The Gödel Horizon got-v1-3a-1 | Canonical paper | Issue 1 https://goto-go.com/machine/documents/got-v1-3a-1/ GOT-V1-3a-1 The Gödel Horizon Relational Availability, Closure, and the Discipline of the Unknown Abstract A finite observer-system does not first encounter reality as named object, completed fact, stable category, proven theorem, or closed explanation. It encounters relation under conditions of availability. Something crosses a boundary, becomes salient, resists, perturbs, contrasts, registers, or demands response. Before recognition, there is availability. Before identification, there is sensed or operational relation. Before knowledge, there is the possibility that something has become present for a locus without yet being closed by that locus’s abstraction regime. Relational availability is the foundational condition of horizon analysis. It asks what relation is available to which locus, through what boundary, interface, scale, horizon, abstraction, branch path, and closure demand. A relation may be unavailable, available in principle, available in potential, coupled, available as contrast, registered, retained, deployed, closed, or residual for a specified locus under stated reference conditions. These statuses are not a strict ladder. They are condition distinctions that prevent existence, access, registration, warrant, closure, and residual from collapsing into one another. A horizon is a boundary in relational availability and closure for a specified locus. It is not an object, location, ontological zone, or mysterious edge of reality. It is a locus-relative boundary condition produced when a specified relation reaches, fails to reach, crosses, stalls within, or overclaims a condition of availability or closure. Availability supplies the access condition. Closure supplies the claim-status. A relation may enter availability without recognition, become coupled without registration, become registered without warrant, become usable through a Soft Horizon while residual remains, or become formulable in a way that discloses the limit of closure within the abstraction regime that formulated it. The Gödel Horizon is the boundary at which relation has become available or formulable under a closure demand, while the current abstraction regime discloses that it cannot close that relation from within its own present resources. Gödel’s incompleteness theorems supply the exact formal exemplar: a system can formulate a statement under its own resources while proof-closure cannot be supplied by those same resources. Gödel did not create the horizon. He disclosed the formal version of a wider finite-observer condition: relation can become available under a regime before the regime can close it. The Soft Horizon is the operational counterpart of this condition. A Soft Horizon is provisional, purpose-indexed closure under residual. It is how a finite observer-system treats relation as resolved enough for recognition, retention, communication, calculation, classification, decision, or action without possessing the relation as total state. Soft Horizons are not defects. Perception, language, mathematics, scientific modelling, legal classification, institutional procedure, and machine classification all depend on provisional closures. The danger begins when such closures are reified: when their residual, scope, tolerance, comparator, and update conditions no longer travel with them. This paper establishes the grammar of horizon analysis. It begins with relation before recognition. It defines relational availability and its main condition distinctions. It then shows how recognition produces Soft Horizons, how misclosure arises, and why horizons should be analysed as availability/closure boundary conditions rather than objects, locations, or metaphysical limits. Later sections extend this grammar to loci, coupled loci, procedural availability, abstraction regimes, latent abstraction boundaries, Gödel disclosure, extension cases, branch-space, and method discipline. The governing rule is simple: no horizon can be analysed without specifying the relation, the locus, the availability condition, and the closure demand. 1. Introduction: availability, closure, and the first boundary The unknown is not first a blank. A finite observer-system does not meet reality as total state and then subtract what it fails to know. It encounters relation through boundary, interface, scale, horizon, discrimination regime, abstraction, registration, retention, redeployment, comparator pressure, and residual limit. What is unknown is therefore not merely what lacks a name, proof, category, explanation, or measurement. It may be relation outside the present access conditions of the locus. It may be relation that has entered availability without recognition. It may be relation recognised under a pattern that later fails correction. It may be relation closed enough for action while remaining residual for explanation. It may be relation formulable within a regime whose own resources cannot close it. This changes the starting point of epistemic analysis. The basic opposition is not known versus unknown. That opposition is too late. It assumes that the relevant relation has already entered a form in which possession or non-possession can be judged. Before that opposition, a prior question governs: under what conditions has relation become available, unavailable, potentially available, coupled, registered, retained, deployed, closed, or residual for the locus under review? Relational availability is the answer to that prior question. It is not a theory of existence as such. It does not decide what reality contains in itself, and it does not reduce reality to what a finite locus can encounter. It states the condition under which a target relation can matter for a specified observer-system, instrument, organism, institution, formal community, machine, or coupled locus. A relation may exist without being available. It may be available in principle without being available in potential for the present locus. It may be available in potential without having entered actual interaction. It may be coupled without being registered. It may be registered without being retained. It may be retained without being warranted. It may be deployed without being closed. It may be provisionally closed while residual remains. A horizon appears wherever these conditions produce a boundary. The boundary may concern access, recognition, registration, retention, proof, explanation, classification, action, or closure. It may appear as absence, anomaly, obscurity, paradox, undecidability, residual, or provisional use. The important point is that the horizon is indexed. It belongs to the relation among locus, target, boundary, interface, scale, abstraction, path condition, and closure demand. It is not a floating metaphysical edge. It is a boundary in relational availability and closure. This makes the Gödel Horizon wider than mathematics, but not loose. The name belongs to the formal case because Gödel’s incompleteness theorems give the most exact instance: a system can formulate a statement under its own resources while proof-closure cannot be supplied by those same resources. The structure is exact because availability and closure are separated. The statement is not absent from the regime. It is available as formulation. The closure demand is not vague. It is proof within the system. The failure is not mere psychological inability or historical inconvenience. It is disclosed by the formal structure itself. The mathematical case is therefore exemplary rather than originative. The horizon begins wherever relation enters availability before the relevant closure demand can be satisfied. The primitive observer senses something without knowing what it is. The perceptual system closes relation into a usable world without possessing the target whole. The scientific system registers anomaly before explanatory closure. The institutional system acts through categories that may preserve procedure while losing relation. The computational system produces outputs under abstraction and training constraints. The formal system expresses what it cannot prove. Across these cases, finite systems operate at boundaries between availability and closure. The Soft Horizon must be named alongside the Gödel Horizon because finite systems do not only fail to close. They also close provisionally in order to function. An observer-system unable to produce provisional closure would be flooded by relation without usable form. It could not distinguish figure from ground, object from field, signal from noise, evidence from event, theorem from inscription, word from sound, risk from background, or action from suspension. Soft Horizons are therefore necessary. They make relation usable under finite conditions. A Soft Horizon becomes legitimate when its residual remains attached. It says, in effect: this relation is closed enough for this purpose, under this abstraction, at this scale, within this tolerance, for this action or inference, with this residual, and under this update condition. Such closure is not weak merely because it is incomplete. It may be the strongest available finite output. The weakness appears only when the Soft Horizon hardens into possession, when provisional closure travels downstream as if it were total closure, settled warrant, final ontology, or non-revisable authority. The discipline required by horizons is therefore double. The observer-system must not deny availability where relation has entered access. It must also not overclaim closure where availability has not satisfied the closure demand. Denial produces blindness to relation. Overclosure produces false possession. Reification allows necessary provisional closure to become unearned authority. This paper develops that discipline from the first boundary upward. It begins with relation before recognition. It defines relational availability as the foundational condition of horizon analysis. It distinguishes availability conditions without turning them into a strict ladder. It then shows how recognition produces Soft Horizons, how misclosure occurs, and why horizons should be understood as availability/closure boundary conditions rather than objects, locations, or ontological zones. Later sections extend this grammar to coupled loci, procedural availability, abstraction regimes, latent abstraction boundaries, Gödel disclosure, and the method discipline required wherever finite systems must claim, classify, model, prove, decide, or act without total possession. The governing rule is simple: No horizon can be analysed without specifying the relation, the locus, the availability condition, and the closure demand. Without relation, the horizon becomes empty metaphysics. Without locus, availability becomes absolute. Without availability condition, access is treated as undifferentiated. Without closure demand, failure becomes vague. Once those conditions are stated, the unknown becomes structured. It can be unavailable, available in principle, available in potential, coupled, available as contrast, registered, retained, deployed, provisionally closed, misclosed, residual, or disclosed as a boundary of the abstraction regime itself. The unknown is then no longer a blank. It is a disciplined field of finite relation. 2. Relation before recognition Observation does not begin with identity. A finite observer-system does not first meet a world already divided into stable objects, names, classes, proofs, causes, values, risks, meanings, and explanations. Those are later achievements of pattern, memory, abstraction, comparison, retention, correction, and use. The more primitive condition is encounter. Something becomes available. A difference crosses a boundary. A contrast enters a field. A perturbation reaches or affects a locus. A pressure, movement, sound, scent, material alteration, resistance, interruption, or anomaly becomes present for that locus. At this level, the observer-system need not know what has been encountered. It need not identify the source, assign the class, explain the cause, prove the relation, classify the event, or close the claim. The first structure is not recognition. It is availability. Availability is narrower than existence and broader than knowledge. A relation may exist without becoming available to the locus under review. A sound may occur outside hearing range. A molecular difference may be real while no receptor can bind it. A risk may exist without entering a decision system. A mathematical relation may hold before any formalism has made it expressible. A social harm may occur before institutional categories can register it. A physical process may produce no accessible physical effect in the available instrument path. Existence alone does not establish availability for a finite observer-system. Availability begins where relation enters access conditions. Something reaches an interface. Something survives a threshold. Something stands out from background. Something becomes salient enough to perturb, register, guide attention, alter state, or invite classification. The relation may remain obscure. It may be weak, partial, noisy, ambiguous, unstable, distorted, or mis-scaled. It may not yet be recognised. It may not yet be retained. It may not yet be usable. But it is no longer outside the access field of the locus in the same way. The difference matters because beginning from recognition gives the wrong structure of ignorance. If observation begins from recognition, then failure to identify appears as ignorance after an object has already been given. The observer seems to possess the thing and merely lack the correct name. If observation begins from availability, the primitive condition changes. The observer does not first possess an object and then fail to label it. The observer first receives relation under constraint. Identity is a later closure attempt. This is visible in ordinary perception. A movement in darkness is sensed before it is identified. A sound behind a wall is heard before its source is known. A bodily pain is felt before its cause is diagnosed. A shadow becomes present before it is resolved as object, animal, person, branch, or illusion. In each case, relation is available before recognition is closed. The same structure extends beyond sensory perception. An experiment may register an anomaly before theory can explain it. A legal institution may receive testimony before its categories can classify the harm. A medical system may observe symptoms before diagnosis. A scientific field may measure an effect before mechanism. A culture may sense instability before inherited concepts can name the transition. A formal system may formulate a statement before it can prove or refute it. A computational system may detect an outlier before it can assign a warranted class. The basic sequence is therefore not knowledge first, but availability first. Availability opens the possibility of recognition. Recognition attempts closure. Closure may be earned, soft, residual-bearing, unstable, false, suspended, or unavailable under the demand being made. The horizon appears wherever availability and closure come apart. It appears where relation becomes available but recognition is not yet possible; where recognition occurs but fails under correction; where provisional closure works for one purpose but not another; where action must proceed before closure; or where an abstraction regime discloses a boundary in its own capacity to close what it can formulate. The primitive horizon is presence without identity. Presence does not mean total possession. It means that a relation has crossed into the access field of the locus sufficiently to matter for that locus under the relevant conditions. The relation may be faint, partial, noisy, or unstable. It may be sensed only as anomaly, disturbance, tension, salience, pressure, or possibility. Yet something has become available. Identity requires more. Identity requires that the observer-system’s abstraction regime can stabilise the available relation as this rather than that. It requires patterning. It requires retained comparators. It requires a regime of sameness and difference. It requires the ability to treat some differences as relevant and others as background. It requires enough closure to say what kind of relation has been encountered. Between presence and identity lies the first horizon. The observer-system is not outside relation. It has not failed to encounter anything. Something is present. But the current pattern regime cannot yet close what is present into stable identity. The relation is sensed but not known as. It is available but not identified. It has crossed a boundary but not yet the recognitional threshold. This state is not empty ignorance. It has structure. The observer-system may know that something is present. It may know that the available relation is unusual. It may know that the relation does not fit inherited categories. It may know that response is required. It may know that further investigation is needed. Such knowledge is not identity closure. It is relation under unresolved availability. Unrecognised availability can be handled in several ways. One response is premature closure. The observer-system forces the available relation into the nearest retained pattern. The unknown sound becomes an intruder. The unfamiliar illness becomes a familiar diagnosis. The anomalous datum becomes measurement error. The strange phenomenon becomes an instance of an inherited class. Premature closure reduces uncertainty by overusing available patterns. Another response is suspension. The observer-system refuses closure and carries the relation as unresolved. Suspension may be disciplined or evasive. Disciplined suspension preserves the relation as available but unclosed. It protects inquiry from false recognition. Evasive suspension avoids the burden of classification without localising the missing condition. A third response is pattern search. The observer-system attempts to extend, refine, or replace the pattern regime. New observations are gathered. Comparators are introduced. The scale is changed. The abstraction is revised. A new category is formed. A stronger model is built. The horizon may move if the relation becomes closeable under a revised regime. The primitive horizon therefore does not merely mark failure. It initiates method. It asks what has become available, for which locus, through which access condition, under which closure demand, and what would be required for recognition, classification, proof, explanation, action, or residual statement to become warranted. The earliest form of the unknown is not absence. It is relation available before identity. 3. Relational availability: conditions and grammar Relational availability is the condition under which a relation becomes unavailable, available in principle, available in potential, coupled, available as contrast, registered, retained, deployed, closed, or residual for a specified locus under stated reference conditions. This definition shifts the unit of analysis. The primary object is not an isolated thing and not an unindexed observer. The primary object is a relation under conditions. The question is not first whether something exists, whether someone knows it, or whether a category already names it. The question is what relation is available to which locus, through what boundary, interface, scale, horizon, abstraction, branch path, and closure demand. A locus is the unit for which the relation is being evaluated. It may be an organism, person, cell, instrument, scientific practice, formal mathematical community, legal institution, computational system, machine-mediated process, social organisation, or coupled observer-network. The same target relation may be unavailable to one locus, available in principle to another, available in potential to a third, coupled to a fourth, registered by a fifth, and closed under a specific demand by a sixth. Availability cannot be detached from the locus for which availability is claimed. The relation is the structured connection under review. It may be physical, perceptual, informational, mathematical, social, legal, institutional, computational, explanatory, evidential, causal, or practical. It is not necessarily a material object. A relation may be a signal entering a detector path, a proof obligation in a formal system, a harm entering a legal process, a category cut in an institution, a statistical pattern in a model, a symptom constellation in medicine, a measurement relation in science, or a possible branch of action under uncertainty. Reference conditions state the boundary under which the claim is evaluated. They include the relevant boundary, interface, scale, operational horizon, discrimination regime, abstraction, noise condition, registration pathway, retention capacity, comparator field, branch-space, and closure demand. Without reference conditions, availability becomes unstable. A claim may slide from one locus to another, from one scale to another, from possible relation to actual registration, from interaction to warrant, or from provisional closure to final possession. Relational availability prevents that slide. It requires the availability status to be indexed. A relation may be open or unindexed where no locus–relation–boundary pair has yet been specified. The claim remains too indeterminate for horizon analysis. A statement such as “the unknown,” “the real,” “information,” “the system,” “the observer,” or “the boundary” may sound substantial, but without a specified relation and locus it does not yet identify an availability condition. Open availability is not mysterious depth. It is an unindexed claim-state. A relation may be unavailable where it does not enter the specified locus under current access conditions. The failure is not necessarily a failure of reality, strength, or importance. It is an access condition. A molecular event may be real but unavailable to the unaided human eye. A future astronomical signal may be real but unavailable to current instruments. A legal harm may be lived but unavailable to a procedure whose categories cannot receive it. A formal relation may be structurally implied but unavailable to a mathematical community before an abstraction has made it formulable. A relation may be available in principle where it is indexable in kind for some suitable locus, channel, or regime. This is not yet actual availability for the locus under review. It means that the relation belongs to a kind that could be made available under some appropriate condition. A photon of a detectable wavelength may be available in principle to a suitable light-sensitive channel even if it has no path relation to the present observer. A formal relation may be expressible in some richer language even if it is not expressible in the current one. A type of evidence may be receivable by some institutional procedure even if the present procedure cannot admit it. A relation may be available in potential where it is available in principle and shares possible branch-space with the specified locus or coupled-locus system. The relation has not necessarily entered actual access, but there is a physically, temporally, technically, biologically, institutionally, mathematically, or operationally specifiable path by which it could become available. An object behind an observer may be visually available in potential because reorientation could bring it into view. A signal may be available in potential because a future detector path could receive it. A disputed institutional relation may be available in potential because a revised evidential procedure could admit the relevant testimony. A relation may be coupled where it has entered interaction with the specified locus or coupled-locus system. Coupling is contact. It is not yet downstream status. A relation may perturb a system without being registered as contrast. It may interact physically without becoming signal-bearing. It may affect a body without becoming recognised. It may enter a machine process without becoming warranted classification. Coupling marks realised interaction, but the downstream status must still be earned. A relation may become available as contrast where it reaches an interface in a way that can be distinguished under the relevant conditions. At this stage, the relation is no longer merely possible or coupled. It has become available for discrimination. But available contrast is still not closure. It may be unclassed. It may be noisy. It may be ambiguous. It may be available for one abstraction and unavailable for another. It may support registration without supporting explanation. A relation may be registered when one available, classed alternative is selected into state; retained when that distinction persists as accessible state over the relevant horizon; or deployed when later use depends on it. Registration selects. Retention preserves. Deployment uses. None of these statuses automatically establishes truth, warrant, proof, explanation, or ontology. A record is not warrant by default. A retained category is not target possession. A deployed model is not closure unless the closure demand has been satisfied under stated conditions. A relation may be closed where the relevant closure demand is satisfied. Closure is always indexed to demand. Recognition closure is not proof closure. Measurement closure is not ontological closure. Legal closure is not explanatory closure. Action closure is not epistemic closure. A claim may close for one purpose while remaining residual for another. A model may be closed enough for prediction and residual-bearing for explanation. A theorem may be closed within a formal system while its interpretation remains open. A decision may close procedure while leaving moral, causal, or evidential residual. A relation may remain residual where some discriminator condition remains unmet, some comparator has not been survived, some scale transition remains unhandled, some live formulation has not been excluded, or some closure demand exceeds the present resources of the locus and regime. Residual is not a rhetorical marker of uncertainty. It is the localised remainder that prevents stronger closure. It should state what is missing, what cannot be claimed because it is missing, and what would force update. These conditions are not a staircase. Finite systems often encounter them out of order, in partial form, or distributed across coupled loci. A telescope may make relation available as contrast while the human observer depends on instrument calibration, image processing, institutional archive, and scientific model to retain and interpret it. A court may register testimony while the lived relation remains poorly captured by legal category. A machine classifier may deploy retained distinctions without warranting the target relation. A mathematical community may manipulate a formal expression before its boundary-status is understood. A scientific practice may act on a model while ontology remains residual. Condition discipline prevents several errors. It prevents treating existence as availability. It prevents treating availability in principle as availability for this locus. It prevents treating potential availability as actual interaction. It prevents treating coupling as registration. It prevents treating registration as warrant. It prevents treating deployment as closure. It prevents treating Soft Horizon closure as total closure. Relational availability therefore locates the unknown without beginning from abstract ontology. It does not deny ontology. It refuses to let ontology substitute for the access and closure conditions under which finite systems make claims. Reality may constrain a locus without being fully available to it. A target relation may correct a model without being possessed as total state. The relation is indexed. The constraint is real. This is why observer-relativity does not mean subjectivism. Availability is relative to locus and regime, but it is not invented by preference. The target constrains what can be preserved, what can be measured, what can be classified, what can be proven, what can be acted on, and what will fail under correction. The observer-system constricts what can enter, be classed, be registered, be retained, be deployed, and be closed. Horizon analysis belongs to the relation between constraint and constriction. The same relation may have different availability statuses across loci. A radio signal may be unavailable to unaided human hearing, available in principle to a suitable receiver, available in potential to an observer with access to such a receiver, coupled to an antenna, available as contrast after filtering, registered as data, retained as record, deployed into a model, and closed only under a specific interpretive demand. At no point should one status be silently substituted for another. The same is true in formal systems. A mathematical relation may be latent in a formal regime before it is expressible. It may be available in principle in a stronger language. It may become available in potential through a line of proof development. It may be coupled to a formal community through anomaly or paradox. It may become formulated. It may be registered as theorem, conjecture, contradiction, or undecidable statement. It may close under one system and remain unclosed under another. The horizon depends on the relation among formal resources, expression, proof demand, and system boundary. Relational Reachability, developed in Part 2a, Paper 1c, sharpens a distinction within this grammar. Physical ingress can make a target available as contrast and support registration under one abstraction while a higher-order relation remains unformable for the locus. The limitation then concerns neither absence of incoming contrast nor closure of an already formulated relation, but the retained organisation and available scaffolding through which a target-supported relation could become stabilisable. Repeated exposure alone need not supply the missing organisation; a physically supported sequence of intermediate distinctions may do so. A relation may therefore be immediately reachable, constructibly reachable, or unreachable under the present regime without those conditions defining new RRP stages or horizon types. Reachability failure is not Gödel disclosure: the latter requires a relation to become available or formulable strongly enough for the regime to disclose its own closure limit. Relational availability therefore prepares the concept of horizon. A horizon is not whatever the observer does not know. It is the boundary condition produced by the relation among availability, locus, abstraction, and closure demand. Some horizons concern access. Some concern recognition. Some concern procedure. Some concern proof. Some concern action. Some concern residual-bearing provisional closure. Some concern internal abstraction-boundary disclosure.1 The discipline begins by refusing to collapse them. 4. Recognition, Soft Horizons, and misclosure Recognition is not passive receipt of identity. Recognition is a closure operation. It takes available relation and stabilises it under a pattern. The observer-system does not merely receive “tree,” “face,” “danger,” “number,” “cause,” “symptom,” “wave,” “particle,” “legal person,” “risk,” “evidence,” or “proof.” It closes available relation into one of these forms by applying retained distinctions under an abstraction regime. This occurs at many levels. Biological recognition may close chemical relation into food, toxin, mate, threat, injury, or self/non-self. Sensory recognition may close visual relation into object, surface, motion, depth, colour, face, path, or obstruction. Linguistic recognition may close sound or inscription into word, sentence, command, question, promise, or insult. Scientific recognition may close measurement into signal, noise, anomaly, confirmation, variable, effect, or mechanism. Mathematical recognition may close symbolic relation into theorem, contradiction, proof step, definition, equivalence, or undecidable statement. Legal recognition may close lived event into evidence, act, intent, harm, responsibility, liability, standing, or remedy. Each recognition event reduces open availability. It converts relation into usable form. Without such closure, the observer-system would remain flooded by unorganised relation. Nothing could be compared, remembered, communicated, corrected, acted upon, or redeployed. Recognition is therefore necessary. But recognition is not total possession. A pattern preserves some relation and omits the rest. It admits selected features, stabilises selected equivalences, and treats selected differences as background. It may be adequate for action while incomplete for explanation. It may preserve one scale while distorting another. It may be useful in one regime and misleading in another. It may close enough for local response while failing under comparator pressure. Recognition is therefore enabling and dangerous at once. It makes relation usable. It can also make usability feel like possession. A recognised object often appears already closed. A face is seen as a face. A sentence is heard as a sentence. A symptom is treated as a diagnosis. A category is treated as if it has captured the event. A mathematical expression is treated as a known form. Once closure occurs, its operation often disappears from view. The observer-system experiences the result, not the cut that produced it. This invisibility is the root of many errors. The observer-system may forget that recognition is indexed to a locus, abstraction, scale, horizon, comparator field, and closure demand. It may forget that the pattern was sufficient for one purpose but not another. It may forget that residual remained. It may forget that alternative closures were excluded, ignored, unavailable, or not yet formulable. It may then treat recognition as if it were possession. A Soft Horizon is the disciplined name for legitimate provisional closure under finite conditions. It is a purpose-indexed closure that makes relation usable while stronger closure remains unavailable, unnecessary, or not yet earned. A Soft Horizon does not claim total possession. It claims sufficient closure for a specified use.2 The structure of a Soft Horizon is therefore conditional. This relation is closed enough for this locus, under this availability condition, through this abstraction, at this scale, over this horizon, for this closure demand, with this residual, within this tolerance, under this update condition. Remove those conditions, and the Soft Horizon is no longer disciplined. It becomes a floating closure. Soft Horizons are everywhere because finite systems cannot operate without them. Perception closes enough for navigation before it closes explanation. Language closes enough for communication before it closes ontology. Scientific models close enough for prediction before they close all mechanism. Legal categories close enough for decision before they exhaust lived relation. Institutional procedures close enough for action before they exhaust moral, causal, or evidential complexity. Machine systems classify under training, architecture, threshold, and deployment constraints before target preservation is guaranteed. Mathematical practice works inside formal systems even when meta-formal limits remain. The Soft Horizon is not a concession to weakness. It is often the strongest possible finite output. A relation closed enough for navigation, measurement, proof within a system, local prediction, procedural decision, or emergency action may be legitimate and powerful. The error is not using Soft Horizons. The error is letting them harden into unindexed closure. Misclosure occurs when a Soft Horizon claims more closure than its conditions support. The available relation is forced into a pattern that does not preserve the target relation under the closure demand being made. An ambiguous movement becomes a threat without sufficient comparator. A symptom becomes a diagnosis without adequate exclusion. A measurement anomaly becomes error before the instrument path is audited. A legal category absorbs a harm it cannot preserve. A machine label is treated as target truth because it is confident, consistent, or institutionally useful. A mathematical formulation is treated as resolved because it is expressible. Optical illusion gives the primitive demonstration. The illusion is not that nothing was seen. Something was seen. Relation became available. The failure lies in closure. The perceptual system produced a pattern that appeared stable until correction exposed its residual. Misclosure is therefore not non-observation. It is overstrong closure over available relation. Misclosure differs from simple uncertainty. In uncertainty, the observer-system may know that it does not know. In misclosure, the observer-system thinks closure has occurred. The relation has been given a form. That form then travels as if it were warranted. Misclosure is dangerous because it reduces correction pressure. A false closure is often more operationally powerful than open residual. It gives the system a usable answer, a record, a category, an action path, or a decision. Its very usability can hide the failure that produced it. Reification is the downstream hardening of Soft Horizon into possession. The provisional closure loses its residual, scope, tolerance, comparator, and update condition. The object is no longer treated as recognised under conditions; it is treated as simply given. The category is no longer treated as an abstraction cut; it is treated as the target. The model is no longer treated as approximation; it is treated as ontology. The record is no longer treated as bounded registration; it is treated as warrant. The procedure is no longer treated as a closure path; it is treated as legitimacy. The machine output is no longer treated as deployment under constraint; it is treated as decision-quality itself. Reification is especially difficult to detect because it often begins with a legitimate closure. The original Soft Horizon may have been necessary and well-grounded for its purpose. The problem arises when the closure travels beyond that purpose. A medical category useful for triage may become personal identity. A legal classification sufficient for temporary protection may become permanent judgement. A scientific model strong within a scale may become total explanation. A machine score useful for screening may become final decision. A formal proof within one system may be narrated as closure under a different demand. The discipline is not to abolish recognition. A system that refused recognition until total closure was available would not become more rational. It would become non-functional. It could not move, classify, speak, model, decide, prove, or act. The discipline is to keep recognition indexed. What relation was closed? For which locus? Under which availability condition? Through what abstraction? At what scale and horizon? For what closure demand? With what residual? Under what update condition? Once those questions remain visible, the Soft Horizon remains soft. It can support use without becoming possession. It can carry action without becoming proof. It can carry classification without becoming ontology. It can carry record without becoming warrant. It can carry model without becoming target. It can carry decision without pretending to exhaust relation. Recognition therefore stands at the first operational boundary of horizon analysis. Availability enters. Pattern attempts closure. Soft Horizon makes use possible. Misclosure shows where the pattern overclaims. Reification shows how provisional closure becomes unearned authority. The horizon appears wherever this process reaches a boundary between relation available and relation closed. The next step is to name that boundary more formally. 5. Horizons as availability/closure boundary conditions A horizon is a boundary in relational availability and closure for a specified locus. This definition must be kept precise. A horizon is not an object. It is not a place. It is not an ontological zone. It is not a mysterious edge of reality. It is not whatever a system happens not to know. It is the boundary condition produced when a specified relation reaches, fails to reach, crosses, stalls within, or overclaims a condition of availability or closure for a specified locus under a specified demand. The horizon therefore has two sides. The first is the availability side. This concerns whether and how the relation enters access for the locus. It may be unavailable, available in principle, available in potential, coupled, available as contrast, registered, retained, or deployed. The second is the closure side. This concerns what the system is entitled to claim once the relation has entered some availability condition. The relation may remain unrecognised, provisionally closed, misclosed, residual-bearing, closed for a limited purpose, or disclosed as uncloseable within the present abstraction regime. The horizon appears where these sides fail to coincide. A relation may be available but not recognised. It may be recognised but not warranted. It may be registered but not retained. It may be retained but not deployable. It may be deployable but not closed. It may be closed for action but not explanation. It may be formulable but not provable. It may be describable as a possible test but procedurally unavailable. It may be latent in an abstraction regime before the regime can expose its own limit. Each case marks a boundary, but not the same boundary. This distinction prevents horizon inflation. If every limit is called a horizon in the same sense, the concept loses force. A blind spot, an inaccessible object, a missing measurement, an unrecognised pattern, an institutional category failure, an undecidable formal statement, and a residual-bearing action decision are not the same condition. They share a general structure only because each involves a relation, a locus, an availability condition, and a closure demand. The horizon must be typed by the status of that relation. An access-side horizon appears where the relation cannot enter the locus under current conditions. The relation may be real, but it has not become available. It may lie outside the relevant boundary, beyond the interface, below the threshold, outside the scale, beyond the operational horizon, or outside the discrimination regime. The correct output is not explanation or closure. It is unavailable relation, or relation available only in principle or potential. A recognition horizon appears where relation enters availability but has not been stabilised as identity. Something is present, but not yet known as. The system may sense anomaly, pressure, contrast, disturbance, or possibility without yet knowing what kind of relation it has encountered. The correct output is available-but-unrecognised relation, not absence and not closure. A registration horizon appears where relation is coupled or available as contrast but no classed alternative is selected into state. The relation may affect the locus without producing registration. The correct output is coupling or available contrast without registration; any claim of record, memory, or carried-forward state requires the further retention condition. A closure horizon appears where relation has been registered, retained, or formulated, but the closure demand exceeds what the current path can support. The demand may be recognition, measurement, proof, explanation, classification, legal finding, institutional decision, machine output, action, or warrant. The failure is not simply that something is unknown. The failure is indexed: this relation, for this locus, under this abstraction, cannot yet satisfy this demand. A residual horizon appears where a claim has reached the limit of its present warrant while preserving the missing condition. Residual is not a decorative acknowledgement of uncertainty. It is the retained remainder that prevents stronger closure. A residual horizon is disciplined only when it states what remains unresolved, what cannot be claimed because it remains unresolved, and what would force update. A Soft Horizon appears where provisional closure is legitimate for a specified purpose while residual remains. This is not a failed horizon. It is a working horizon. It permits the system to move, speak, classify, model, decide, or act without claiming total possession. The danger is reification: the Soft Horizon travels downstream without its scope, tolerance, residual, comparator, or update condition. A Gödel Horizon is more specific. It appears where a relation becomes available or formulable within an abstraction regime and that regime discloses a limit of closure under its own present resources. This is not mere absence, not procedural inaccessibility, not misrecognition, and not ordinary residual. It is a boundary disclosed from within the regime’s own structure once the relevant relation and closure demand become available. A horizon is therefore not merely the limit of an observer. It is the boundary produced by the relation among observer-system, target relation, abstraction regime, path condition, and closure demand. If any of these is missing, the horizon has not been analysed. This is why the governing rule must remain strict: No horizon can be analysed without specifying the relation, the locus, the availability condition, and the closure demand. Without relation, the horizon becomes empty metaphysics. Without locus, it becomes false absoluteness. Without availability condition, it loses access discipline. Without closure demand, every failure becomes vague. A horizon is intelligible only when the system can say what has become available, what has not, what kind of closure is being sought, and what prevents or limits that closure. This also explains why horizons can move without disappearing. A new instrument may turn unavailable relation into available contrast. A new category may turn unrecognised relation into classed relation. A new record path may turn transient interaction into retained evidence. A new proof technique may close what had been residual. A new abstraction may formulate what had been latent. But the movement of a horizon is not the abolition of horizon. It is the replacement of one availability/closure boundary with another. A finite system does not escape horizons by learning. It relocates them. 6. Loci, coupled loci, and procedural availability A horizon is always relative to a locus. The locus is the unit for which availability, registration, retention, deployment, closure, and residual are being evaluated. It may be a person, organism, cell, instrument, legal institution, scientific community, mathematical practice, computational system, machine-mediated process, administrative body, cultural formation, or coupled observer-network. It is not always an individual human observer. Human knowing is one important case, but it is not the root form. Locus discipline matters because availability is easily misassigned. A relation available to one locus may be unavailable to another. A signal available to an instrument may be unavailable to unaided perception. A harm available to lived experience may be unavailable to legal procedure. A pattern available to a machine model may be unavailable to human inspection. A proof relation available to a stronger formal system may be unavailable to a weaker one. A measurement available to a scientific community may be unavailable to the individual who relies on its published result. If the locus shifts silently, the claim changes without being marked. The first failure is locus inflation. This occurs when availability for one locus is treated as availability for all. The instrument registers; therefore the person is said to know. The institution records; therefore the affected relation is said to be recognised. The model classifies; therefore the target is said to be understood. The formal system expresses; therefore closure is said to be available. Each move exports status from one locus to another without showing the path by which the relation travels. The second failure is locus contraction. This occurs when a coupled system is reduced to one visible agent. A scientist is treated as if she alone carries the observation, ignoring instrument, calibration, laboratory practice, statistical method, archive, peer correction, and theoretical background. A judge is treated as if he alone carries the legal finding, ignoring procedure, evidence rules, institutional history, precedent, statutory structure, and record path. A machine output is treated as if it belongs only to the model, ignoring data provenance, labelling practice, architecture, threshold, interface, deployment context, and user interpretation. The third failure is locus drift. This occurs when the explanatory work moves between source, carrier, receiver, institution, model, user, affected party, and analyst without being stated. A relation may be available to the analyst but not to the system being analysed. It may be available to the institution but not to the person subject to its decision. It may be available to a model during training but not in deployment. It may be available to a technical team but not to a governance process. The locus must be redrawn whenever the evaluated relation changes. Many modern horizons are coupled horizons. They arise not at the boundary of an isolated observer but inside observer-networks. Laboratories, courts, hospitals, archives, platforms, markets, administrations, sciences, cultures, and formal communities all extend availability by coupling many loci together. These systems can make relations available that no individual locus could access alone. They also multiply Soft Horizons. A coupled locus is not merely a collection of observers. It is a relation-carrying arrangement through which access, transformation, registration, retention, deployment, and closure may be distributed. In a scientific laboratory, an instrument may receive relation, software may process it, a database may retain it, a theory may classify it, a community may compare it, and a publication may redeploy it. The claim belongs to the coupled path, not to any single element by default.3 This is powerful. Coupled loci expand reach. They allow finite systems to see farther, measure finer, remember longer, compare more widely, formalise more rigorously, and act at greater scale. But coupling also increases the risk of residual loss. A relation may pass through many transformations before it becomes a claim. At each transformation, something may be filtered, compressed, renamed, normalised, excluded, mis-scaled, or overclosed. The final claim may appear stable precisely because the earlier residuals have disappeared from view. Coupling marks contact. It does not settle the status earned by that contact. A relation that enters an institutional process has not necessarily been recognised adequately. A relation that enters a machine system has not necessarily been preserved. A relation that enters a formal community has not necessarily been closed. A relation that enters a scientific instrument has not necessarily been explained. The downstream status must still be earned. The stronger the coupled locus, the more important its horizon discipline becomes. An unaided observer may misrecognise a shadow. A coupled institution may misrecognise a population. A machine-mediated system may scale a misclosure across millions of cases. A formal community may stabilise an abstraction so powerfully that its boundaries become visible only through internal proof. Coupling increases reach. It also increases the need to preserve the path by which relation travels. This leads to a distinct horizon type: the Procedural Availability Horizon. A Procedural Availability Horizon occurs where an abstraction can state what would count, but the process, data path, evidence path, instrument, material condition, temporal access, or realised branch required to bring the relation into registration, comparison, demonstration, or closure is not available to the specified locus.4 This horizon is availability-side, but it is not simple non-availability. The relation is not wholly outside thought. The observer-system can describe what would be needed. It can state the test, evidence, branch, instrument, proof path, material condition, observation, or procedure that would make the relation available. The failure lies not in conceivability but in access to the required path. The distinction is important because describability often masquerades as availability. A system can state what would count as evidence and then act as if the evidence relation is already available. It can describe a possible experiment and then speak as if the experiment has constrained the claim. It can imagine an institutional process and then treat the process as if it had registered the affected relation. It can define a test and then speak as if the test has been run. It can specify a proof demand and then behave as if the proof path exists. This is procedural overclaim. A Procedural Availability Horizon blocks that overclaim. It says: the abstraction can describe the condition, but the condition is not presently available through a realised path. The simplest scientific case is an experiment that can be specified but not performed. A theory may state what measurement would decide between alternatives. The measurement may be coherent. The required relation may be available in principle. It may even be available in potential under some future technology. But if the instrument, energy scale, temporal access, sample condition, isolation regime, or observational pathway does not exist for the current locus, the relation has not entered registration. The correct output is not closure and not Gödel diagnosis. It is describable but procedurally unavailable. A historical case has the same structure. A claim may require evidence from an event for which no relevant record, material evidence, or witness account was preserved. The evidential demand can be stated. The missing archive, lost material, absent witness, destroyed context, or unrecorded condition can be identified. But if the path by which the relation could enter present comparison is unavailable, the claim remains procedurally residual. The system may know what would count without possessing a route by which what would count can be made to count. Institutional cases show the same danger. A legal or administrative system may know that a certain lived relation would matter if it could be admitted, classified, evidenced, or procedurally recognised. But if the category, evidential path, record form, standing rule, or institutional interface cannot receive it, the relation remains procedurally unavailable to that system. The failure is not necessarily that the relation is unreal. It is that the process cannot make the relation count. Machine systems intensify the problem. A model may be evaluated by metrics that state what would count as performance while lacking the data path required to test the target relation. A benchmark may describe a task but fail to preserve the deployment relation. A monitoring system may specify a drift condition but lack the instrumentation to detect the relevant drift. The condition is describable. The path is absent. This distinction is central to the Gödel architecture. A Procedural Availability Horizon is not a Gödel Horizon proper. In a Gödel Horizon, the relation has become available or formulable within the abstraction regime, the closure demand is well-posed inside that regime, and the regime discloses that it cannot close the relation from within its own resources. In a Procedural Availability Horizon, the closure condition can be described, but the path required to bring the relation into registration, comparison, demonstration, or closure is not available to the locus. Procedural availability also protects historical analysis. Once a later branch exists, it is tempting to read it backward as if earlier observers already had the relation in view. But earlier abstraction may have had only principle-level compatibility, not potential availability, not coupled interaction, not registration, and not closure. Retrospective analysis must preserve the difference between what was structurally possible, what was procedurally reachable, what was historically available, and what was actually disclosed. A Procedural Availability Horizon can move. A new instrument may make the measurement possible. A newly accessible archive may supply a relevant record. A new legal category may admit evidence. A new data path may capture the target relation. A new formal language may make a question expressible. A new institution may create the procedure required for registration. When this happens, the relation may move from procedural unavailability into available contrast, registration, comparison, or formal boundary-disclosure. But the movement of the horizon is not its abolition. It is the replacement of one availability/closure boundary with another. 7. Closure, abstraction regimes, and latent boundaries Closure is not a single endpoint. A relation may be closed for recognition, but not explanation. Closed for measurement, but not ontology. Closed for procedure, but not justice. Closed for action, but not certainty. Closed within one formal system, but not within another. Closed under a threshold, but not under a stronger demand. Closure must therefore be treated as claim-path status, not as possession of the target relation as total state. A claim path is the route by which relation becomes available, is classed, registered, retained, formulated, compared, provisionally closed, residualised, acted upon, or updated. It may be short, as in a perceptual recognition. It may be long, as in a scientific claim that depends on instrumentation, calibration, theory, statistical method, publication, replication, and interpretation. It may be distributed, as in institutional decisions, legal findings, machine outputs, or formal mathematical developments. What matters is that claim-strength must not exceed the closure earned along that path. Closure demand specifies what the claim is asking the relation to satisfy. Recognition demand asks whether this is identifiable as that. Measurement demand asks whether the value has been established within tolerance. Proof demand asks whether the proposition follows under a formal regime. Explanation demand asks whether the mechanism or relation has been preserved. Classification demand asks whether the case belongs inside the category. Legal demand asks whether the procedural or evidential burden has been met. Action demand asks whether enough has been closed to justify intervention. These demands must not be substituted for one another. A closure may be strong without being total. A measurement may be closed within tolerance. A theorem may be closed within a formal system. A legal decision may close a procedural question. A model may close prediction over an operating range. A perception may close navigation. These closures are legitimate when their demand, scope, scale, horizon, comparator, residual, and update condition remain attached. They become dangerous when they are narrated as unbounded possession. Closure failure is likewise demand-relative. A claim may fail under proof demand while remaining usable for conjecture. It may fail under explanation demand while remaining predictive. It may fail under legal demand while remaining morally significant. It may fail under institutional category while remaining socially real. It may fail under machine classification while remaining available to human judgement. Without closure-demand discipline, success and failure become vague. Closure is made possible by abstraction regimes. An abstraction regime is the pattern architecture through which relation becomes classable, retainable, communicable, comparable, usable, correctable, and sometimes closeable. It supplies the cuts by which the observer-system treats this as the same as that, different from that, within scope, outside scope, signal, noise, object, event, number, category, proof step, record, harm, risk, or decision. Without abstraction, availability remains insufficiently organised for closure. Abstraction is not optional. A finite system cannot receive reality as total state. It must select. It must ignore some differences and preserve others. It must compress, stabilise, and redeploy relation. Every perceptual category, word, measurement, model, institution, legal test, scientific theory, mathematical formalism, and machine classifier depends on abstraction. To reject abstraction would be to reject finite cognition, communication, science, law, mathematics, governance, and action. But abstraction is also a closure machine. It does not merely describe relation. It shapes what can become available as relation for the system. It determines what counts as same or different, relevant or irrelevant, admissible or inadmissible, measurable or immeasurable, signal or noise, proof or non-proof, evidence or background, harm or non-harm, valid or invalid. It opens some paths and closes others. It makes some relations visible and others residual. This gives abstraction regimes their double character. They enable closure by making relation usable. They also produce boundaries because their cuts cannot preserve every relation under every demand. A regime may be powerful precisely because it filters. But what it filters may later become the relation that matters. A perceptual abstraction regime stabilises object, surface, depth, movement, and figure. It enables navigation, but it can misclose illusion. A scientific abstraction regime stabilises variables, measurements, models, and explanatory forms. It enables prediction and explanation, but it can hide scale transition or anomalous relation. A legal abstraction regime stabilises standing, evidence, harm, liability, procedure, and remedy. It enables decision, but it can fail relations that do not enter its categories. A machine abstraction regime stabilises labels, features, embeddings, thresholds, and outputs. It enables scalable classification, but it can mispreserve target relation under deployment shift. A formal mathematical regime stabilises symbols, axioms, inference rules, and proof conditions. It enables exact closure, but it may disclose statements it cannot prove within itself. The abstraction regime is therefore not merely an instrument applied after relation is available. It partly determines how relation can become available at all. It forms the channel through which availability is classed into closure. This is why no claim can be audited without exposing its abstraction. If the abstraction is hidden, the closure appears natural. The system forgets that a cut has been made. An abstraction boundary appears where the regime’s cuts fail to preserve the target relation under the closure demand being made. Some abstraction boundaries are visible. A model may state its operating limits. A legal category may specify exclusions. A measurement instrument may state tolerance. A formal system may state its axioms and inference rules. In such cases, the boundary is part of the regime’s explicit discipline. Other boundaries are residual. The regime knows that some relation remains unclosed, but it cannot yet localise the boundary completely. A scientific model may carry unresolved mechanism. A legal category may carry unresolved affected-locus relation. A machine classifier may carry deployment uncertainty. A formal system may carry untested closure ambition. Residual boundaries are known enough to be stated, but not closed enough to be resolved. A latent abstraction boundary is different. It is a closure limit structurally implied by an abstraction regime before that limit has become available, formulable, demonstrable, or disclosed to the locus. The regime may operate successfully across many cases while containing a boundary it has not yet encountered. The boundary is not visible as failure because the relevant pressure has not yet appeared. A legal category may work across standard cases while failing at an edge. A scientific model may predict within a domain while carrying a scale boundary. A mathematical programme may produce proofs while assuming a closure ambition it cannot satisfy. A machine classifier may perform well on benchmark data while carrying deployment failure. The latent boundary is not visible as long as the regime is not pressed at the relevant point. This is why stabilised abstraction can become overconfident. Success hides boundaries until the right pressure reveals them. The regime’s strength within its domain may make its latent limits harder to see. A familiar closure becomes background. Its residual appears irrelevant because ordinary use does not expose it. When the boundary is finally disclosed, the system may experience the disclosure as crisis, paradox, anomaly, contradiction, exception, failure, or impossibility. The better interpretation is often more disciplined: the regime has encountered the edge of its own closure power. A latent abstraction boundary is not a weakness by itself. It is a finite-regime condition. Every abstraction regime selects, compresses, and closes under limits. Some limits are known. Some are residual. Some are latent. The discipline is to avoid pretending that undisclosed boundaries do not exist, while also avoiding the claim that they are already known. Latency must not become mystery. It must remain indexed: this regime may contain a closure limit not yet available, formulable, demonstrable, or internally disclosed to the specified locus. The sequence matters. A relation may be latent in a regime. It may later become describable as a possible test. It may later become procedurally available. It may later become formulated. It may then disclose closure failure. Each stage has a different availability and closure status. If they are collapsed, the analysis will either overclaim early disclosure or understate the force of later boundary revelation. This prepares the refined definition of the Gödel Horizon. The Gödel Horizon is not every limit, every unknown, every procedural gap, every residual, or every latent boundary. It is the disclosure of a latent abstraction boundary as a present closure limit of the regime under a specified demand. 8. The Gödel Horizon defined A Gödel Horizon occurs where a latent abstraction boundary becomes available or formulable within an abstraction regime and is disclosed as a limit of closure under that regime’s own present formal, inferential, descriptive, or classificatory structure. This definition keeps the name Gödel because the formal mathematical case remains the cleanest exemplar. But the definition is relational. The horizon is not a property of the target alone, not a property of the observer alone, not a property of language alone, and not a general name for mystery. It appears at the boundary among four conditions: locus, abstraction regime, target relation, and closure demand.5 The locus is the observer-system or coupled system for which the relation becomes available. It may be a person, organism, instrumented practice, legal institution, mathematical community, computational system, scientific field, or coupled observer-network. The abstraction regime is the pattern architecture through which relation is made available and closeable. It may be perceptual, linguistic, formal, mathematical, scientific, legal, institutional, computational, cultural, or procedural. The target relation is the relation the system attempts to preserve, identify, prove, explain, classify, decide, or act upon. It constrains the claim. The observer-system does not invent the target by naming it. The closure demand is the kind of closure being sought. It may be recognition, proof, explanation, measurement, classification, warrant, decision, or action. A Gödel Horizon cannot be diagnosed unless the demand is well-posed. A vague sense of difficulty is not enough. The Gödel Horizon appears only when the relation has entered the regime strongly enough to expose the regime’s own limit. This is the crucial point. If the relation is simply unavailable, the condition is not Gödel-type. If the system can only state what would count but cannot access the path, the condition is procedural. If the system has applied a wrong pattern, the condition is misclosure. If the system carries unresolved remainder without internal boundary disclosure, the condition is residual. If the relation is structurally implied but not yet available or formulable, the condition is latent. The Gödel Horizon requires disclosure. Disclosure means that the regime is confronted with a relation it can formulate, register, or otherwise make available under its own terms, while the closure demand cannot be satisfied by those same terms. The boundary is not imported from outside as a mere objection. It appears from within the regime’s own resources. The regime can bring the relation into view, but cannot close it. This separates the Gödel Horizon from ordinary ignorance. In ignorance, the system lacks the relation, the information, the method, the evidence, or the proof. In a Gödel Horizon, the relation has become available in a way that shows something stronger: the present regime cannot close what it has made available under the relevant demand. It also separates the Gödel Horizon from simple error. An error may be corrected within the regime. A miscalculation, mismeasurement, mistranscription, faulty observation, or mistaken classification may fail without disclosing a boundary of the regime. Gödel-type failure is not merely that the system produced the wrong answer. It is that the regime’s own closure resources are insufficient for the relation now in view. It also separates the Gödel Horizon from paradox as spectacle. A paradox may indicate confusion, ambiguity, self-reference, category error, abstraction failure, or genuine boundary disclosure. It becomes Gödel-type only where the paradox or limit reveals that the regime can formulate the relation but cannot close it under its own resources. Not every paradox is a Gödel Horizon. Some are bad formulations. Some are linguistic traps. Some are resolved by sharper abstraction. Some disclose a deeper boundary. It also separates the Gödel Horizon from the Observer Event Horizon. An Observer Event Horizon concerns the access boundary beyond which relation cannot enter the observer-system under the current regime. The Gödel Horizon concerns the closure boundary that appears after relation has entered availability or formulation. One is primarily about ingress into access. The other is about closure failure within a regime that has already made the relation available in the relevant sense. It also separates the Gödel Horizon from the Procedural Availability Horizon. In a Procedural Availability Horizon, the system can state what would count but lacks the path required to bring the relation into registration, comparison, demonstration, or closure. In a Gödel Horizon, the relation has crossed that procedural threshold far enough for the regime to disclose its own closure limit. This definition also prevents overextension of Gödel language. It is tempting to call any hard problem, mystery, anomaly, impossibility, open question, or institutional failure a Gödel Horizon. That inflation should be resisted. The concept is useful only where the structure is disciplined. The relation must be specified. The locus must be specified. The abstraction regime must be specified. The closure demand must be specified. The boundary must be disclosed from within the regime’s own relation to the demand. The Gödel Horizon is therefore not a metaphor for difficulty. It is a name for a precise finite-observer condition: availability or formulation has outrun closure within the current abstraction regime. The mathematical case will show this most exactly. A formal system may be strong enough to formulate a statement whose proof-status cannot be closed within that same system. The statement is not outside the regime in the ordinary sense. It is expressible within it. The closure demand is not subjective. It is proof. The limit is not historical inconvenience. It is structurally disclosed. This is why Gödel gives the horizon its name. But the architecture is wider because finite systems repeatedly encounter the same structure outside mathematics. Perception may make relation present before identity closure. Science may register anomaly before explanatory closure. Institutions may receive relation that their categories cannot preserve. Machines may produce outputs under abstractions whose target relation they cannot warrant. Legal systems may close procedure while leaving affected relation residual. These cases are not identical to Gödel’s theorem. They are structurally related only where the same availability/closure boundary is disciplined. The Gödel Horizon names the boundary where a regime becomes answerable to what it can make available but cannot close.6 That boundary is not the end of inquiry. It is the point at which inquiry must become more honest about its conditions. The system may need a stronger abstraction, a different locus, a revised formal regime, a changed procedure, a new comparator, a residual statement, or an explicit Soft Horizon. What it must not do is treat availability as closure merely because the relation has entered view. The Gödel Horizon is where the regime learns that seeing, stating, formulating, or registering is not the same as closing. 9. Gödel’s theorem as formal exemplar Gödel’s incompleteness theorems supply the formal exemplar of the Gödel Horizon because they separate formulation from proof-closure with unusual precision.7 The point is not that all horizons are mathematical. Nor is it that every difficult problem should be described as Gödelian. The point is narrower. In the formal case, a system of sufficient strength can formulate a statement under its own resources while proof-closure cannot be supplied by those same resources. The relation is not absent. It is expressible. The closure demand is not vague. It is proof within the system. The failure is not historical inconvenience, psychological limitation, poor technique, or missing effort. The boundary is disclosed by the structure of the regime itself. This is why Gödel gives the horizon its name. A formal system creates a highly disciplined abstraction regime. It specifies symbols, formation rules, axioms, inference rules, proof conditions, and admissible transformations. It determines what can count as expression and what can count as proof. Inside such a regime, closure is exact in a way that ordinary perceptual, legal, scientific, or institutional closure rarely is. A proof either follows under the rules or it does not. This exactness makes the boundary visible. The first lesson is that expression is not closure. A statement may belong to the system’s language without being provable in that system. It is not outside the regime as noise, inaccessibility, or unformed possibility. It has entered the regime. It is available as formulation. Yet the closure demand that the regime itself makes available cannot be satisfied by the regime’s own proof resources. The second lesson is that internal availability may disclose internal limit. The boundary is not merely imposed by an external critic who says the system has failed. The system’s own structure makes the relevant relation formulable, and the analysis of that structure discloses the proof boundary. The regime becomes answerable to what it can state but cannot close. The third lesson is that stronger formalisation does not abolish horizons by default. Formalisation can sharpen closure, remove ambiguity, define admissible operations, and preserve proof paths. It can make some claims much stronger than ordinary language can. But once a system is strong enough to express certain relations, it may also become strong enough to disclose limits in its own closure power. More discipline can reveal a horizon rather than remove one. This matters for the wider architecture because finite observer-systems often mistake expression, registration, or availability for closure. A system can name a relation and treat the name as possession. It can measure a physical effect or registered output and treat the measurement as explanation. It can classify a case and treat the category as the target. It can generate a machine output and treat the output as warrant. It can state a possible proof condition and treat the proof as available. Gödel’s formal case blocks that movement in its exact domain. Formulation is not proof. The same distinction must then be carried carefully into weaker domains. In perception, a relation may become present without becoming identified. The availability is real, but identity closure remains open. In science, an anomaly may be registered without being explained. The registration is real, but explanatory closure remains open. In law, a harm may enter institutional process without being preserved by the available category. The procedural entry is real, but relation-preserving closure remains open. In machines, an output may be produced without warranting the target relation. The output is real, but target closure remains open. These cases are not Gödel’s theorem. They do not have the same formal force. They do not prove incompleteness in the mathematical sense. They inherit only the methodological separation that the formal case makes exact: a regime may make relation available in a form that exceeds its own closure resources. The difference must be preserved. Gödel’s theorem remains unique in its formal content. It should not be flattened into a general metaphor for confusion, complexity, paradox, mystery, or institutional failure. Its technical force belongs to formal systems of the relevant kind. But its methodological force lies in the clean separation it reveals: a regime may be able to state what it cannot close. That separation is the horizon. The formal exemplar also protects the paper from two opposite errors. The first is overextension. This error treats every uncertainty, paradox, undecided question, difficult problem, or unsolved scientific puzzle as a Gödel Horizon. That weakens the concept. A Gödel Horizon is not ordinary ignorance. It is not mere lack of data. It is not simply hard inquiry. It requires a specified relation, a specified locus, a specified abstraction regime, and a specified closure demand. The boundary must be disclosed through the relation between the regime and the demand. The second error is confinement. This error treats Gödel’s theorem as technically isolated and therefore irrelevant to wider horizon analysis. That also misses the point. The mathematical case is exemplary because it shows in exact form a structure that finite systems encounter more generally: availability and closure can come apart. A system can receive, register, formulate, express, model, classify, or act through relation before it can close the relation under the demand being made. The correct position is therefore disciplined extension. The formal theorem remains formal. The horizon grammar travels only where its conditions travel. This is why the paper does not claim that every domain is mathematical. It claims that finite domains encounter relation through abstraction regimes, and that abstraction regimes can expose boundaries between what they make available and what they can close. Mathematics gives the cleanest case. Other domains give weaker, messier, but still important cases where the same architecture appears through perception, anomaly, paradox, institutional procedure, machine classification, and action under residual. The Gödel Horizon therefore marks the moment when a system learns that making relation available is not the same as possessing it. A finite system becomes more disciplined when it can preserve that difference. A formal system becomes stronger when it distinguishes expression from proof. A scientific system becomes stronger when it distinguishes model from ontology. A legal system becomes stronger when it distinguishes procedure from relation-preservation. A machine-mediated system becomes stronger when it distinguishes output from warrant. A person becomes stronger when they distinguish recognition from possession. The lesson of Gödel’s theorem is not despair. It is closure discipline. 10. Horizon symptoms and extension cases The Gödel Horizon should not be flattened into a general name for difficulty. A difficult problem is not automatically a Gödel Horizon. A confusing perception is not automatically a Gödel Horizon. A paradox is not automatically a Gödel Horizon. A scientific anomaly is not automatically a Gödel Horizon. A machine failure is not automatically a Gödel Horizon. A relation becomes horizon-significant only when its availability condition and closure demand are specified. The same case may involve unavailable relation, available-but-unrecognised relation, procedural unavailability, Soft Horizon closure, misclosure, residual, or internal abstraction-boundary disclosure. The diagnostic work lies in distinguishing them. The extension cases matter because they show the same architecture under different abstraction regimes. They are not separate foundations. They are horizon symptoms: different ways availability, closure demand, abstraction regime, Soft Horizon, and residual become visible under finite conditions. Perception gives the primitive symptom. Something becomes present before it is identified. A movement in darkness, a sound behind a wall, a shape at the edge of vision, or a bodily sensation before diagnosis does not begin as a named object. It begins as available relation. The observer-system has enough relation to be perturbed but not enough to close identity. The horizon lies between availability and recognition. It asks what has entered access, what pattern is being attempted, and what would close the relation as this rather than that. Optical illusion gives the primitive misclosure symptom. The visual system closes relation under an ordinary pattern, but the closure fails when a stronger demand is applied. The percept is not nothing. It is a real perceptual closure. The error lies in treating that closure as target-preserving under measurement, correction, or altered comparator. Illusion therefore shows that the Soft Horizon of ordinary perception is powerful but bounded. It produces a usable world, not horizonless possession. Scientific anomaly gives a more developed symptom. An experiment may register a relation that the current theory cannot explain. The anomaly is not mere absence. It is registered relation under failed or incomplete closure. The first question is whether the anomaly arises from measurement error, artefact, scale mismatch, weak abstraction, procedural gap, missing variable, comparator weakness, or genuine boundary of the current model. Many anomalies disappear under improved method. Some become evidence for theory revision. Some disclose that a regime has reached the edge of its closure power. The horizon status depends on the claim-path, not on the emotional force of surprise. Paradox gives a conceptual symptom. A relation becomes available in a way that supports competing closures that cannot be held together under the current regime. The observer-system can stabilise the relation one way or another, but not both without revising the abstraction. Some paradoxes are merely verbal confusions. Some are category errors. Some are resolved by better definitions. Some reveal self-reference, circularity, or expressive limits. A paradox becomes horizon-significant only when it shows that availability has outrun the closure architecture of the regime. Infinity gives a scale symptom. Finite systems can formulate infinite structures, sequences, limits, spaces, or operations, but their closure demands vary. Some infinite relations can be handled by formal abstraction. Some can be closed under proof. Some are only approximated. Some generate limits in computation, measurement, representation, or intuition. The point is not that infinity is mystical. The point is that finite access to infinite relation is always mediated by abstraction. The horizon lies in what the abstraction can preserve, what it can prove, what it can compute, what it can approximate, and what remains residual. Recursion gives a self-referential symptom. A regime may be able to apply its operations to outputs, descriptions, or structures of its own operation. This can produce great strength: self-correction, proof theory, meta-language, learning, institutional review, model monitoring, and recursive abstraction. It can also generate closure problems when the system attempts to close its own conditions from within the same level. The horizon appears where the regime’s capacity to re-enter itself creates a relation it can formulate but cannot close without meta-level movement, restriction, or residual. Quantum observation gives an access-and-closure symptom. Measurement does not simply reveal a pre-packaged object to a passive observer. It involves boundary, apparatus, interaction, scale, formalism, probability, and interpretation. The horizon question is not answered by saying that quantum reality is mysterious, nor by saying that measurement simply creates reality. The disciplined question is: what relation becomes available through which experimental arrangement, under which formal abstraction, with what closure demand, and with what residual between model, measurement, and ontology? The horizon lies where formal success, measurement registration, and ontological closure do not coincide. The invention/discovery problem in mathematics gives a formal-practice symptom. Mathematical work feels discovered because formal relations constrain what can be proven once definitions and rules are in place. It feels invented because the choice of abstraction, axioms, notation, problem framing, and proof regime is made within a practice. Horizon analysis does not need to collapse the problem into either side. Mathematical relation is constrained, but access to it is mediated by formal regime. A mathematical community may invent a language in which a relation becomes discoverable. The horizon lies between formal construction, relation constraint, proof closure, and residual interpretation. Cosmological horizons give an access symptom. Some relations may be unavailable because no signal can reach the locus under the physical conditions of the universe, or because the relevant scale, time, or observational path lies outside the available regime. These are not Gödel Horizons by default. They are primarily availability horizons. They become Gödel-relevant only if a theory can formulate the relevant relation while disclosing limits in its own closure resources under the demand being made. Institutional horizons give procedural and classificatory symptoms. Institutions encounter relation through categories, records, rules, thresholds, admissibility conditions, forms, roles, and procedures. A harm may be lived but not admissible. A social relation may be consequential but not recordable. A category may close a case procedurally while failing the relation it purports to govern. The horizon lies in the gap among lived relation, institutional availability, category abstraction, procedural closure, and residual. A legal decision may close a proceeding without closing the underlying relation. Machine-mediated horizons give abstraction and deployment symptoms. A machine system encounters relation through data, labels, tokenisation, architecture, loss functions, thresholds, interfaces, benchmarks, feedback channels, and deployment environments. Its output is not automatically warrant. A model may classify, predict, rank, or generate while preserving only a proxy relation. The horizon lies between input availability, internal abstraction, output production, target preservation, comparator performance, and deployment residual. A machine system does not become horizonless because it is fast, consistent, scalable, or mathematically implemented. These cases show why horizon analysis must remain typed. The same event may contain several horizons. A scientific anomaly may begin as available-but-unrecognised relation, pass through procedural availability questions, be misclosed by a weak model, become residual under comparator pressure, and later disclose an abstraction boundary. A legal case may contain lived availability for the affected locus, procedural unavailability for the institution, category misclosure in the decision path, and residual in the final judgement. A machine output may contain registered data, retained labels, active deployment, proxy closure, and target residual. The diagnostic question is always the same, but the answer differs: What relation is available to which locus, under which availability condition, through which abstraction, and under what closure demand? Without that question, extension cases become metaphor. With it, they become disciplined horizon symptoms. The extension of the Gödel Horizon beyond formal mathematics must therefore be constrained. The term should be reserved for boundary-disclosure where relation is available or formulable within a regime while closure cannot be supplied by the regime’s own present resources. Other cases should be named more exactly: access horizon, recognition horizon, procedural availability horizon, Soft Horizon, misclosure, residual horizon, abstraction gap, or action threshold. This does not weaken the architecture. It strengthens it by preventing one powerful term from absorbing all distinctions. The extension cases show how often finite systems live at availability/closure boundaries. They do not make every boundary Gödelian. 11. Stabilisation and excluded branch-space A successful abstraction regime does not stabilise against an empty background. It stabilises against a larger branch-space of failed, abandoned, latent, locally useful, excluded, suppressed, unrealised, or not-yet-available alternatives. Once the surviving branch is named, taught, institutionalised, formalised, encoded, or recorded, it can appear more inevitable than it was. The closure becomes familiar. The path that produced it becomes background. The alternatives disappear from view. This produces retrospective overclosure. A scientific theory that survived may appear to have been the natural destination of inquiry. A mathematical formalism may appear as if it were the only possible language for the relation. A legal category may appear as if it simply named the event. An institutional procedure may appear as if it were the obvious way to govern the relation. A machine benchmark may appear as if it captured the target because it became standard. A word may appear to have always meant what its current authoritative use requires. Positive stabilisation leaves records. Failed stabilisations often do not. The surviving abstraction produces textbooks, proofs, methods, categories, standards, instruments, archives, institutions, software, policies, and training data. It builds continuity around itself. It creates memory. It becomes easier to reproduce, teach, cite, test, automate, and defend. By contrast, failed or excluded alternatives may leave weaker surviving records or evidence. They may be forgotten, caricatured, reduced to error, treated as primitive, or lost because they did not become institutionally retained.8 This asymmetry matters because horizon analysis depends on branch-space. A closure cannot be assessed only by looking at the surviving output. The system must ask what alternatives were excluded, what made them fail, what residual they carried, what residual the surviving branch carries, and whether later authority has hidden the conditions of selection. Positive exclusion is stronger than forgetting. A formulation is positively excluded when it fails under stated conditions. A hypothesis may be ruled out by evidence. A proof path may fail. A legal argument may be rejected under a standard. A category may be shown not to preserve the relation. A machine proxy may be demonstrated inadequate. Positive exclusion reduces branch-space in a disciplined way because the reason for exclusion is retained. Forgetting is weaker. A branch may disappear because it lacked institutional support, because it was technically premature, because no instrument path existed, because the social setting could not receive it, because the formal language was unavailable, because the evidence path was procedurally blocked, or because the successful branch became dominant before the alternative was fully tested. A forgotten branch is not thereby live, but neither is it positively excluded. The distinction matters. Latent branches are different again. A relation may not yet have become available enough to form an alternative. It may be structurally implied, available in principle, or available in potential, but not yet coupled to a practice. Later developments may make it visible. The existence of latent branch-space does not license arbitrary speculation, but it prevents the surviving abstraction from claiming inevitability merely because no rival was retained. Local usefulness also complicates stabilisation. An abstraction may be locally strong while globally limited. A model may predict well in one regime and fail at another scale. A legal category may work for standard cases and fail at edge cases. A machine metric may correlate with a target in one deployment and fail in another. A perceptual pattern may guide ordinary action and misclose under unusual conditions. A locally useful Soft Horizon becomes dangerous only when its local success is treated as total warrant. Stabilisation therefore requires Counterfactual Comparison Discipline. The observer-system must ask what the closure was stabilised against. It must preserve not every imaginable alternative, but the relevant excluded branch-space: alternatives that were live, nearly live, structurally latent, procedurally unavailable, locally useful, or positively excluded under known conditions. Counterfactual Comparison Discipline is epistemic: presence in this branch-space does not establish that an alternative was physically available to the evaluated locus as a causal branch. This does not weaken closure. It strengthens it by showing what the closure has actually survived.9 A theorem survives because a proof path closes under rules. A scientific model survives because it predicts, explains, and withstands comparator pressure inside a domain. A legal judgement survives because it meets procedural and evidential standards. A machine output survives only if it preserves the target relation under deployment conditions and comparator challenge. A method survives because it reduces relational uncertainty without hiding new residual. In each case, strength depends not merely on positive form but on the pressure survived and the residual retained. The danger is that stabilised branches create their own Soft Horizons. The name of the theory, the category, the method, the model, the procedure, the proof, the benchmark, or the institution begins to stand for the whole claim-path. Later users inherit the closure without inheriting the branch-space it overcame. The stabilisation becomes reified. What was once an earned closure becomes a label. The label carries authority. The authority hides residual. This is one reason formal systems, institutions, and machines can magnify error. They retain and redeploy stabilised closures at scale. Retention is necessary. Without it, no cumulative knowledge or coordinated action is possible. But retention also compresses. A retained closure may lose the alternatives against which it was stabilised. A record may not carry the alternatives excluded. A training dataset may retain labels without the cases that challenged them. A legal archive may preserve judgements without lived residual. A scientific field may teach the winning theory without the latent assumptions that later become boundary conditions. Counterfactual Comparison Discipline keeps the closure honest. It asks which alternatives were excluded and how. It asks whether the exclusion was earned or merely inherited. It asks whether the surviving branch has been tested against the right comparators. It asks whether a Soft Horizon has travelled beyond the domain in which it survived. It asks whether residual has been converted into silence. This is especially important for the Gödel Horizon because disclosed boundaries often appear after long stabilisation. A formal system may look complete because it has closed many relations. A scientific theory may look total because it has succeeded repeatedly. A legal procedure may look legitimate because it is settled. A machine metric may look adequate because it is standard. Then a boundary appears. The system may treat the boundary as an external disruption. Horizon analysis treats it as a relation between stabilisation and residual. The boundary was not necessarily visible, but it was not arbitrary. It emerged where branch-space, abstraction, and closure demand met pressure. Stabilisation is therefore neither proof of total closure nor evidence of illusion. It is a finite achievement. It says that a closure has survived enough to become retained and redeployed. The discipline is to keep its conditions attached. The surviving branch may be strong. It is not inevitable merely because it survived. 12. The discipline of the horizon The discipline of the horizon begins from the fact that finite observer-systems live through horizons. They do not possess reality as total state. They encounter relation through boundary, interface, scale, abstraction, registration, retention, deployment, comparator pressure, closure demand, and residual. They produce Soft Horizons because without provisional closure they cannot recognise, speak, model, decide, prove, coordinate, or act. The task is not to abolish horizons. The task is to prevent horizon conditions from being hidden by the closures they make possible. The first discipline is relation specification. The system must state what relation is under review. Without relation, the horizon becomes empty language. A claim about “the unknown,” “the system,” “the real,” “risk,” “harm,” “proof,” “objectivity,” “intelligence,” or “evidence” remains under-specified until the relation being carried by the term is named. The second discipline is locus specification. The system must state for which locus the relation is available, registered, retained, deployed, closed, or residual. Without locus, availability drifts. A relation available to an instrument is treated as available to a person. A relation available to an institution is treated as available to an affected party. A relation available to a machine output is treated as available to the target relation itself. The third discipline is availability-status control. The system must distinguish unavailable relation, availability in principle, availability in potential, coupling, available contrast, registration, retention, deployment, closure, and residual. Without this discipline, existence becomes availability, availability becomes warrant, and warrant becomes possession. The fourth discipline is closure-demand control. The system must state what kind of closure is being sought. Recognition closure is not proof closure. Measurement closure is not ontological closure. Legal closure is not explanatory closure. Action closure is not epistemic closure. A claim may be strong under one demand and weak under another. The fifth discipline is abstraction exposure. The system must state the cut by which relation becomes classable, measurable, formal, admissible, computational, or usable. Without abstraction exposure, the category appears natural, the model appears identical to the target, the machine output appears objective, and the legal form appears to preserve the lived relation by default. The sixth discipline is Soft Horizon marking. Where provisional closure is being used, the system must state the purpose, scope, tolerance, residual, and update condition of that closure. A Soft Horizon is legitimate only while it remains recognisable as provisional closure under conditions. If its conditions disappear, it becomes reified authority. The seventh discipline is misclosure detection. The system must test whether the pattern applied preserves the relation under the demand being made. It must look for illusion, proxy substitution, category failure, scale drift, procedural overclaim, model-target collapse, proof-expression confusion, and output-warrant collapse. Misclosure is not solved by confidence. It is exposed by comparator pressure and update. The eighth discipline is residual localisation. Where closure has not been earned, the system must state what remains unresolved, what cannot be claimed because it remains unresolved, and what would force update. Residual is not a vague admission that uncertainty exists. It is the localised remainder that prevents stronger closure. The ninth discipline is boundary-disclosure. Where a regime makes relation available or formulable but cannot close it under its own resources, the system must name the boundary rather than hide it inside failure, paradox, mystery, or action. Boundary-disclosure is a disciplined output. It may lead to meta-level movement, abstraction revision, changed demand, residual, or a new Soft Horizon. The tenth discipline is branch-space preservation. The system must retain relevant excluded alternatives, not every fantasy but those live, latent, procedurally unavailable, locally useful, or positively excluded alternatives that bear on the claim’s strength. Without branch-space, the surviving closure appears more inevitable than it was. With branch-space, the closure can show what it has actually survived. The eleventh discipline is action-threshold control. When action must proceed under residual, the system must state the threshold, safeguard, and update condition. Action does not close the uncertainty under which it acts. A decision may be justified while the relation remains unclosed. The system must not later narrate action as if it had been proof. These disciplines converge on one central rule: Do not let a claim carry more strength than the relation-preserving path that supports it. This rule applies to perception. A recognised pattern should not become target possession. It applies to science. A model should not become ontology beyond its conditions. It applies to law. Procedure should not become exhausted truth. It applies to institutions. A category should not become the relation it governs. It applies to machines. Output should not become warrant by execution. It applies to mathematics. Expression should not be confused with proof. It applies to method. Completion of a form should not become evidence that relation has been preserved. The discipline of the horizon is therefore not scepticism in the loose sense. It is not doubt for its own sake. It does not flatten all claims into uncertainty. It does not deny earned closure. It does not prevent action. It does not reduce knowledge to perspective. It does not treat reality as invented by the observer-system. It asks each claim to preserve the relation it asserts under the strength it claims. Some claims are strong. Some closures are earned. Some proofs close. Some measurements are robust. Some categories preserve their target relation well enough for the use made of them. Some decisions are justified under threshold. Some Soft Horizons are legitimate and indispensable. The discipline protects these strengths by preventing them from being inflated, mislocated, or detached from their conditions. The unknown then becomes disciplined. It is not worshipped as mystery and not erased by premature closure. It is carried as non-availability, potential availability, unrecognised availability, procedural unavailability, misclosure, residual, live formulation, action threshold, or boundary-disclosure. Each form has a different status. Each requires a different response. At the Gödel Horizon, the required response is especially exact. The system has reached a boundary where relation is available or formulable but closure cannot be supplied by the present regime under the demand being made. The system may move to a meta-level, revise its abstraction, change its closure demand, adopt a Soft Horizon, carry residual, act under threshold, or build a new procedural path. What it may not do is pretend that the boundary has vanished because a usable closure was adopted. The Soft Horizon remains necessary after boundary-disclosure. A mathematician continues working inside formal systems after incompleteness. A scientist continues modelling after anomaly. A court continues deciding under evidential limits. A machine system may continue assisting under constrained deployment. A person continues navigating through perception. The point is not to stop finite operation. The point is to keep finite operation answerable to its residual. The discipline of the horizon therefore gives finite systems a way to proceed without pretending to possess. It allows claim, classification, model, decision, and action while preserving the relation among availability, abstraction, closure, residual, threshold, safeguard, and update. The unknown is not defeated by this discipline. It is made accountable. 13. Conclusion: working at the boundary A finite observer-system does not escape horizons by becoming more intelligent, more formal, more institutionalised, more scientific, more mathematical, or more machine-mediated. It moves horizons.10 Better instruments move availability boundaries. Stronger abstractions move recognition boundaries. More disciplined proof systems move formal boundaries. Better procedures move institutional boundaries. Better data paths move machine boundaries. Stronger comparators move model boundaries. Wider archives move historical boundaries. More careful methods move residual boundaries. But no finite system becomes horizonless. Every extension of availability introduces new cuts. Every closure regime carries residual. Every Soft Horizon that makes relation usable can be reified if its conditions stop travelling. The Gödel Horizon names the disciplined moment when this finite condition becomes visible as a closure boundary. Relation has become available or formulable. The closure demand is well-posed. The abstraction regime discloses that it cannot close the relation from within its own present resources. In mathematics, Gödel’s incompleteness theorems give the formal exemplar. A system can express what it cannot prove under its own rules. The boundary is not a failure of thought. It is thought learning the condition of its own closure. The wider lesson is not that everything is undecidable. It is not that all systems fail. It is not that closure is impossible. It is that closure is indexed and must be earned. A system may close strongly under one demand and remain residual under another. A Soft Horizon may be legitimate for action and inadequate for explanation. A record may preserve registration without warrant. A model may predict without possessing ontology. A category may govern without exhausting the relation it names. A proof may close within a system while the system’s own closure power remains bounded. The discipline is therefore neither scepticism nor finalism. It does not deny closure. It denies unearned migration of closure. It does not forbid Soft Horizons. It keeps them marked. It does not paralyse action. It requires threshold, safeguard, and update where action proceeds under residual. It does not dissolve reality into perspective. It preserves the fact that the target constrains while the finite locus constricts. To work at the boundary is to proceed without pretending to possess. It is to recognise that availability precedes recognition, recognition precedes many forms of closure, closure is demand-relative, residual must remain visible, and stronger abstraction may reveal rather than erase horizon. A finite observer-system becomes disciplined not by abolishing the unknown, but by preserving the conditions under which the unknown can answer back. The horizon is not the end of thought. It is the boundary at which thought becomes accountable to its own conditions. Endnotes References Gödel, K. (1931). Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I. Monatshefte für Mathematik und Physik, 38, 173-198. https://doi.org/10.1007/BF01700692 Humphreys, P. (2009). The philosophical novelty of computer simulation methods. Synthese, 169(3), 615-626. https://doi.org/10.1007/s11229-008-9435-2 Husserl, E. (1982). Ideas pertaining to a pure phenomenology and to a phenomenological philosophy, first book: General introduction to a pure phenomenology (F. Kersten, Trans.). Martinus Nijhoff. https://doi.org/10.1007/978-94-009-7445-6 Hutchins, E. (1995). Cognition in the wild. MIT Press. Kuhn, T. S. (1970). The structure of scientific revolutions (2nd ed., enlarged). University of Chicago Press. Platt, J. R. (1964). Strong inference. Science, 146(3642), 347-353. https://doi.org/10.1126/science.146.3642.347 Sahdev, D., Sundararaman, R., & Modgil, M. S. (2006). The Gödel universe: A practical travel guide. arXiv:gr-qc/0611093. https://doi.org/10.48550/arXiv.gr-qc/0611093 Simon, H. A. (1955). A behavioral model of rational choice. The Quarterly Journal of Economics, 69(1), 99-118. https://doi.org/10.2307/1884852 Szangolies, J. (2018). Epistemic horizons and the foundations of quantum mechanics. Foundations of Physics, 48(12), 1669-1697. https://doi.org/10.1007/s10701-018-0221-9 Wolpert, D. H. (2008). Physical limits of inference. Physica D: Nonlinear Phenomena, 237(9), 1257-1281. https://doi.org/10.1016/j.physd.2008.03.040 Endnotes 1.　Horizon provenance: Edmund Husserl, Ideas Pertaining to a Pure Phenomenology and to a Phenomenological Philosophy, First Book (1982 English translation), develops the phenomenological notion of an intentional horizon in which what is presently given is surrounded by further actual or potential possible experience. That lineage is acknowledged here because “horizon” has an established technical-philosophical history. The present horizon grammar is not inherited from phenomenology: it is locus-relative but non-conscious by default and requires an explicit relation, availability condition, abstraction, and closure demand. 2.　Bounded-rationality comparator: Herbert A. Simon, “A Behavioral Model of Rational Choice” (1955), gives a restricted decision-theoretic case in which finite agents simplify the choice environment and operate under bounded informational and computational conditions rather than exhaustive optimisation. The present Soft Horizon is broader: it applies to purpose-indexed provisional closure across perceptual, scientific, legal, institutional, formal, and machine relations and requires locus, closure demand, residual, and update to remain explicit. No bounded-rationality machinery is inherited. 3.　Distributed-locus comparator: Edwin Hutchins, Cognition in the Wild (1995), moves the unit of cognitive analysis beyond an individual and treats ship-navigation practice as a distributed cognitive and computational system. The present coupled-locus relation is lower-level and non-cognitive by default: it tracks where a relation is actually carried through distributed access, transformation, registration, retention, deployment, and closure without inferring a distributed mind or cognition from coupling alone. 4.　Procedural-opacity comparator: Paul Humphreys, “The Philosophical Novelty of Computer Simulation Methods” (2009), identifies epistemic opacity in computational science where relevant elements of a computational process are not epistemically accessible to an agent. The Procedural Availability Horizon generalises the missing-path condition beyond computation to evidence, instruments, archives, procedures, material conditions, temporal access, and realised branch paths, and it distinguishes such unavailability from internally disclosed closure limits. No theory of computational opacity is inherited. 5.　Terminological disambiguation: “Gödel horizon” also has an established but unrelated use in work on Gödel spacetime; Deshdeep Sahdev, Ravishankar Sundararaman, and Moninder Singh Modgil, “The Gödel Universe: A Practical Travel Guide” (2006), use the term for a causal and geodesic boundary in the Gödel universe. The present Gödel Horizon is epistemic and methodological and has no conceptual dependence on that spacetime usage. 6.　Epistemic-horizon comparator: Jochen Szangolies, “Epistemic Horizons and the Foundations of Quantum Mechanics” (2018), uses epistemic horizons for in-principle restrictions on information obtainable about a system and investigates paradoxical self-reference as a possible source of such limits. The correspondence is specific but limited: the present Gödel Horizon belongs to a wider typed availability/closure grammar and is not restricted to quantum-foundational or information-gathering limits. No categorical or quantum-foundational machinery is inherited. 7.　Formal dependency: Kurt Gödel, “Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I” (1931), supplies the formal incompleteness result used here as the exact mathematical exemplar. The theorem is an operational formal dependency only. Relational availability, Soft Horizon, Procedural Availability Horizon, and the wider finite-observer horizon grammar are not attributed to or inherited from Gödel. 8.　Stabilisation comparator: Thomas S. Kuhn, The Structure of Scientific Revolutions (2nd ed., 1970), provides a science-historical specialisation in which anomaly becomes visible against a stabilised paradigm and post-revolutionary teaching and textbook practices can make development appear more linear and inevitable than it was. The present account generalises that relation into stabilisation and excluded branch-space across domains, and it separately distinguishes positive exclusion, forgetting, latency, local usefulness, and residual. Kuhnian paradigm theory is not inherited. 9.　Comparator-discipline specialisation: John R. Platt, “Strong Inference” (1964), advocates constructing alternative hypotheses and using discriminating experiments to eliminate them rather than developing one favoured explanation in isolation. Counterfactual Comparison Discipline generalises the comparison burden beyond empirical hypothesis testing to any closure whose strength depends on relevant alternative branches, including formal, legal, institutional, model, and machine-mediated cases; it also distinguishes epistemic comparison branches from physically included causal alternatives. Platt’s method is a comparator, not an inherited procedure. 10.　Formal physical-limit comparator: David H. Wolpert, “Physical Limits of Inference” (2008), formalises physical inference devices that perform observation, prediction, or recollection and proves existence and impossibility results that constrain what such embedded devices can infer, independently of quantum indeterminacy or chaotic dynamics. The present claim that stronger finite systems move rather than abolish horizons is broader and does not adopt Wolpert’s inference-device formalism. The result is therefore a formal specialisation, not a dependency.