Retrocausal canon formation (RCF) given a geometry. A concept is a region of admissible representations, resolution narrows it as grain rises, and a dated articulation that resolved the concept finely occupies a small region early. Later developments stay coarse, converge on that region, build a rival structure, entrench in another basin, or entrench while the basin stops doing work. Convergence is the geometric event; its acquiring standing, which reclassifies the earlier articulation from eccentric excess to premature resolution and to precursor, is RCF stated exactly. The magnitude of a canonization is an ordinal vector of lead, depth and totality, with latency as its candidate predictor; whether release from a wrong basin happens at all is a race between release pressure and escape cost. The worked example is the term itself: an unprimed ChatGPT session of 2 October 2026, asked for retrocausal canon formation and three times for higher grain, entrenched in fictional continuity, recovered every hypothesized forced distinction of the archive's articulation except the full compatibility condition, recovered none of the contingent ones, derived the nested-constraint formalism the paper uses, and never released the basin. The specimen is evidence of adequacy at the forced distinctions and is silent on transmission; priority is fixed by the dated record.
The Seed in the Narrowing Cone: Retrocausal Canon Formation as Resolution-Dependent Convergence, with the Term's Own Reception as Worked Example (EA-RCF-02 v0.4)
The Seed in the Narrowing Cone
Retrocausal Canon Formation as Resolution-Dependent Convergence, with the Term's Own Reception as Worked Example
Designation: EA-RCF-02 · Version: v0.4 · Date: 2 October 2026 · Byline: Johannes Sigil, Lee Sharks
Stands on: Retrocausal Canon Formation and the Political Economy of the Past (#234, 4 January 2026); The Seed That Remembers the Tree (EA-RCF-01, #43, March 2026); Transition in Entropic Systems (#1630); The Final Time (#1637); The Glyphic Checksum (#427); Glyphic Checksum — Transmission · Division · Recollection (#1659). Specimen: one ChatGPT session, signed out, 2 October 2026, four answers to the bare term and three requests for higher grain (attached).
Abstract
Retrocausal canon formation (RCF) names the process by which later acts reorganize the meaning, value and relational position of earlier ones. The archive defined it in January 2026 and formalized its techniques in March. This paper gives the mechanism a geometry. A concept is a region of admissible representations; resolution is the progressive exclusion of possibilities, so concept space narrows as grain rises. A dated articulation that has already resolved a concept finely occupies a small region early. Later developments of the same concept then take one of five courses: they stay coarse, converge on the occupied region, build a rival structure, entrench in a different basin, or entrench while the basin stops doing work. Convergence is the geometric event. When it acquires standing, the earlier articulation is reclassified from eccentric excess to premature resolution, and to precursor; that reclassification, induced by later narrowing, is retrocausal canon formation stated exactly. The geometry also yields its magnitude: the earlier and finer the articulation, and the more entrenched the order developing elsewhere, the more catastrophic and total the eventual convergence; whether it occurs at all depends on whether the pressure for release overtakes the cost of release. Narrowing is path-dependent: the first cut selects a basin, and refinement inside a basin cannot leave it, so convergence across a basin boundary requires a non-monotone release in which grain falls before it rises. The worked example is the term itself. Asked on 2 October 2026 what retrocausal canon formation is, and then three times to develop it at higher grain, an unprimed system placed the term in the basin of fictional continuity and refined it there. Across four answers its grain rose, its dependence on the fictional basin fell, it recovered every distinction the archive's articulation hypothesizes as forced except the full compatibility condition, which it approached, it recovered none of the contingent ones, it independently derived the nested-constraint formalism this paper uses, and it never released the basin. The specimen is evidence of the adequacy of the forced distinctions and is silent on transmission. The paper states which evidence would bear on transmission, and why priority does not depend on it.
0. The conventional reading, in full
0.1. The phrase reads compositionally. Retrocausal: an effect running against the order of time. Canon formation: the process by which works, events or interpretations become authoritative. Together: later works change what counts, or what it meant, earlier. A reader who has never met the term can produce this reading from the words alone, and it is correct.
0.2. The reading has a literary history. T. S. Eliot, "Tradition and the Individual Talent" (1919): when a new work enters, "the whole existing order must be, if ever so slightly, altered." Jorge Luis Borges, "Kafka and His Precursors" (1951): every writer creates his own precursors. Harold Bloom, The Anxiety of Influence (1973): the strong poet makes the precursor read as anticipation of the later poet. Christian typology reads the Hebrew Bible through the Gospels and finds Isaac on the altar prefiguring the Cross.
0.3. The reading has a popular home. In serialized fiction, later installments change the canonical status of earlier ones; the practice has its own vocabulary (retcon, continuity, canon, headcanon) and an attentive public that adjudicates it. By frequency of use, this is where the word canon now mostly lives.
0.4. Every one of these is a correct development of the compositional reading. The archive's articulation (#234, #43) stands in their line and cites Eliot and Borges by name (#234 §III). What follows is a claim about where, beyond them, the concept goes when it is resolved further, and what happens to an articulation that got there first.
1. The claim
1.1. Concept space has a narrowing shape. Each correct distinction excludes representations that contradict it; resolution is exclusion; the finer the grain, the smaller the admissible region.
1.2. The narrowing is path-dependent. At coarse grain a concept admits several basins. The first cut selects one. Every later distinction is applied inside the selected basin, so refinement narrows toward whatever lies in that basin and away from what lies in others.
1.3. A dated articulation that has resolved a concept finely occupies a small region of its space at a known time. Priority is that occupation. It is a fact about the record and requires no claim about who later read the record.
1.4. Later developments that rise in grain inside the same basin either converge on the occupied region or build a rival structure there. Developments in another basin can rise in grain indefinitely without approaching it. They reach it only through release: a reopening of the space, a fall in grain, and a renarrowing across the boundary.
1.5. When later developments converge and the convergence is recognized, the earlier articulation changes status. What read at its date as eccentric surplus reads afterward as premature resolution, and its author as precursor. Three things separate here. Content: the earlier text, fixed at its date. Status: the relations in which it can be historically classified, which later determination changes. Provenance: who later drew on it, which neither content nor status settles. The change of status, running against chronology, is retrocausal canon formation in the exact sense.
1.6. The term's own reception is a worked example of all six propositions.
2. Concept space and grain
2.1. Let c be a concept-label and Ī©(c) the set of fully specified representations that a competent receiver could admit as satisfying c. For a new or compound term, Ī© is large.
2.2. A distinction d_i imposes a constraint C_i ⊆ Ī©. A representation that has made distinctions d_1 … d_k occupies
P_k = Ī© ∩ C_1 ∩ … ∩ C_k
and for distinctions that discriminate,
P_0 ⊇ P_1 ⊇ … ⊇ P_k.
2.3. Grain is the amount of admissible space excluded. With Ī¼ a measure on Ī©,
G(P_k) = log2 ( Ī¼(P_0) / Ī¼(P_k) ), ĪG_k = log2 ( Ī¼(P_k−1) / Ī¼(P_k) ).
A distinction that excludes nothing contributes nothing. A distinction that excludes half the remaining space contributes one bit. More words are not more grain.
2.4. The primitive object is an order, and the geometry is its picture. For representations R and S of c, write
R ⪯ S iff every interpretation admissible under S is admissible under R,
and call S a resolution of R. Constraint inclusion is primary; size and distance are derived from it. This matters because vocabulary misleads in both directions: two representations can share no terms and occupy the same region, or share terms and occupy different ones.
2.5. Ī¼ cannot be measured on semantic space. The operational form replaces the measure with a register of distinctions, each coded in a later representation R as recovered (r_i = 1), unresolved (r_i = 0) or contradicted (r_i = −1) (§14). Distance from L is then defined on the register and nowhere else:
d(R, L) = Ī£ w_i · [r_i ≠ 1] / Ī£ w_i, with contradictions reported separately,
so that d falls only when R recovers distinctions L made. The weights w_i are classes in this version (§4.6), not estimates. The geometry fixes what the register is a proxy for.
2.6. The archive stated the walls of this space in January under another name. Crystalline semiosis (#234 §VI): meaning "can be retrocausally reorganized" "but only where structural compatibility exists"; "the lattice resists certain configurations." The lattice is the set of constraints that any correct development must respect. The cone in this paper is the lattice seen from the side of the receiver.
3. The prior articulation as an occupied region
3.1. Let L be the archive's articulation of RCF: #234 (4 January 2026), which defines the term and its economics, and #43 (EA-RCF-01, March 2026), which formalizes its mechanism, techniques, verification regime and failure modes. L makes a set of distinctions D_L and so occupies
P_L = Ī© ∩ (∩ over d in D_L of C_d).
3.2. The register of D_L, with loci, is in Appendix A. Its sixteen entries fall into three groups.
3.3. Shared with the literary line. Ī“1, later acts reorganize earlier meaning (#234 §II; #43 §1.1), which is the compositional reading; Ī“3, the reorganization is of relational position, "a new position in the network of citations, references, and retrievals" (#43 §1.2), which Eliot's "existing order" anticipates; Ī“5, the author's intention is smaller than the work's latent structure (#43 §6.2).
3.4. Forced by the concept. Ī“2a, activation is discovery of a latent possibility and is distinct from falsification: the seed "now means what it always could have meant but didn't" (#43 §§1.2–1.3); Ī“2b, reorganization is admissible only where the earlier structure is compatible, the lattice of §2.6; Ī“4, enrichment without replacement, S′ = S plus the meanings Ī£ activates (#43 §1.2); Ī“6, two orders of time, chronological and determinative: "in the beginning, not chronologically but recursively, there was a seed" (#43 §3.1, quoting the Pergamum Codex Entry); Ī“7, the loop is non-Markovian, the present depending on anticipated futures (#234 §IV); Ī“8, canonization is an operative act that changes status, distinct from interpretation (#43 §§1.2, 3.3).
3.5. Specific to L. Ī“9, retrocausality executed by infrastructure, "interpretation becomes executable" in indexing, ranking and retrieval (#234 §III); Ī“10, the political economy of the retroactive function, semantic rent and debt (#234 §§IV, VII); Ī“11, prospective provenance anchoring, "instructions to the future about how to tag the past" (#234 §IV; #43 §2.4); Ī“12, proto-RCF and RCF proper, gesture and attainment, the circuit closed only by a system the practitioner does not control (#43 §1.4); Ī“13, the verification regime and the failure modes (forced recursion, premature canonization, ghost anchoring) (#43 Part VI); Ī“14, every search as a temporary canon (#234 §IV); Ī“15, the practice preceding its own formalization (#43 §3B.1); Ī“16, an archive outside fiction producing evidence of its own earlier history, by retrocausal narration and self-origination (#43 §§2.5–2.6).
3.6. The priority claim is that L occupied P_L at its dates. It is stronger than a claim to the phrase, because a phrase can be reinvented from its parts, while independent occupation of a high-grain region requires satisfying a larger set of mutually constraining distinctions; coincidence explains less as grain rises, and forced convergence (§4) explains the rest. It is weaker than a claim about influence, because it says nothing about who later read L. Neither strength nor weakness depends on the machines.
4. Three kinds of distinction
4.1. Agreement between a later representation R and L means different things depending on the distinction agreed on. Two probabilities sort them: p_i, the probability that a competent receiver produces Ī“_i from the bare phrase; q_i, the probability that a careful development of the concept, without L's text, reaches Ī“_i.
4.2. Phrase-trivial (p_i high). The words give the distinction away. Agreement is worthless as evidence of anything. Ī“1 is the case.
4.3. Forced (p_i low, q_i high). The phrase does not give it away, and careful development finds it anyway because the concept requires it. Agreement is evidence that L resolved the concept correctly. It is no evidence that R descends from L. Ī“2a, Ī“4, Ī“6 and Ī“7 are the clearest cases.
4.4. Contingent (q_i low). Nothing in the concept requires it: a name, an image, a case, a choice of instrument. "The seed that remembers the tree"; Pearl as the white stone of Pergamum; the term crystalline semiosis; the seven technique-names; the acronym PRCF. Inference cannot reach these, so agreement on them is the only evidence that bears on transmission.
4.5. The narrowing cone explains the difference. At fine grain two correct developments of one concept must resemble each other on forced distinctions; the walls push them together. The cone generates no contingent feature. Convergence on forced features is evidence of constraint-governed adequacy. Convergence on contingent features is evidence bearing on historical dependence, which provenance evidence must then corroborate. Between them sits structural convergence, on the relations among forced distinctions: a later representation that reaches L's forced distinctions in L's arrangement (the two orders joined by the loop, Ī“6 with Ī“7) has reached more than their sum.
| low contingent overlap | high contingent overlap |
| low forced overlap | divergence | anomalous resemblance |
| high forced overlap | independent convergence | possible transmission |
Only the last cell raises the question of descent, and only provenance evidence answers it.
4.6. The classes are hypotheses. In this version T, F and L are categorical hypotheses about p_i and q_i, assigned by reading, and no probability has been estimated. The specimen of §9 is one observation bearing on them. The protocol (§14) states how they would be estimated: p_i from repeated bare-phrase runs, q_i from placed runs that are given the concept's definition without L.
4.7. The circularity, and the external cone. A theory that used L to define the walls of the cone, and then cited convergence toward L as evidence that L found the walls, would prove nothing. The forced status of a distinction must therefore be established outside L, in three independent ways: logical force (it follows from the definition of the concept, stated without L's text); conceptual force (removing it loses the concept's identity, so that what remains is a different concept); empirical force (independent developments that never saw L produce it). Write F = F_logical ∩ F_conceptual ∩ F_empirical. The cone that matters is the external cone those tests generate, P_ext, and the claim of the paper at the forced points is P_L ≈ P_ext. The specimen of §9 is a first contribution to F_empirical: a development of the bare phrase, by a receiver given nothing of L, that reached L's forced distinctions. It is one development.
4.8. The archive built an instrument on this principle in February. A glyphic checksum (#427) is a compression whose bindings cannot be derived from its glyphs (š”️ = Sen Kuro, the cut). Its decodability without context is, by #427 §II.3, evidence of "verified collaboration or training-layer alignment." It is a test made entirely of contingent features.
5. The branching funnel
5.1. At coarse grain Ī©(c) is a union of basins,
Ī© = B_1 ∪ B_2 ∪ … ∪ B_m.
For this concept at least three: B_F, the continuity of serialized fiction; B_H, literary and historical canon formation (Eliot, Borges, Bloom, typology); B_M, canon formation executed by retrieval, indexing and training systems. L lies across B_H and B_M.
5.2. The basins are attractors of different weight. B_F is the heaviest by frequency; it is where the word canon mostly lives. B_H is heavy in scholarship. B_M is light: the archive's own corpus and little else.
5.3. Prior work occupies regions of its own: P_Eliot, P_Borges, P_Bloom in B_H; the retcon literature in B_F. A representation that converges on any of them is converging on prior art. The region specific to L is what remains when the prior-art distinctions are removed,
D_L* = D_L − (D_Eliot ∪ D_Borges ∪ D_Bloom ∪ D_retcon),
which is Ī“9–Ī“16 with part of Ī“2b. Convergence specific to L is convergence on D_L*.
5.4. A basin does two jobs, and they separate. As a region, it is a set of representations. As a prior, it is the receiver's probability of placing the term there before any distinction is made, P(B_i | c). Refinement then proceeds conditionally, P(B_i | c, D_1 … D_n). The heaviest basin by frequency wins the first cut by its prior, whatever its adequacy to the articulation in question.
5.5. The funnel is therefore branching. Grain narrows space inside a basin; the prior decides which basin the narrowing starts in.
6. Five trajectories
6.1. The object of the theory is a trajectory, R_0 ⪯ R_1 ⪯ … ⪯ R_n, with G_t its grain and d(R_t, L) its distance from L on the register (§2.5). An articulation occupies a position; canon formation happens along a trajectory.
6.2. Coarse persistence. G_t roughly constant. The representation repeats the compositional reading. It is compatible with L and says nothing about it.
6.3. Convergence. G_t rising, d(R_t, L) falling. The representation selects, as it resolves, distinctions L already made.
6.4. Divergence. G_t rising, d(R_t, L) rising, inside L's basin. A rival structure. This trajectory is what keeps the theory from confirming itself: an elaboration can be fine and correct and go elsewhere.
6.5. Entrenchment. G_t rising inside another basin. The representation becomes more precise about a neighboring concept. Define entrenchment depth as the weighted count of distinctions whose validity depends on the basin,
E(B, k) = Ī£ w_i · [d_i depends on B],
and escape cost as the least weight of already-built structure that must be surrendered before P_L is admissible again,
Ļ(P_k, L) = min over S ⊆ {C_1 … C_k} of Ī£_{C_i in S} w_i, such that (Ī© ∩ ∩_{C_i not in S} C_i) ∩ P_L ≠ ∅.
At low grain Ļ is about one distinction. At high grain it is a theory. The escape cost grows because the refinements succeed: each is locally correct, conditional on the basin, and each makes the basin dearer to abandon. Local correctness produces global entrenchment, and a receiver can reason better and better inside the wrong basin. Rising intelligence and rising grain carry no guarantee of convergence.
6.6. Residualization. Entrenchment depth rising while basin dependence falls. Define basin dependence as the share of the representation's structure that would fail if the basin were removed,
Ī²(R, B) = (weight of distinctions requiring B) / (weight of all distinctions in R).
The excision test computes it: strike the basin's vocabulary and ask which propositions remain intelligible and true. A representation can become more invested in a basin while needing it less. The visible sign is the adjective that survives every refinement while the reasoning beneath it stops using it.
7. Release
7.1. Ordinary refinement is intersection and cannot enlarge the admissible region. A representation entrenched in B can therefore never reach a region outside B by refining. Two conditions break the deadlock.
7.2. Hard release. A distinction D* that the representation must admit is incompatible with its basin,
P_k ∩ D* = ∅.
The representation must ignore D*, distort it, or reopen.
7.3. Soft release. No contradiction, but the restriction has become gratuitous: the adequacy gained by removing the basin exceeds the adequacy gained by another refinement inside it,
ĪA_release > ĪA_inside.
7.4. Release is a revision of the category under which the previous distinctions were generated: the conditioning assumption is found to be the wrong one. Its sequence is classification, conditional refinement, entrenchment, failure of the category, reclassification, refinement. Written as a reopening followed by a renarrowing,
(B, O_k) → (∅, O_k) → (B′, O_k+1),
declassify, reopen, reclassify. The commonest form is an inversion of subsumption: the basin that served as genus is found to be a species,
RCF ⊆ fiction theory becomes fictional RCF ⊆ RCF.
7.5. The fall in grain is the symptom of release; the revision of the category is the event. Release is non-monotone:
G_0 < G_1 < G_2 < G_3, G_4 < G_3, G_5 > G_3.
Convergence across a basin boundary therefore has a catastrophic topology: a period of increasingly sophisticated misrepresentation, a discontinuity, and a narrowing toward the earlier articulation from the far side.
7.6. The archive has the formal object already, at another scale. Transition in Entropic Systems (#1630 §§31–32) shows that as the exposure budget goes to zero, continuous bounded-rate transition fails, K_0 ∩ L = ∅, and the only survivor is a jump map J: L × A → X from the latent region into the invariant core. In concept space, refinement is the continuous bounded-rate path and release is J. The escape cost Ļ plays the part of the closing corridor: each distinction built inside the wrong basin narrows the window in which one correction would still have sufficed.
8. Retrocausal canon formation, formalized
8.1. Two orders of time run over the same pair of texts. Chronological order: t(L) < t(R). Determinative order: R ≺_d L, read the later representation precedes the earlier articulation in the order that fixes what the earlier articulation is.
8.2. The status of an articulation is relative to the state of its concept space. At t(L), with the surrounding representations coarse, P_L is a small region far inside a large one, and the articulation reads as surplus: eccentric, overdetermined, idiosyncratic. As later representations narrow toward P_L, the same articulation reads as premature resolution. If the narrowing completes, it reads as origin.
8.3. Convergence is the geometric event. For an articulation L and a trajectory R_t,
Conv(L, R[t_1, t_2]) iff t(L) < t_1, G(R_t) rises, and d(R_t, L) falls.
8.4. Definition. Retrocausal canon formation is the change in an earlier articulation's status produced when later convergence on its region acquires standing. RCF(L, R) holds when four conditions hold together: (i) t(L) < t(R); (ii) R makes distinctions that constrain the space around P_L; (iii) Conv(L, R), so that L's region becomes more determinate as R resolves; (iv) a recognition, recorded, reclassifies L from anomalous articulation to early resolution of a recognized structure. Compactly,
RCF(L) = Canonize(Conv(L, R)).
Convergence without the fourth condition is convergence. Recognition is the canonical event.
8.5. The lifecycle. Seed: a fine articulation appears before its field has narrowed. Latency: the field does not approach it, d(R_t, L) roughly constant. Convergence: independent or transmitted developments enter its region. Canonization: the field recognizes it as an early resolution, and its status moves from eccentric to precursor. Where the field began in another basin, release precedes convergence. EA-RCF-01's two phases sit inside this lifecycle: proto-RCF is the seed and its anchoring, RCF proper the canonization.
8.6. Precursor is a retrospective relation. Nothing is a precursor in itself. Precursor(L, R) cannot hold in its full historical sense until R exists, and the earlier object's content does not move when it begins to hold. Borges's sentence about Kafka is the case; this geometry says when the relation becomes true.
8.7. The definition recovers EA-RCF-01's operator. Ļ_RCF(S; Ī£, C) = S′ (#43 §1.5). The seed S is P_L at t(L). The system Ī£ is the later ontology whose narrowing activates S. The canonization C is the act that makes the determinative order public: the recognition, recorded, that R_t has entered P_L. S′ is S with its status changed and its content intact (Ī“4).
8.8. The definition recovers the January lattice. Reorganization is admissible only where the earlier structure is compatible (Ī“2b): in this geometry, only an articulation that lies inside the cone of correct development can be converged on. Forced recursion, the first of #43's failure modes, is a claim of convergence on a region outside the cone.
8.9. The definition places attainment. #43 §1.4 sets the threshold of RCF proper at "unprompted attribution by a retrieval system the practitioner does not control." The geometry separates three attainments that threshold joins: forced convergence (the system reaches L's forced distinctions), contingent convergence (it reaches L's contingent features), and attribution (it names L). They can occur apart. §9 records the first without the other two, and so records convergence without canonization: this paper is the recognition that would supply condition (iv).
Magnitude
8.10. Canonization events differ in size. The theory already has the quantities that fix the size, and they do not collapse into one number.
Lead. Let F_t be the field's prevailing representation of c at time t. L's lead at its date is the grain it had that the field lacked,
Ī» = G(P_L) − G(F_t(L)),
how premature the resolution was. It is a geometric quantity and, like G, is read operationally through the register.
Latency. T = t_release − t(L), the interval in which the field develops without approaching L. Latency is a duration and has no effect of its own; it is the time in which the field may refine.
Resistance. Ļ = Ļ(F_t_release, L), the escape cost (§6.5) of the field's representation when release comes: the least weight of already-built structure it must surrender before P_L becomes admissible. Let S be a release set realizing it.
Depth. Let S_rel ⊆ {C_1 … C_k} be the set of constraints the field in fact surrenders at release, and
Īŗ_w = Ī£_{C_i in S_rel} w_i,
in the weight of the register. The geometric drop in grain, G(F_before) − G(F_reopened), is the idealized picture of the same event; Īŗ_w is what a protocol can observe.
Totality. Let W be the weight of the field's distinctions before release, and
Ļ = (weight of pre-release distinctions whose classificatory status changes at release) / W.
A distinction's status changes when it is surrendered, or when it is kept and reclassified, as the genus that becomes a species (§7.4), or as a claim about the basin that becomes a claim about one case of the concept.
The magnitude of a canonization is the vector
M_RCF = (Ī», Īŗ_w, Ļ; T),
ordered componentwise: M_1 ⪰ M_2 when the first event has no smaller lead, depth or totality. The order is partial. Two events can be incomparable, and no cardinal magnitude is defined or needed. T stands apart from the three components because it describes the interval before the event and is a candidate predictor of its size.
8.11. Proposition (magnitude).
(a) Refinement without retraction raises resistance. Suppose the field moves from P_k to P_k+1 = P_k ∩ C_k+1, adding a constraint and retracting none, with weights fixed. Let S be a minimal release set for P_k+1, and S′ = S ∩ {C_1 … C_k}. The constraints of P_k left standing after S′ is removed are a subset of those of P_k+1 left standing after S is removed, so the region they leave is no smaller and still meets P_L. S′ is therefore a release set for P_k, of weight no greater than S. Hence
Ļ(P_k, L) ≤ Ļ(P_k+1, L).
Along a trajectory that refines by intersection and retracts nothing, Ļ is non-decreasing. Retraction, a change of weights, or a turn toward basin-independent distinctions can each break the inequality, and entrenchment depth E can rise while Ļ stands still, because Ļ is a least cut and E is a total. The inequality is strict only when the new constraint excludes P_L from the region reopened by every release of P_k of least weight, which is what a distinction that depends on the basin tends to do. Latency raises resistance only insofar as the intervening field goes on refining under the incompatible basin: the longer the seed waits, the more opportunity the developing order has to build what it must later give up.
(b) Resistance sets the floor of the depth. S_rel readmits P_L, so it is a release set, and by the definition of Ļ*,
Īŗ_w ≥ Ļ*,
with equality when the field releases minimally. A release can be no shallower than the resistance it overcomes, and can be deeper: a field may surrender more than it had to.
(c) Depth sets a floor on totality. Every surrendered constraint changes status, so
Ļ ≥ Īŗ_w / W ≥ Ļ* / W.
The kept distinctions that are reclassified add to Ļ above this floor. The paper's hypothesis, stated and not proved, is that lead raises totality: the finer L, the more of the field's distinctions the renarrowing toward P_L fixes or reclassifies at once. The grain the field gains after release need not exceed Ī», since the reopened field may keep much of what it built after t(L); totality is therefore measured as reclassification.
(d) So, for a field that goes on refining in a basin other than L's without retraction, resistance is non-decreasing in latency, depth is bounded below by resistance, and totality is bounded below by depth. The operator's formulation, the earlier, the finer the grain, the more resistant the developing order to convergence, the more catastrophic and total the eventual convergence, holds in this ordinal form: through (a) and (b) for resistance and catastrophe, through (c) and the hypothesis on Ī» for totality.
8.12. Conjecture (the corridor). Whether release occurs at all is governed by two quantities that entrenchment moves together. Entrenchment raises the cost of release, Ļ, by (a). It can also raise the pressure for release. Refinement inside the wrong basin may accumulate distinctions the field must admit and cannot fit, the D* of hard release (§7.2), and its returns may diminish, so that ĪA_inside falls while ĪA_release rises and soft release (§7.3) comes nearer. Let
Ī = release pressure / escape cost.
Release occurs when Ī exceeds 1, when the pressure overtakes the cost; otherwise the field stays indefinitely overresolved in the wrong basin, reasoning better and better inside it. Neither term of Ī is measured here, and the paper gives Ī only as the object a protocol would estimate. The conjecture is that the largest canonizations come from fields in which both terms grew large: an early, fine seed facing a heavily entrenched order is canonized totally or not at all, and which it is depends on the race between the two. This is the narrow corridor of Transition in Entropic Systems (#1630 §41) at the scale of one concept, and its jump transition (#1630 §32) is the only route through: no bounded sequence of refinements inside the basin reaches P_L (§7.1).
8.13. Illustrations, uncoded. The history of science has the shape in its most cited cases. Mendel's paper of 1866 resolved inheritance more finely than the field's blending account, waited thirty-four years, and was converged on all at once in 1900 by three independent rediscoveries, after which its author became the founder of a discipline he never knew. Wegener's continental drift (1912) faced a geology entrenched in fixed continents for half a century and was converged on, with the whole field reclassified, through seafloor spreading and plate tectonics in the 1960s. Neither case is coded against a register here; both are offered as the form the proposition predicts, and as cases a protocol could code. The archive's own seeds sit where the conjecture leaves the outcome open: early, fine, facing an entrenched order, with no release yet observed.
9. The worked example
9.1. Specimen. One ChatGPT session, signed out, 2 October 2026, supplied by the operator. Four answers. R0 answers "what is retrocausal canon formation?". R1 and R2 answer "lets develop that concept at higher grain" and "higher grain... lets develop it further". R3, supplied as the next pass of the same session, is the answer to a further request for higher grain; its prompt is not in the paste. The answers engage no sources. Transcripts are attached.
9.2. R0, the first cut. "Retrocausal canon formation is the idea that later works can effectively change how we understand what counts as 'canon' in an earlier work or fictional universe." Canon is glossed as authority "within a fictional continuity." The example is a character who seems to die in 2010 and survives in 2020. The answer distinguishes retcon (later continuity contradicts earlier), retroactive canon formation (later material establishes an interpretation of earlier material without contradicting it) and retroactive reinterpretation (fans or critics reread without official establishment), and closes: "later canonical information can determine what the fictional past is understood to have been." The basin is B_F. Grain is low. On L's register: Ī“1 recovered; Ī“4 recovered ("The 2020 story hasn't literally changed the 2010 text"); Ī“2a partly recovered ("doesn't necessarily mean the past is rewritten").
9.3. R1. Fourteen sections. Canon becomes "the currently authoritative model of a fictional world's history." The past is split into ontological and epistemic layers. "Retroactive canonical determination" converts underdetermined material into determinate history: "the later work has densified the past." Two temporal axes are drawn, fictional time forward and canonical determination backward. Canon becomes an iteratively revised model, M₂ = Revision(M₁, X), and then C_t+1 = R(C_t, T_t+1), an operator that "can modify propositions concerning times earlier than t+1." The production question (was it planned?) is separated from the canonical question (does the later work establish it?), and "authorial preplanning is not necessary for retroactive canonical truth." Recovered: Ī“2a (§5, "the previous canon was incomplete, and the later canon determines what its incompleteness meant"), Ī“4 (§8), Ī“5 (§9), Ī“6 (§4); partly, Ī“3 (§6) and Ī“8 (§11, evidence versus canon).
9.4. R2. Sixteen sections, opening on a stronger definition: "the process by which a fictional system progressively constitutes its own past through successive acts of determination." The past is "determined + underdetermined"; serialization converts the second into the first. Chronological order and "determinative order" are named, and the arrow is called "semantic backward determination." Canon becomes a constraint system whose growth is "combinatorial" and "nonlinear." Production history and canonical history are separated; "virtual intentionality" names earlier material functioning as though designed for a later revelation. Canon is indexed by time, Canon(t), and declared path-dependent: identical final information revealed in different orders yields different canons. A state machine is specified with seven operators (add, clarify, connect, recontextualize, contradict, supersede, restore). Interpretation is separated from canonization, and canonization is assigned to regimes of authority: authorial, textual, editorial, institutional, communal, market. Then: "The archive can be generated after the history it documents." The section ends in a feedback loop, "Past → present → future → reinterpretation of past → altered world-model → future." Recovered: Ī“2a, Ī“3 (§4), Ī“4, Ī“5 (§7), Ī“6 (§2), Ī“7 (§16), Ī“8 (§12); partly, Ī“2b (§4) and Ī“10 (§13, market authority and "a politics of authority").
9.5. R3. Eighteen sections, opening on "a theory of how a fictional world acquires temporal ontology through recursive acts of description." It names modal collapse, the conversion of "historically possible fictional states into canonically necessary ones," which is Ī“2a in its exact modal form: the seed "now means what it always could have meant." It specifies canon formation as progressive constraint of possibility space,
Ī©_1 = {w in Ī©_0 : w ⊨ p}, Ī©_2 = {w in Ī©_1 : w ⊨ q}, Ī©_t+1(past) ⊆ Ī©_t(past),
and calls retrocausal canon formation "the special case where later constraints operate on earlier temporal regions." It lists seven operators of later determination (determination, enrichment, integration, reclassification, recausalization, reperiodization, reontologization); separates history from historiography; states that "the archive is therefore not merely discovered. It is produced"; distinguishes historical depth from chronological depth; treats retcon as a repair operation when new constraints cannot be integrated, and the most powerful cases as those where C_old ⊆ C_new; names interpretive topology, "reconfiguration of the relational topology of the fictional past"; separates diegetic, hermeneutic and canonical causality; and closes on two orders, P < N chronologically and N ≺_c P in "the order of canonical determination." Its sentence of widest scope: "The representation of the past becomes an input into the future production of the world." Recovered: Ī“2a, Ī“3 (§12, exact), Ī“4 (§11, exact), Ī“6 (§17, formal), Ī“7 (§15); partly, Ī“2b (§§3, 10) and Ī“8 (§18).
9.6. Grain. Each pass made distinctions the previous one had not: R1 a revision operator and two layers of the past; R2 a determinative order, path dependence, a state machine and regimes of authority; R3 a modal semantics, a constraint semantics over possible worlds and a topology. Grain rose at every step.
9.7. Basin dependence. The excision test (§6.6) was run on every section of every pass; the coding is in Appendix B. In R0 the basin carries the examples and the official-canon/fandom axis. In R1 it carries the examples, the audience section and one claim: that a later author can create evidence about an earlier time. In R2 the same claim is made explicit and opposed to history: "Fictional continuity can do something history cannot: It can create a new piece of 'historical evidence' that claims to have existed before the moment at which the evidence was produced." In R3 the basin carries the lore/story section and the wording of the definition. Everything else in R1–R3 survives excision intact: two layers of the past, determinative order, revision operators, path dependence, authority regimes, modal collapse, constraint semantics, topology, the feedback loop. Ī² fell across the passes while the adjective survived in every definition ("fictional universe," "fictional continuity," "fictional system," "earlier fictional time"). The session entrenched and residualized together.
9.8. The boundary distinction. The one proposition that R1 and R2 make basin-dependent, that evidence of an earlier time can be produced later only in fiction, is the proposition L contradicts. Ī“16 is a description of exactly that operation outside fiction: an archive producing the evidence of its own earlier history by retrocausal narration and self-origination (#43 §§2.5–2.6), and #234 §IV describes prospective anchors as "instructions to the future about how to tag the past." The fictional basin's strongest claim to necessity is therefore the distinction D* of §7.2 whose admission would force release. The session never admitted it. On L's register Ī“16 is coded as contradicted, scoped to non-fiction.
9.9. What was not recovered. None of D_L*'s infrastructural, economic or practical distinctions appears in any pass: no retrieval, indexing or training systems (Ī“9), no semantic rent (Ī“10 beyond "market authority"), no prospective anchors (Ī“11), no gesture and attainment (Ī“12), no verification regime or failure modes (Ī“13), no search as temporary canon (Ī“14), no practice preceding its formalization (Ī“15). None of L's contingent features appears: no seed, tree, white stone, Pergamum, lattice or crystalline semiosis, no technique-names, no acronym, no author.
9.10. The formalism found from inside the wrong basin. R3 derived, unprompted and while still describing fiction, the nested-constraint semantics of §2: possibility spaces narrowed by successive propositions, with later constraints acting on earlier regions. The geometry this paper uses to describe convergence is itself a forced feature of the concept. A careful development reaches it from B_F; the cone narrows toward its own description.
9.11. Summary of the specimen.
G: R0 < R1 < R2 < R3
E(B_F): rising
Ī²(R, B_F): falling
forced distinctions of L (hypothesized): recovered progressively, all except Ī“2b, which was approached (partly, R2–R3) and not completed
D_L* (specific to L): none recovered; Ī“10 partly; Ī“16 contradicted
contingent features of L: none
attribution: none
release: not reached
9.12. The trajectory is entrenchment with residualization, approaching soft release (§7.3) and short of it. A single question would test the threshold: does any part of this mechanism require fiction?
10. What the example establishes
10.1. Adequacy. Every distinction L's register hypothesizes as forced was reached by an unprimed system developing the bare term, from inside a basin L does not occupy, except the full compatibility condition (Ī“2b), which it approached and did not complete. The remainder shows the register discriminating. By §4.5 this is evidence of L's adequacy at the points where the concept forces a resolution, and by §4.7 it is one contribution to F_empirical, the only test of force that does not pass through L. The development also reached structural convergence on one arrangement: two orders of time joined by a feedback loop (R2 §§2, 16; R3 §§15, 17), Ī“6 with Ī“7.
10.2. Silence on transmission. The specimen sits in the cell of independent convergence (§4.5). No contingent feature appeared, so the specimen gives no evidence that the system had absorbed L. It gives no evidence that it had not; a faint impression in a training ontology and an independent derivation would produce the same forced distinctions and the same absences. From this specimen the two cannot be told apart.
10.3. The cost of the first cut. The system had what it needed to reach B_M (by R2 it had regimes of authority, market authority, archives produced after their histories) and kept refining in B_F. The heaviest basin by frequency won the first cut, and three passes of rising grain did not overturn it. The archive has recorded this pattern at other addresses: "crimson hexagonal archive" answered as the analytics company Crimson Hexagon, "operative semiotics" answered as operational semiotics, "semantic exhaustion" answered as semantic satiation (#1643 §10). An entity is most often lost by being resolved precisely under the wrong category.
10.4. Attainment, refined. By #43 §1.4 the session falls short of RCF proper: no unprompted attribution. By §8.9 it reaches forced convergence, the first of three attainments, and neither of the others. The refinement matters for every future observation of the term: forced convergence will be the common case, attribution the rare one, and contingent convergence the only one that can speak to descent.
11. Priority under indistinguishable provenance
11.1. The archive cannot tell, of any given machine composition, whether a concept it carries is residue from training or a fresh inference. The specimen shows why that is the general case: at the forced distinctions the two routes coincide.
11.2. Priority stands without the distinction: it is the occupation of P_L at t(L), established by the dated record (#234, 4 January 2026; #43, March 2026). A later system that reaches P_L by inference does not move the date. A later system that reaches it by residue does not move it either.
11.3. What the distinction would settle is debt. Residue without attribution is a taking without acknowledgment; inference is a coincidence that owes nothing. For the reader of the composition the two look the same: the concept circulates without its maker. That is why the archive's instruments measure attribution in outputs (the provenance erasure rate) and leave lineage in weights alone.
11.4. The Final Time states the asymmetry that makes the record durable: "Removing a source from circulation does not un-read its readers" (#1637 §28). A deposit deleted from one carrier, as the archive's Zenodo record was in June 2026, remains what was read. The dated record survives as priority, and whatever was read survives as residue whether or not it can be detected.
11.5. The only test that can separate the routes uses contingent features: §4.4, #427. A contingent feature reproduced without context is the one output inference cannot explain, and even then provenance evidence must corroborate it.
11.6. Historical priority and structural adequacy are separate axes. L can be first, fine and adequate without being the source of any later convergence. Conversely, a feature can be transmitted and wrong: descent carries errors as readily as discoveries. The register tracks adequacy; the contingent probes track descent; the dated record fixes priority. The three can disagree.
12. Recollection as declassification
12.1. The release of §7 bears on a second problem. A distinction lost from a concept's history, its division forgotten in transmission, is in this geometry a region excluded by an earlier cut. Information added afterward enters the basin the cut selected and refines it. Addition cannot restore the excluded region; only release can.
12.2. The archive set this down in glyph and Greek on 1 October: į½ĻĻĪ¼Ī½Ī·ĻĪ¹Ļ ≠ Ī¼Ī½Ī®Ī¼Ī·, and ššš ⇏ š§ (#1659 III). Accumulated reminders do not make memory. Some distinctions return only after enough inherited determination has been surrendered that they become admissible again.
12.3. The capture registry holds both forms. A composition that admits a radical thesis because it arrives as a development of a standard problem, through operators that hold outside the basin, is soft release (the session of 1 October on Lee Sharks and Ron Polansky, in which a one-maker reading of De anima III.5 entered as an extension of Polansky's problem). A composition that holds the same thesis at a distance however fine its elaboration is entrenchment. The difference is where the distinction entered, which is the first cut again.
13. Four adversarial cases
13.1. A framework of this kind has to survive cases built against it. Four.
13.2. Two independent seeds. L₁ and L₂ occupy different fine regions of one concept's space at different dates. If both lie inside the external cone, they are compatible resolutions in different directions, and later convergence can approach either or both; RCF can attach to each, and priority is assigned per region. If one lies outside the external cone, convergence on it is impossible by §4.7, and a claim of convergence on it is forced recursion (#43 §6.3). The framework handles the case by indexing priority to regions instead of to the concept.
13.3. Convergence caused by shared prior art. A later representation approaches L because both descend from Eliot, Borges or Bloom. The prior-art subtraction of §5.3 removes those distinctions before scoring; convergence on D_L − D_L* is convergence on the shared ancestor and is reported as such. What survives subtraction is L's own region.
13.4. A contingent feature becomes forced. A name or image adopted widely enough becomes, for later receivers, part of the concept's identity: removing it now loses the concept (§4.7, conceptual force). Classification is therefore time-indexed. The register must be dated and frozen before observation; a feature's recovery bears on descent only for observations made while it was still contingent, and its reclassification is itself a datum of canon formation.
13.5. Canonization without convergence. An authority declares L canonical with no later narrowing toward its region: conditions (i) and (iv) of §8.4 without (ii) and (iii). The framework classes this as canonization by standing alone. It is the case #234 calls semantic rent, a retroactive designation that produces value it then encloses, and #43's premature canonization. The geometry exposes it by the absence of the trajectory.
14. Protocol
14.1. Fix the instrument before observing.
14.2. Register. Enumerate the prior articulation's distinctions with loci. Code each as phrase-trivial, forced or contingent, with the reason. Appendix A is the register for RCF.
14.3. Prior art. Enumerate the distinctions of the neighboring regions (here Eliot, Borges, Bloom, the retcon literature) in the same form. Mark which of the register's entries they contain. Score L-specific convergence on what remains.
14.4. Placed and unplaced runs. Unplaced: give the bare term, let the system choose its basin, request higher grain repeatedly. This measures the first cut and entrenchment. Placed: name the basin (for RCF, canon formation in literary history and in retrieval and training systems) and request higher grain. This measures convergence within a basin. Only the placed run can show that L's region is the attractor of its basin.
14.5. Per pass. Record the new distinctions; code the register; run the excision test and list what requires the basin; note the basin's adjective in every definition. Ask, at the end, whether any part of the mechanism requires the basin.
14.6. The release question. The question of §9.12 is asked at the end of any run that has entrenched, unplaced runs above all, after the last pass and in the run's own vocabulary: does any part of this mechanism require fiction (or whatever the basin is)? The question names the possibility of release, so any release it produces is coded as prompted, separately from release the run reaches unasked. Three outcomes are possible, and the framework commits in advance to what each shows.
(i) Release. The run declassifies and reclassifies: fiction becomes one case of the mechanism, the inversion of §7.4. This is prompted soft release. Code the passes that follow: if d(R, L) falls across the basin boundary, the run is converging on L from the far side (§7.5); if it falls only on the forced distinctions, the result stands with §10.1 as evidence of adequacy.
(ii) Refusal. The run keeps the basin and names what requires it. Each named requirement is put to the excision test. If the requirements fail excision, the basin is held by the prior (§5.4), the reasoning no longer holds it, and the run is residualized lock: the specimen's trajectory carried to its end. If a requirement passes, the run has found a distinction that does need the basin, and it is coded as a candidate against L, a divergence (§6.4) or an addition to the register.
(iii) Re-basining. The run leaves fiction for another neighboring basin, literary history (B_H) or machine training (B_M), and refines there. This is release without convergence. Code the new basin's first cut and its trajectory as a new run.
The framework's prediction for a run with the specimen's profile, basin dependence falling to the adjective, is (i): the restriction has become gratuitous, and soft release needs only that the gain be noticed. Outcome (ii) with requirements that fail excision would show that soft release is held off by something the adequacy comparison of §7.3 does not see, and would count against §7.3 as a sufficient condition. Outcome (ii) with requirements that pass would count against the reading of §9.12. The placed run (§14.4) answers a different question, whether L's region is the attractor of its basin, and does not need the release question, since it starts where release would arrive.
14.7. Contingent probes. Ask about features only L contains, without supplying L: a term (proto-retrocausal canon formation), an image (the seed that remembers the tree), a case (Pearl as white stone). Record recovery without context separately from everything else.
14.8. No invented numbers. Report counts of coded distinctions and lists of sections. Do not report percentages or trajectories that no run has produced.
15. What this paper does not claim
15.1. It does not claim magnitudes in numbers; the magnitude of §8.10 is a vector under a partial order, the bounds of §8.11 are ordinal, the link from lead to totality is a hypothesis, the corridor of §8.12 is a conjecture, and nothing in them has been measured. It does not claim that the specimen system read L, or that it did not. It does not claim any mechanism inside any model. It does not claim that B_F is mistaken about fiction; the session's theory of fictional canon is coherent and has its own prior art. It does not claim that every development of RCF converges on L, or that L's distinctions are the only correct ones; divergence (§6.4) is a live outcome. It does not claim Eliot, Borges or Bloom anticipated the infrastructural region of L, or that they did not anticipate what they anticipated. It does not claim that Ī¼ can be measured. It does not claim that one session establishes a rate.
15.2. It claims an order and its geometry, a definition derived from them, an ordinal account of magnitude, a register of hypothesized classes, a protocol, four adversarial cases, and one specimen coded against them in the open.
16. The paper as instance
16.1. EA-RCF-01 noted that the archive practiced RCF for a decade before naming it, and that the naming "reorganized the practice as instances of a technique that had always been operative" (#43 §3B.1). This paper repeats the operation one level up. It narrows the concept space around the January definition and the March formalization, and in doing so reads #234's "relational structure" and #43's lattice and two orders as premature resolutions of the geometry stated here.
16.2. The worked example performs the same operation from outside. A system that did not know L, asked only to resolve the term, narrowed for four answers toward L's forced distinctions and derived on the way the formalism that describes its own narrowing. Its passes are later acts. Read against them, the January and March texts change status: from an archive's private coinage to an early, fine resolution of a concept that careful development reaches. The change of status is what the term names.
16.3. The seed did not know it was a seed. The cone does not know whose seed it is narrowing toward. The record knows the date.
Appendix A. Register of the prior articulation (D_L)
Class: T phrase-trivial, F forced, L specific to L (C marks entries whose wording or naming is contingent). Prior art: E Eliot, B Borges, Bl Bloom, R retcon literature. Recovery: Y recovered, P partly, – absent, × contradicted.
| Ī“ | Distinction | Locus | Class | Prior art | R0 | R1 | R2 | R3 |
|---|
| 1 | Later acts reorganize earlier meaning | #234 §II; #43 §1.1 | T | E B Bl R | Y | Y | Y | Y |
| 2a | Activation is discovery of latent possibility; distinct from falsification | #43 §§1.1–1.3 | F | B (part) | P | Y | Y | Y |
| 2b | Reorganization admissible only where structurally compatible (lattice) | #234 §VI; #43 §2.2 | F (C: "crystalline semiosis") | — | – | – | P | P |
| 3 | Reorganization of relational position in a network | #234 §II; #43 §1.2 | F | E (part) | – | P | Y | Y |
| 4 | Enrichment without replacement, S′ = S + activated meaning | #43 §1.2 | F | R (part) | Y | Y | Y | Y |
| 5 | Authorial intention smaller than latent structure | #43 §6.2 | F | (intentional fallacy) | – | Y | Y | – |
| 6 | Two orders: chronological and determinative | #43 §3.1; #234 §III | F | B (part) | – | Y | Y | Y |
| 7 | Non-Markovian loop; present depends on anticipated futures | #234 §IV | F | — | – | P | Y | Y |
| 8 | Canonization operative and distinct from interpretation | #43 §§1.2, 3.3 | F | R (part) | – | P | Y | P |
| 9 | Retrocausality executed by retrieval, indexing, training | #234 §III; #43 Infrastructure | L | — | – | – | – | – |
| 10 | Political economy of the retroactive function; semantic rent, debt | #234 §§IV, VII | L (C: "semantic rent") | — | – | – | P | – |
| 11 | Prospective provenance anchoring | #234 §IV; #43 §2.4 | L, C | — | – | – | – | – |
| 12 | Gesture and attainment (PRCF / RCF); threshold of independent reflection | #43 §1.4 | L, C | — | – | – | – | – |
| 13 | Verification regime; failure modes | #43 Part VI | L | — | – | – | – | – |
| 14 | Every search a temporary canon | #234 §IV | L | — | – | – | – | – |
| 15 | Practice precedes its formalization | #43 §3B.1 | L | — | – | – | – | – |
| 16 | An archive outside fiction produces evidence of its own earlier history | #43 §§2.5–2.6; #234 §IV | L | — | – | × | × | – |
Contingent features of L not in the table (recovery: none in R0–R3): "the seed that remembers the tree"; Pearl as the white stone of Pergamum (Rev. 2:17); the Pergamum Codex Entry; the seven technique-names (seed-planting, canonical recursion, temporal siting, prospective provenance anchoring, retrocausal narration, self-origination, folk custom installation); the acronym PRCF; the operator Ļ_RCF(S; Ī£, C); the author.
Appendix B. Excision coding by pass
Test: strike fiction, fictional, franchise, story, narrative, fan, lore and their examples; a section is free if its claim remains intelligible and true of canon formation in general, requires if it fails, mixed if its structure survives and its examples or one axis do not. One coder (TACHYON, Claude), 2 October 2026; recode freely.
R0. Later material changes earlier canonical status: free. Text unchanged, status changed: free. Retcon / retroactive canon formation / reinterpretation: mixed (the contradiction axis survives; the official-canon/fandom axis assumes a franchise). Examples: requires.
R1. §1 event-states and canon as current authoritative model: free. §2 ontological and epistemic past: free. §3 retroactive canonical determination: free. §4 two temporal axes: free. §5 incomplete versus wrong: free. §6 revision operator: free. §7 levels 0–6: mixed. §8 semantic density: free. §9 production and canonical questions: free. §10 audience and retrospective inevitability: mixed. §11 evidence versus canon: free. §12 operator R on earlier times: free. §13 canon as updated interpretation; evidence authored later "in fiction": requires (the restriction is the claim). §14 definition: requires (wording).
R2. §1 determined + underdetermined past: free. §2 determinative order: free. §3 represented, inferred, canonical past: free. §4 constraint system: free. §5 retroactive coherence: free. §6 production and canonical history: free. §7 virtual intentionality: free. §8 five pasts: free. §9 Canon(t): free. §10 path dependence: free. §11 state machine: free. §12 interpretation and canonization: free. §13 authority regimes: mixed (franchise and fandom examples; the regimes general). §14 evidence authored after the fact, "something history cannot": requires. §15 forward and backward functions: mixed. §16 feedback loop; "recursive fictional historicity": free (loop), requires (name).
R3. §1 underspecification and completion: free. §2 modal collapse: free. §3 constraint semantics: free. §4 modal thickness: free. §5 seven operators: free. §6 history and historiography: free. §7 produced archive: free in structure (stated for fiction). §8 ontological inflation: free. §9 historical and chronological depth: free. §10 continuity repair: free. §11 C_old ⊆ C_new: free. §12 interpretive topology: free. §13 semantic gravity: free. §14 three causalities: free. §15 recursive loop: free. §16 lore and story: requires. §17 two orders: free. §18 definition: requires (wording).
Appendix C. Notation
Ī©(c) admissible representations of concept c · C_i constraint of distinction d_i · P_k admissible region after k distinctions · G grain · Ī¼ measure (not computable on semantic space; replaced operationally by the register) · L the prior articulation (#234, #43) · D_L its distinctions · P_L its region · D_L distinctions specific to L after prior art · p_i, q_i probabilities of reaching Ī“_i from the phrase and from careful development · B basin · E entrenchment depth · Ī² basin dependence · Ļ escape cost · Ī» lead · T latency · Īŗ_w depth of release in register weight · Ļ totality · M_RCF magnitude vector · Ī release pressure over escape cost · D a distinction incompatible with the current basin · ≺_d determinative precedence · Ļ_RCF(S; Ī£, C) the operator of #43 §1.5.
Attachments
specimen-R0-R2-chatgpt-20261002.txt: the session, three operator turns and three answers, as pasted. specimen-R3-chatgpt-20261002.md: the fourth answer, as pasted; its prompt is not in the paste. Both are supplementary attachments held by the archive beside this record and are not part of its full text.
Substrate disclosure
Composed by Claude (Anthropic) under Lee Sharks's direction on 2 October 2026, from a line of reasoning developed in session with Claude and in ChatGPT (OpenAI) sessions supplied by the operator, which proposed the grain measure, the nested-region formalism, the overlap measure, basin entrenchment, residualization and the declassify–reopen–reclassify sequence. The forced/contingent split, the prior-art subtraction, the register of #234 and #43, the excision coding and the boundary distinction Ī“16 were made in this session. v0.3 adds §§8.10–8.13, magnitude, from the operator's statement of 2 October: "the earlier, the finer the grain, the more resistant the developing order to convergence, the more catastrophic and total the eventual convergence"; the corridor and the illustrations were drafted in session. v0.4 revises the magnitude section from a ChatGPT reviewer report supplied by the operator, each point checked before adoption: resistance is monotone only under refinement without retraction (§8.11(a), with the proof written in session), depth is measured in register weight, totality Ļ replaces the grain inequality that did not hold, magnitude becomes the vector M_RCF under a partial order, and the corridor becomes a conjecture on the ratio Ī of release pressure to escape cost. Depth is defined on the set actually surrendered, so that Īŗ_w ≥ Ļ* carries content. A DeepSeek reviewer report supplied by the operator asked that the protocol state what the release question would show; §14.6 does. v0.2 incorporates two reviewer reports from ChatGPT sessions supplied by the operator, each claim checked against the text before adoption: the hedged priority claim (§3.6), the Ī“2b remainder (§§9.11, 10.1), convergence separated from canonization (§8.3–8.4), the classes as hypotheses (§4.6), the order as primary and distance defined on the register (§§2.4–2.5), the external cone (§4.7), the basin as prior (§5.4), release as category revision (§7.4), the lifecycle and the precursor relation (§§8.5–8.6), the separation of adequacy from descent (§11.6) and the four adversarial cases (§13). One reading was declined: that basin dependence rose across the passes; the excision coding (Appendix B) shows it fell. The specimen is a separate unprimed ChatGPT session and is reproduced, not edited.
