Is the problem recognition, classification, explanation, legal judgement, institutional action, machine deployment, certification, retrospective audit, release governance, or update? A formulation may be sufficient for one closure demand and insufficient for another. If the amendment fails to state the demand, it may expand or weaken the method without justification. The amendment must identify comparator cases. A revision must show why the proposed formulation handles a class of cases better than the current formulation. This requires comparison. Which cases does the current version mishandle? Which alternative formulations were considered? What fork, dissent, edge case, affected-locus challenge, or implementation record tests the revision? What would count against it? Without comparator cases, amendment becomes assertion. The amendment must state residual reduced and residual introduced. No revision eliminates residual altogether. A stronger revision may reduce one uncertainty while introducing another. It may make affected-locus challenge clearer but complicate implementation. It may improve machine deployment but require stronger residual flags. It may improve legal translation but narrow generality. It may improve domain specificity but increase fragmentation. The amendment must not hide the new residual it creates. The amendment must state its own failure condition. A revision is not disciplined unless it says where it would fail. What evidence would weaken it? What domain would test it? What affected-locus challenge would reopen it? What implementation result would show it is too narrow, too broad, too rigid, too permissive, too difficult to update, or too easy to certify without relation-preservation? An amendment with no failure condition is a bid for doctrine. A constitutional challenge carries a higher burden. It must show that the failure cannot reasonably be localised to adaptive application or to a replaceable technical mechanism. It must identify a defined class of cases in which the proposed principle itself prevents, distorts, or obscures relation-preservation, and it must show why the replacement preserves the intended relation more strongly while carrying its own comparator field, residual, and failure condition. Calling a disagreement constitutional does not make it so. Amendment burden is therefore scaled rather than uniform. It rises with the level being altered, the dependency reach of the proposed revision, and its downstream consequences. This proportionality is a governance requirement; it does not prescribe a universal numerical threshold. Any operational or mathematical representation of that burden remains a corrigible technical formalisation. No numerical score, confidence level, statistical threshold, formal function, or procedure is sufficient for amendment merely by satisfying its own form. The representation must itself demonstrate that it preserves the governance burden, under comparator discipline and with explicit residual and failure conditions. Demonstrated failure of the representation requires its correction. Revision of the underlying governance principle requires demonstration that the failure reaches the principle itself. This gives the amendment grammar, but the object of amendment must remain explicit: adaptive application, current technical formalisation, or a proposed constitutional element.6 A legitimate amendment says: the current formulation fails to preserve this relation, for this locus, under this availability condition, under this closure demand, in these comparator cases, leaving this residual hidden or mismanaged; the proposed revision preserves the relation more strongly by this mechanism, reduces this residual, introduces this new residual, and would itself fail under these conditions. This is not bureaucratic excess. It is the minimum requirement for changing a discipline whose purpose is to prevent ungrounded authority. If the method could be changed without this burden, it would no longer govern formalisation. It would become one more formal language available for institutional preference. Ease of adoption is not an amendment argument. A revision that makes the method easier to adopt may be good if it preserves burden while reducing unnecessary friction. But ease alone is not strength. The same is true of clarity. A clearer formulation is better only if clarity exposes relation more effectively rather than hiding residual. The same is true of certifiability. A certifiable version is better only if certification preserves the burden rather than replacing it. The same is true of machine-readability. An encodable version is better only if encoding carries relation, residual, comparator, safeguard, and update rather than flattening them. A proposed amendment may therefore be rejected even if it improves uptake. It may be rejected because it weakens affected-locus challenge, hides residual, collapses threshold into closure, makes carrier path invisible, turns comparator into theatre, treats release as doctrine, or allows schema fit to replace relation-preservation. A proposed amendment may also be accepted even if it complicates implementation. It may be accepted because it exposes a hidden failure, preserves a relation the old formulation lost, prevents compliance closure, gives affected loci a challenge route, strengthens update, or prevents machine-mediated misclosure. A method governed by relation-preservation must sometimes become harder to apply in order to remain honest. Amendment records matter. Accepted amendments should preserve the failure class, comparator cases, residual reduced, residual introduced, and reason for acceptance. Rejected amendments should also be preserved. A rejected amendment may become relevant later if conditions change, if new cases emerge, if the current formulation fails in a way the rejected amendment anticipated, or if a fork develops its alternative more strongly. Rejection without record becomes forgetting; acceptance without record becomes unexplained authority. The amendment process must therefore preserve branch-space. A method that records only the current formulation creates retrospective inevitability. Later users see the release, not the contested field that shaped it. They may assume that the current version was obvious, natural, or final. A disciplined governance system preserves serious alternatives, reasons for exclusion, residual doubts, and conditions for reopening. Amendment is not weakness. It is the condition under which a finite method remains answerable. But amendment is not preference. It must be earned. The final test therefore begins one level earlier: which proposed constitutional burden is being preserved, strengthened, or challenged, and at what level does the demonstrated failure occur? For revisions to the current technical formalisation, the further questions remain practical and exacting. Does the revision make the claim path more recoverable? Does it make the relation more visible? Does it reduce locus drift? Does it expose abstraction more clearly? Does it distinguish evidence path, record path, and carrier path more strongly? Does it improve comparator discipline? Does it localise residual? Does it prevent threshold from becoming closure? Does it strengthen safeguard? Does it make update more actionable? Does it reduce relational uncertainty without hiding a new residual? Where a constitutional element itself is challenged, these questions remain relevant but are not sufficient: the revision must also show that the failure lies in the principle rather than in its present implementation. Where the answer is yes, amendment may be justified. Where the answer is no, the method must not change merely because change is useful. 8. Forks, dissent, and competitive correction A governed method must preserve dissent. This does not mean every objection is equally strong, every criticism requires amendment, or every dissatisfied actor deserves a competing release. It means that challenge must remain capable of becoming evidence. A method that cannot preserve serious dissent cannot learn from the relations it fails to carry. Dissent is a governance form of epistemic counterfactual correction. It records that the current formulation may not have exhausted the live formulation field under the relation, locus, availability condition, and closure demand being challenged. A critic may expose hidden residual, affected-locus exclusion, comparator weakness, semantic drift, schema inadequacy, certification closure, institutional theatre, prior capture, or an alternative formulation still live under available evidence. The challenge may be wrong, weak, misunderstood, factional, institutionally resistant, or preferential. But if the governance structure cannot preserve it long enough to test whether its exclusion is earned, the method protects its own Soft Horizon rather than the relation under review. Dissent should therefore be localised. What relation does it say the current formulation fails to preserve? For which locus? Under which availability condition? Under which closure demand? What residual does it expose? What comparator case does it introduce? What update does it require? A challenge that cannot answer these questions may remain politically, ethically, or institutionally important, but it has not yet become a method-revision claim. A challenge that can answer them must not be dismissed merely because it is inconvenient. Affected-locus dissent requires special care. Affected loci often encounter method failure before custodians do. They may experience category collapse, record-warrant inflation, residual burial, threshold overreach, safeguard failure, or update exclusion before the institution recognises the problem. Their challenge may initially look anecdotal, emotional, disruptive, or domain-specific. But a governance structure that records only accepted expert objections may miss the first evidence of failure. This does not mean affected-locus challenge is automatically decisive. It means affected-locus challenge must remain capable of becoming evidence. It must have a path by which it can be localised, compared, weighted, and carried into amendment, domain guidance, fork comparison, or residual register. A system that cannot hear affected-locus dissent will eventually treat its own categories as sufficient. Forks are the stronger form of dissent. A fork is not merely disagreement. It is an alternative formulation that preserves enough structure to become a comparator. It says: the current formulation fails here, and this rival formulation preserves the relation better under these conditions. A fork may challenge an operational implementation, a current technical formalisation, or, under the higher burden, a proposed constitutional element. It may also challenge release structure, certification practice, schema design, affected-locus access, comparator discipline, residual carriage, safeguard, or update. Its value is not that it fragments the method. Its value is that it makes the alternative explicit enough to be tested. A fork is therefore a comparator. It allows the current formulation to be tested against a rival rather than protected by incumbency.7 Forks are necessary because a stewarded formulation can become self-protective. It may treat all objections as misunderstandings. It may classify failures as bad implementation. It may delay amendments to preserve stability. It may overvalue continuity. It may privilege actors already inside the custodial process. It may confuse release status with strength. Competitive forks prevent the stewarded formulation from becoming the only imaginable path. But forks are not stronger by existing. A fork must satisfy the same burden as the amendment depth it claims. It must show improved relation-preservation. It must state what residual it reduces and what residual it introduces. It must define its comparator cases. It must show where the current formulation fails. It must identify conditions under which the fork itself would fail. A fork challenging a proposed constitutional element must meet the higher constitutional-challenge burden rather than magnify ordinary disagreement by terminology. No fork becomes stronger by attracting supporters, institutional adoption, public enthusiasm, commercial sponsorship, technical elegance, or symbolic resistance. A fork can be captured too. A regulator may favour a fork because it is easier to administer. A vendor may support a fork because it is easier to encode. A public campaign may support a fork because it produces preferred outcomes. A profession may support a fork because it protects discretion. A steward may oppose a fork because it threatens custody. A community may support a fork because it symbolises resistance. None of these facts decides epistemic strength. Forks must compete by relation-preservation, not faction. This requires fork-comparison discipline. The comparison should not ask which formulation is more popular, more official, more convenient, or more aligned with institutional interests. It should ask which formulation better preserves relation under the cases in dispute. It should identify where the current formulation is stronger, where the fork is stronger, where both fail, where each carries residual, and what further evidence would decide the issue. A fork may have several outcomes. It may be rejected because it fails comparator discipline. It may be retained as an unresolved challenge. It may be partially incorporated as clarification. It may become domain-specific guidance. It may expose an implementation failure rather than a formulation failure. It may trigger a formal amendment. It may replace the current formulation. It may remain as a competing branch where evidence is not yet sufficient to close the question. This range matters because not every unresolved fork should be forced into acceptance or rejection. Some forks identify live residual. The current formulation may remain best, but not closed. A fork may be weaker overall while exposing a local failure. A fork may be premature but important. A fork may be strong in one domain and weak elsewhere. A fork may reveal that the method needs adaptive correction rather than technical amendment, or technical amendment rather than constitutional challenge. A disciplined governance structure must preserve these distinctions. Dissent and forks also protect against retrospective inevitability. Once a formulation survives, it can appear as if it was always the natural path. Alternative formulations disappear from active consideration. Current language becomes familiar. Familiarity becomes authority. Authority becomes trust signal. The method’s own history then repeats the error it criticises in other systems: the surviving formulation is treated as inevitable while the history of the live formulation field and its contraction disappears. Dissent archives preserve branch-memory across Live Field Contraction. They record which formulations remained live, which were excluded, deferred, or left residual, and why the field contracted. If the current formulation later fails, the archive may show that a prior challenge had located the weak point or that an alternative ceased to remain live without adequate discriminatory pressure. This is epistemic and historical branch-memory; it does not establish that every archived alternative was physically available within any Information Processing Locus’s Counterfactual Inclusion Field. If the challenge is erased, the method loses memory. If preserved, the method can update more intelligently. Forks also protect against machine capture. Machine-mediated implementations tend to stabilise the fields they can represent. Once a schema is built, the schema becomes hard to challenge because institutions, software, vendors, and records depend on it. A fork may be needed to show that the machine-readable version has become too narrow. It may preserve non-encodable residual, qualitative affected-locus relation, or comparator structures that implementation excluded. Without such forks, the schema may become the practical method. Forks also protect against certification capture. A certification market may favour standardised, passable, repeatable criteria. A fork may show that certification has drifted from relation-preservation into process completion. It may preserve stronger comparator burden, better residual visibility, affected-locus relation, or more honest limits on what certification can claim. Without fork pressure, certification becomes closure by market habit. Dissent and forks therefore perform a constitutional function. They keep the method from becoming identical with its current stewarded form, its founding text, the authority of its originating author, or any other operational Soft Horizon. The current formulation may be strongest. It may deserve stability. It may reject many challenges. But it must remain answerable to the possibility that a challenge identifies a relation it has failed to preserve. This does not require fragmentation. Governance should not encourage endless branching for its own sake. Too many weak forks can create confusion, reduce trust, and make application unstable. The point is not to multiply alternatives indefinitely. The point is to preserve a disciplined route by which serious alternatives can be stated, compared, and retained where necessary. The governing sentence is: Dissent and forks are legitimate where they preserve relation-bearing challenge; they become capture where they seek strength from faction, adoption, ownership, or preferred outcome. A mature method should not fear forks. It should fear untested stability. 9. Prohibited capture routes Governance must identify capture routes before they become normal practice. Capture rarely announces itself as capture. It usually appears as implementation, clarity, adoption, assurance, certification, accountability, efficiency, scale, accessibility, urgency, or stewardship. Each may be legitimate. Each may also become a route by which something other than demonstrated relation-preservation carries epistemic strength. The first prohibited route is compliance closure. Compliance closure occurs when completion of a method process is treated as adequacy. A form is completed. A template is filled. A checklist is satisfied. A claim packet exists. The institution then treats completion as evidence that the relation was preserved. This must be prohibited because completion is not warrant. The question remains whether the completed form preserved relation, residual, comparator, safeguard, and update. The second route is certification closure. Certification may support reliance where it records that a process was performed under stated conditions. But certification becomes capture when the certificate is treated as proof of substantive adequacy. A certificate may show that a review occurred. It does not by itself show that the review preserved the affected relation. Certification must therefore state what it certifies, what it does not certify, what residual remains, and what conditions would require re-audit.8 The third route is schema-fit capture. A schema is useful because it makes some structures stable, interoperable, machine-readable, and reviewable. But schema fit becomes capture when what can be encoded becomes what the method is allowed to see. Relation becomes a field. Residual becomes a status. Comparator becomes a menu. Safeguard becomes a flag. Update becomes a scheduled event. The schema then replaces the method’s burden. This must be prohibited because machine-readability is not relation-preservation. The fourth route is foundation, steward, originating-author, canonical-text, or constitutional-label authority capture. Stewardship and provenance are necessary: releases, archives, amendments, dissent records, residual registers, fork comparisons, originating reasoning, and canonical text all need recoverable custody. Capture occurs when that custody or priority is treated as epistemic authority: steward acceptance becomes truth, original wording becomes privileged because it is original, canonical status becomes doctrine, or the label constitutional becomes a shield against challenge. Provenance should be preserved and custody should remain stable, but neither author, text, steward, nor constitutional label owns the method’s epistemic conditions. A related prohibited migration is Responsibility Warrant laundering. Governance may need to allocate answerability, supervision, stewardship, liability, duty, or decision ownership. Those allocations may be legitimate Responsibility Warrants where their scope and authority are explicit. They become capture when the office, role, or mandate is treated as proof of Responsibility Provenance for the underlying consequence, or when provenance is treated as transferred merely because governance has allocated answerability elsewhere. Authority may allocate warrant. It cannot manufacture or erase provenance. The fifth route is proprietary implementation capture. A vendor may build tools that assist formalisation. Such tools may preserve logs, prompt missing conditions, track residuals, compare cases, and support update. But proprietary implementation becomes capture when the tool hides the audit path, narrows the method to what the product can represent, or makes users dependent on a black-box implementation of governance. A tool may assist the method. It must not become the method by market position. The sixth route is adoption validation. Adoption may show usefulness, visibility, legitimacy demand, or operational need. It does not show epistemic strength. A method is not stronger because regulators, courts, corporations, platforms, vendors, public bodies, or campaign groups adopt it. Adoption becomes capture where the fact of uptake is used to silence criticism, close residual, or weaken amendment burden. The seventh route is clarity-as-closure. A clearer formulation may be an improvement where it exposes relation more accurately. But clarity may also simplify away residual. A method may become easier to teach because it has removed the difficult burden. It may become easier to certify because it has narrowed the audit path. It may become easier to encode because it has flattened qualitative relation. Clarity must therefore be judged by whether it improves recoverability, not by whether it feels more definite. The eighth route is comparator theatre. Comparator discipline requires real alternatives capable of testing the claim. Comparator theatre occurs when a method, institution, or certification process lists alternatives only to create the appearance of review. The alternatives are too weak, irrelevant, already excluded by design, or never allowed to affect the outcome. Comparator theatre is prohibited because it simulates correction while protecting closure. The ninth route is action-threshold conversion. Finite systems often must act before final closure. Threshold action may be justified by urgency, precaution, reversibility, duty, or risk. But action-threshold conversion occurs when the fact that action was justified is later narrated as if the relation had been closed. This route must be prohibited because action under residual is not proof. The stronger the action demand, the more important residual, safeguard, and update become. The tenth route is safeguard theatre. Safeguard is real only where it constrains exposure. A review right that cannot be used, an appeal that cannot reach the relevant relation, a human oversight step that rubber-stamps machine output, a consultation that cannot change the decision, or a monitoring process that cannot trigger update is not an effective safeguard. It is theatre. Governance must distinguish safeguard language from safeguard function. The eleventh route is update inaccessibility. Update inaccessibility occurs when revision is said to be possible, but the actors most likely to encounter failure cannot trigger review. Affected loci lack access. Machine users cannot report residual. Records cannot be corrected. Appeals are too narrow. Monitoring looks only at metrics that miss the target relation. Implementation teams cannot escalate conceptual failures. Review occurs only on a timetable disconnected from harm. This route must be prohibited because update is not real unless relation can answer back. The twelfth route is moral urgency capture. Moral urgency occurs where harm is real, stakes are high, and action may be required before closure. That urgency may justify threshold action. It does not erase residual, comparator, affected-locus complexity, safeguard, or update. Moral urgency capture occurs when a desired outcome is treated as reason to weaken epistemic burden. The system may need to act quickly. It may not pretend that speed has closed the relation. These routes share a common structure. Each allows something other than demonstrated relation-preservation to carry epistemic strength. Compliance carries it. Certification carries it. Schema fit carries it. Foundation approval, originating authority, canonical status, or constitutional label carries it. Proprietary implementation carries it. Adoption carries it. Clarity carries it. Comparator display carries it. Action threshold carries it. Safeguard language carries it. Inaccessible update carries it. Moral urgency carries it. In each case, an operational or governance closure hardens beyond its authority. The governance discipline rejects each substitution. A method that permits these routes may still look formal, disciplined, transparent, public, modern, scalable, and institutionally useful. It may be widely adopted. It may be easy to teach. It may be certifiable. It may be machine-readable. It may be politically reassuring. But it will no longer perform its central function. It will have become a route by which unresolved relation is converted into authority. The governing sentence is: No form of use, approval, authorship, canonical status, implementation, certification, adoption, clarity, urgency, constitutional labelling, or stewardship may substitute for relation-preservation. These prohibitions do not prevent practical use. They make practical use honest. An institution may comply, but compliance is not closure. A certificate may be issued, but it must state its limits. A schema may be used, but it must mark what it cannot carry. A steward may release, but not own. A proprietary tool may assist, but not hide the audit path. Adoption may spread, but not validate. Clarity may help communication, but not erase residual. Urgency may justify action, but not certainty. Safeguards may reassure only where they constrain. Update may be claimed only where it can be triggered. Governance succeeds where these distinctions remain visible under pressure. Governance fails where the method becomes easier to cite than to obey. 10. Release discipline and best available formalisation Formalisation needs release discipline. A method cannot operate as a permanently fluid conversation. Institutions need to know what version they are using. Courts need to know which formulation is being cited. Regulators need to know which requirements apply. Implementers need stable reference. Users need common terms. Critics need a target for challenge. Machine systems need defined structures before any governed encoding can occur. Without release discipline, every application begins from scratch, and no comparator can test whether the method is improving. A release is therefore a necessary custodial Soft Horizon. It closes the current formulation into usable form while preserving residual, amendment history, dissent, fork comparison, and update condition. A governed release should also distinguish what it presently treats as proposed constitutional elements, what belongs to the current technical formalisation, and what is domain guidance or operational implementation. It may stabilise terms, burdens, support documents, schemas, templates, and amendment notes, but stable publication must not constitutionalise technical detail by repetition. It may support teaching, application, review, and public challenge. But release discipline must not become release theology. Release theology occurs when the release is treated as doctrine because it has been released. The current version becomes authoritative because it is current. The archive becomes sacred rather than useful. The steward becomes guardian of truth rather than custodian of correction. The release number becomes the argument. Users ask whether the current release was followed, not whether relation was preserved. Critics are treated as enemies of stability. Amendments are treated as threats to legitimacy. A release must therefore carry its own limits. It should state what it settles, what it leaves residual, which proposed constitutional elements remain under live challenge, which technical structures are provisional, contested, or being monitored, what domain translations remain local or provisional, what implementation risks exist, what forks remain live, what rejected amendments may become relevant under future conditions, and what would force revision. A release without these limits is not disciplined. It is an overclosed Soft Horizon. Release discipline also requires version memory. A later release should not merely replace an earlier one. It should state why the change occurred, which failure class prompted it, what relation was better preserved, what comparator cases supported it, what residual was reduced, what residual was introduced, and what prior branches were rejected or deferred. Without this memory, release history becomes authority without argument. The proper governance state is Best Available Formalisation. A Best Available Formalisation is not final doctrine and not constitutional settlement. It is not mere draft. It is the current strongest formal expression of the proposed constitutional burden under stated conditions: strong enough to organise use and corrigible enough to remain answerable to failure. It is available not because it is merely present, but because its claim path and reliance path can be inspected, challenged, compared, and revised. Best Available Formalisation status does not elevate every component of the current technical architecture into a constitutional element. This concept is necessary because formalisation faces two opposite dangers. The first danger is premature finality. A formulation is released, adopted, taught, certified, encoded, or cited, and then treated as doctrine. Its current structure becomes protected by familiarity. Its terms become institutional. Its forms become habits. Its authority becomes easier to invoke than to test. Correction becomes expensive because many actors have built on the formulation. The Best Available Formalisation is then misread as final truth. The second danger is permanent provisionality. A formulation is never allowed to stabilise because every challenge, preference, uncertainty, domain difficulty, and implementation problem is treated as equally destabilising. The method cannot become usable. Institutions cannot learn it. Comparators cannot test it. Forks cannot be meaningfully compared. Every application starts from scratch. The Best Available Formalisation is then never allowed to perform its stabilising function. Both failures must be avoided. A Best Available Formalisation must be strong enough to resist manipulation. It must not bend whenever a regulator wants administrability, a vendor wants certification, a court wants procedural fit, a profession wants discretion, a public body wants reassurance, or a machine system wants encodable fields. It must resist preference-led removal of proposed constitutional burdens and unjustified weakening of current technical burdens while remaining open to demonstrated technical amendment and, under the higher burden, constitutional challenge. It must also be weak enough to remain corrigible. It must not protect itself from affected-locus challenge, failure cases, comparator pressure, implementation evidence, domain difficulty, or fork correction. It must preserve the conditions under which it can be shown to fail. Strength without corrigibility becomes doctrine. Corrigibility without strength becomes drift. The Best Available Formalisation holds the middle. It should therefore carry its own residual. It should say what remains unresolved, which domains require further application, which edge conditions are not yet settled, which implementation risks remain, which terms require monitoring, and which failure modes would force revision. A formalisation that does not carry residual is not best in the required sense. It is overclosed. It should also carry its own comparator field and branch-memory. It should preserve rejected alternatives, serious forks, unresolved challenges, reasons for exclusion, and enough of the live formulation field to show how the present formulation stabilised. If prior alternatives and their exclusion conditions disappear, the current formulation can appear inevitable and correction weakens. Where the history of Live Field Contraction remains recoverable, stability remains accountable. It should carry its own update conditions. It should state what kind of evidence would force clarification, domain translation, support-form correction, fork reconsideration, or amendment. This does not mean every possible update can be predicted. It means the formalisation does not pretend that release ends the question. The Best Available Formalisation is therefore a governance state, not merely a document state. It is the condition in which a formulation is strong enough to be used, taught, applied, and tested, while still explicitly answerable to future correction. It is the condition under which a finite discipline can enter public life without pretending to have escaped finitude. This matters because governance must not confuse the strength of a current formulation with possession of final form. The method may be the best available way of preserving claim-strength under finite conditions. It may still require later correction. It may be strong enough to reject weak challenges. It may still preserve unresolved residual. It may be stable enough for institutional use. It may still refuse certification closure. It may be precise enough for guidance. It may still reject schema reduction. The release should therefore say, in effect: This is the current strongest formalisation. It is stable enough for use and explicit enough to expose which elements are presently proposed as constitutional, which belong to the current technical architecture, and which are local applications. None is final merely because it has been released. The constitutional proposal remains open to demonstrated challenge, the technical architecture remains open to justified amendment, and operational translations remain local to their conditions. Its authority lies in the relation-preserving burden it carries, not in release, authorship, or canonical status. That is release discipline without release theology. 11. Conclusion: governance without ownership Formalisation cannot govern ungrounded authority if it becomes ungrounded authority itself. This is the constitutional problem. A discipline introduced to keep load-bearing claims answerable to relation will fail if its own name, release, steward, template, certificate, schema, compliance process, institutional adoption, or machine implementation begins to carry authority beyond its grounding. The method can become the very Soft Horizon it was designed to discipline. The solution is not to keep formalisation outside institutions. That would make it pure but unusable. A method that cannot be adopted, taught, implemented, cited, contested, revised, or embedded in practice cannot help govern claim-strength where claim-strength becomes consequential. Formalisation must enter institutions, law, governance, professional practice, public administration, certification, and machine-mediated systems. It must become usable. The solution is to govern its usability. Operational Soft Horizons are necessary. Releases, templates, certificates, schemas, stewards, and implementations reduce repeated audit. They allow finite users to rely on prior formalisation-work. They stabilise memory, support review, and make application possible. But they remain trust carriers, not trust grounds. They are legitimate only where their reliance-strength remains indexed to recoverable relation-preservation, residual, scope, safeguard, and update. Governance without ownership follows from this. No originating author, founding or canonical text, foundation, institution, profession, court, regulator, public body, vendor, machine system, or political majority owns the conditions by which claims become evaluable. They may propose, preserve provenance, inspect, challenge, evidence, apply, contest, fork, steward, implement, and correct. They may not make authorship or canonical status the ground. They may not vote away relation. They may not replace locus with institutional vantage. They may not treat availability as warrant. They may not collapse threshold into closure. They may not hide abstraction behind clarity. They may not treat record, certificate, schema, release, compliance, adoption, or constitutional label as epistemic ground. Stewardship is necessary, but stewardship is not ownership. A steward maintains releases, archives, dissent records, amendment histories, residual registers, and fork comparisons. It keeps correction paths recoverable. It does not make the method true. Institutional use is necessary, but use is not validation. A court, regulator, hospital, platform, public body, corporation, or professional body may apply the method within domain constraints. It does not prove adequacy by application. Machine implementation may assist, but implementation is not definition. A schema may support consistency, logging, monitoring, and comparison. It must not become the practical boundary of what the method can see. Public inspection is necessary, but preference is not epistemic authority. The public may challenge, evidence, test, and fork. It may not remove audit burdens because they are inconvenient or unpopular. The method may change only through demonstrated relation-preservation. Its procedures remain answerable to its proposed principles, and those proposed principles remain answerable where relation demonstrates their failure. A revision is legitimate where it shows that the current formulation fails to preserve relation in a defined class of cases, and that the proposed revision preserves relation more strongly under finite conditions. It must state whether the failure lies in adaptive application, the current technical formalisation, or a proposed constitutional element. It must state the relation, locus, availability condition, closure demand, comparator cases, residual reduced, residual introduced, and its own failure condition. Where constitutional revision is claimed, it must show that the principle itself remains implicated after plausible operational and technical explanations have been tested. It must show stronger recoverability, not merely easier adoption, clearer packaging, better certification, smoother encoding, institutional comfort, or preferred outcome. This rule keeps the method stable enough to resist manipulation and corrigible enough to learn. That is the aim of constitutional governance here. Not ownership. Not popularity. Not custodial control. Not institutional convenience. Not final doctrine. Not permanent drift. The aim is correction without capture. The proposed constitutional elements belong inside that correction path. Their purpose is to preserve the methodological identity of TEM across technical change, not to create a final layer immune from correction. A constitution that could not be challenged by demonstrated relation would reproduce the closure this paper is designed to prevent. Formalisation remains legitimate only where it remains answerable to relation-preservation. It may enter the world through releases, stewards, templates, certificates, schemas, courts, regulators, vendors, public bodies, and machines. But wherever it travels, it must carry its own residual, comparator field, safeguard, and update condition. A method that cannot be corrected becomes doctrine. A method that can be altered by preference becomes capture. A governed method must remain between those failures: usable without being owned, stable without being final, public without being preference-led, stewarded without becoming priestly, implemented without becoming reduced, and corrigible without surrendering its burden. The method does not belong to its steward. It belongs to the relation it is able to preserve. The handoff to the final paper is therefore precise. Constitutional governance has established how formalisation remains stable enough for use, exposed enough for challenge, and corrigible without allowing release, stewardship, adoption, authority, or implementation to become epistemic ground. The burden of amendment must remain proportionate to the level being altered, the dependency reach of the proposed revision, and its downstream consequences. The Epistemic Method must now consolidate the inherited architecture into a general discipline of relational audit under finite conditions, while preserving the same final guardrail: method-compliance must not become truth or closure. Endnotes References Cihon, P., Kleinaltenkamp, M. J., Schuett, J., & Baum, S. D. (2021). AI certification: Advancing ethical practice by reducing information asymmetries. IEEE Transactions on Technology and Society, 2(4), 200-209. Hart, H. L. A. (1961). The Concept of Law. Clarendon Press. Hirschman, A. O. (1970). Exit, voice, and loyalty: Responses to decline in firms, organizations, and states. Harvard University Press. Longino, H. E. (1990). Science as social knowledge: Values and objectivity in scientific inquiry. Princeton University Press. Meyer, J. W., & Rowan, B. (1977). Institutionalized organizations: Formal structure as myth and ceremony. American Journal of Sociology, 83(2), 340-363. https://doi.org/10.1086/226550 Nyman, L. (2015). Understanding Code Forking in Open Source Software: An Examination of Code Forking, Its Effect on Open Source Software, and How It Is Viewed and Practiced by Developers. Hanken School of Economics. O'Mahony, S., & Ferraro, F. (2007). The emergence of governance in an open source community. Academy of Management Journal, 50(5), 1079-1106. https://doi.org/10.5465/amj.2007.27169153 Ostrom, E. (2005). Understanding Institutional Diversity. Princeton University Press. Power, M. (1999). The Audit Society: Rituals of Verification. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780198296034.001.0001 Stigler, G. J. (1971). The theory of economic regulation. The Bell Journal of Economics and Management Science, 2(1), 3-21. https://doi.org/10.2307/3003160 Timmermans, S., & Berg, M. (1997). Standardization in action: Achieving local universality through medical protocols. Social Studies of Science, 27(2), 273-305. https://doi.org/10.1177/030631297027002003 Endnotes 1.　Institutional and verification comparators: Meyer and Rowan (1977) show how formal organisational structures can acquire legitimacy while becoming decoupled from ongoing activity; Power (1999) examines the expansion of audit and formal scrutiny and the possibility that verification practices exceed their operational capabilities or generate dysfunctional effects. Paper 7 uses these as comparators for symbolic success and form-as-authority. It does not inherit institutional theory or audit-society machinery; its additional burden is that governance carriers remain answerable to relation-preservation, residual, comparator, safeguard, and update. 2.　Restricted capture comparator: Stigler (1971) develops an economic theory in which regulation may be acquired and operated primarily for the regulated industry's benefit. Paper 7 uses regulatory capture only as a domain specialisation. Its capture concept is wider, covering any operational or custodial pressure that preserves a method's authority or legitimacy function while weakening its relation-preserving burden; no public-choice framework is inherited. 3.　Social-objectivity comparator: Longino (1990) argues that scientific objectivity depends in part on social critical interaction. Paper 7 treats public and affected-locus criticism as a necessary correction pathway, but not as epistemic ownership: criticism may expose relational failure, while claim-strength still depends on demonstrated relation-preservation rather than participation or consensus. No wider social-epistemology framework is inherited. 4.　Standards, adaptation, and level-structure comparators: Timmermans and Berg (1997) describe 'local universality', in which protocols become workable through situated transformation while remaining grounded in practice. Ostrom (2005) separately distinguishes operational-choice, collective-choice, and constitutional-choice levels in institutional analysis. These are strong specialisations of Paper 7's adaptive-application and level-separation problems. Paper 7's three levels - adaptive application, current technical formalisation, and proposed constitutional elements - govern an epistemic method rather than institutional rule-making authority, and no IAD or standards-governance framework is inherited. 5.　Governance specialisation: O'Mahony and Ferraro (2007) show an open-source community developing formal authority while limiting and revising it through democratic mechanisms. The case is a useful operational comparator for corrigible governance and authority distribution. Paper 7's epistemic, custodial, and operational authority distinction is different: democratic or custodial authority can expose and govern correction but cannot create epistemic strength by status. 6.　Rules-of-change comparator: Hart (1961) distinguishes secondary rules governing how a legal system identifies, adjudicates, and deliberately changes its rules. That makes rules of change a useful jurisprudential specialisation of Paper 7's now-explicit amendment grammar. The paper does not inherit Hart's legal validity architecture: amendment here is not legitimate because an authorised rule of change permits it, but only where demonstrated relational failure is localised and the proposed revision preserves the relevant relation more strongly under stated comparator, residual, and failure conditions. 7.　Fork and dissent comparators: Hirschman (1970) analyses exit and voice as distinct responses to organisational decline. Nyman (2015) uses 'competitive forks' in open-source software for independently developed competing projects, providing prior terminology and an operational fork specialisation. Paper 7's dissent and competitive-fork architecture is broader: internal challenge and explicit rival formulations become correction mechanisms only where they preserve relation-bearing comparison, residual, scope, and update conditions. A fork is not merely exit, dissent is not merely voice, and no open-source or organisational-response framework is inherited. 8.　Certification comparator: Cihon et al. (2021) analyse AI certification as a mechanism for reducing information asymmetries and emphasise the need for certification symbols to correspond to substantive implementation and programme goals. This is a close domain specialisation of Paper 7's certification-closure guardrail. The present architecture treats a certificate as a potentially legitimate operational trust carrier, not as epistemic ground: certification must remain limited to what was actually assessed, preserve residual and re-audit conditions, and must not substitute for demonstrated relation-preservation.