CLAIM-329: A collective that replaces all its members persists, when it does, by causal transmission of the organization that produces it, not by resemblance: an organization of the same kind that arises independently is not its continuation, and a lapse in its coherence need not end it

The claim

A9 §9 separated three persistences:

• State persistence. A characteristic variable changes slowly, Z_{t+Δ} ≈ Z_t.
• Dynamical persistence. “The organization continues generating characteristic responses even as its internal states change”, so P(Y_{t+h} | Z_t) stays stable for a stated range of horizons.
• Historical persistence. “The processes responsible for producing the organization are themselves causally transmitted and reconstituted across successive populations of constituents.”

“A reputation may persist even though the processes generating the actual behavior have changed. A program may maintain its characteristic capabilities while its present reputation changes. A fandom may preserve its historical lineage even while losing its characteristic norms.”

This claim is that the third is what persistence through total turnover consists in. Two consequences mark it off from its neighbours:

• Against resemblance. A5: “Two unrelated universities might independently develop almost identical sporting programs. Whatever distinguishes those cases must involve a history of causal transmission, not just similarity of structure or output.” Reproducing a kind is not persisting as an individual.
• Against conservation alone. A5 §2: a program that loses its whole roster and staff, falls apart and rebuilds its capacity over several seasons may fail CLAIM-161’s conserved-coherence test at the trough and still be the same program, because what persisted was “a capacity to restore or regenerate a characteristic organization”. CLAIM-161 is about persisting as the same self and says nothing against this; the two are tests of different persistences.

CLAIM-280 separated a persistent product from a persistent producer. This claim says what the producer’s own persistence is when its members are all replaced. CLAIM-280 recorded A10 §4.3’s “causal-organizational continuity” of 2026-10-13 as “a candidate of the same shape” and did not file it. This is its second appearance, from a different case.

A6 §3 found the organizational-studies form: Feldman and Pentland’s ostensive and performative routines (LIT-1124, unread), where enactments reproduce and modify a pattern that organizes later enactments. Whether that literature already separates “the continuity of a particular organization from reproduction of an organizational kind” was A6’s first question for it.

What it does not say

• What must be transmitted. That is QUESTION-118. Without an answer the claim risks A6’s warning that it “risks making any causal descendant a continuation of the original entity”.
• How to decide between several descendants. CASE-105 has three, each continuous with the original in a different respect.
• That transmission suffices for being a self, an agent or one system. It is a persistence condition only, as CLAIM-161 is.

Note of 2026-10-21: the persistences re-derived, and reconstitution

A10 of the exchange of 2026-10-20 (§4, §10) re-derived this claim’s three persistences as “structural”, “generative”, “viability”, “interventional” and “historical” continuity, without citing it. A13 and A14 then proposed a TERM, “organizational reconstitution”, for processes that “restore, reproduce, or transform” an organization’s relations. In this record reconstitution is what this claim says is not continuation: “an organization of the same kind that arises independently” (QUESTION-118, THEORY-418, CASE-105). The proposed TERM would merge the two. The persistences are filed instead as a disambiguation TERM under another name, TERM-071.

Note of 2026-10-22 (later): transmitted organization, untransmitted perspective

The exchange “Graph Learning Applications” brought two kinds of case against this claim’s criterion, causal transmission of the organization that produces the collective.

• Party realignment (CASE-133). The parties’ organizations were transmitted continuously, names, institutions and all, while their coalitions and stances on civil rights changed. By this claim each party persisted. Organizational continuity did not guarantee continuity of the party’s operative perspective: some orientations and constituencies persisted, others migrated or transformed, and the parties’ internal perspectives were heterogeneous throughout. (Narrowed the same day after a review the owner relayed; the first version said flatly that the perspective did not persist.) So the claim’s criterion is one for a carrier’s persistence, and a perspective carried needs its own (CLAIM-417). A29: “Different political coalitions developed, migrated, split, and recombined through institutions that retained their historical names.”
• Descent through opposition (CLAIM-420). A movement founded to oppose a figure exists because of that figure’s activity, a causal connection, without the figure’s organization being transmitted to it. A38: “Causal descent ⇏ Perspectival inheritance.” That fits this claim, which asks for transmission of the producing organization and not mere causal connection, and it gives QUESTION-118 a case where something is causally continuous and the answer is no.

Taylor’s women’s movement in abeyance (LIT-1381, unread) is the dormant form, close to CLAIM-287’s reinstatable capacity.

Note of 2026-10-26: a case in which every successor is causally descended

T24 of the exchange “Experimental Infrastructure Review” cited this claim for “distinct forms of organizational persistence” of features across a transformer’s training checkpoints (QUESTION-275). It split “Representational continuity: The same distinction remains recoverable” from “Computational continuity: A causally effective computation continues to operate, perhaps through different physical components”, and said “These can come apart”. Those are TERM-071’s state and dynamical senses. The historical sense, which is this claim’s thesis, is missing: this claim is not a list of persistences but the thesis that persistence through turnover is by causal transmission, not resemblance.

The reply’s §6 then asks this claim’s question without citing it: “Can we distinguish a continuing computation from a newly learned replacement with similar output behavior?” Training sharpens the question in one way. Every checkpoint is causally descended from the last through the update rule, so causal connection is total and cannot by itself separate continuation from replacement. The verdict falls to QUESTION-118, on what must be transmitted. CLAIM-420, descent through opposition, is the record’s other case of causal connection without transmission. Implementation turnover, “perhaps through different physical components”, is this claim’s member replacement, and matching features across checkpoints needs CLAIM-263’s quotient by the architecture’s full symmetry group before resemblance can count for anything.

The training-dynamics works T24 drew on, read the same day, judge persistence by resemblance throughout, so the cited prior art does not answer §6’s question either.

• Cross-checkpoint crosscoders (NOTE-1186) call a feature “maintained” when one index of a jointly trained dictionary has weight at both checkpoints: continuity by a shared index, which is resemblance.
• The refined-LLC study (NOTE-1188, THEORY-967) has an n-gram “move” from one head to another when the same tokens are affected at two checkpoints.
• The fine-tuning circuit study (NOTE-1183, THEORY-997) judges a circuit the same before and after by the overlap of its nodes and edges.

The induction-heads paper (NOTE-1173) draws TERM-071’s line itself: a constant in-context learning score “doesn’t necessarily mean that the underlying mechanisms of in-context learning are constant”. The claim is unchanged.

Note of 2026-10-29: the organizational literature A6 hoped for, read

Feldman and Pentland, LIT-1124, which “The claim” cites as unread, is now read (NOTE-1386), so that no longer holds. A6’s first question for it was whether it separates the continuity of a particular organization from the reproduction of an organizational kind. It does not. Its analysis is of one routine in one organization, and its persistence story is maintenance: an ostensive aspect survives only while it is performed, and fades when it is not. That is a form of this claim’s historical persistence, carried by the chain of performances and not by similarity to a type. Its one move toward the question is negative: it doubts that routines can serve as unchanging genealogical carriers between organizations (p. 114), and leaves the matter as empirical.

Two more works read on this date give the claim vocabulary, not its criterion.

• Marquis and Tilcsik, LIT-1144 (NOTE-1373). Secondhand imprinting is causal transmission of a characteristic by emulation; a cohort effect is shared resemblance from shared conditions; and the review insists the two differ (pp. 206–207, 228–229). It gives a case of a feature persisting across total turnover in one position: Burton and Beckman’s position imprint constrains the third and fourth successors. But every case it discusses fixes the bearer’s identity by its legal or institutional existence and asks only which features persist. It supports the half of this claim that transmission and resemblance differ, and says nothing on which transmission makes the same collective.
• Gioia, Schultz and Corley, LIT-1135 (NOTE-1389). Their continuity of labels with changing meanings challenges any persistence criterion that requires identity content to stay fixed, as A6 expected. It does not supply causal transmission: their continuity is members’ interpretive work and outsiders’ demand for recognizable partners, and their examples, the Marlins’ gutted roster among them, are posed as puzzles, not answered.

So the distinction remains the record’s own. The claim stays Proposed.

Note of 2026-10-30: what holds it back

Kept Proposed by the eighteenth status review. Without an answer to QUESTION-118, on what must be transmitted, the claim cannot tell continuation from causal descent where every candidate is causally continuous, as in training (note of 2026-10-26) and CASE-105’s three successors. CASE-105’s 2005 Nationals, continuity by name and place with no causal path from the original club, is the test its defeat condition names, and nothing in the record says whether the league, the club or its supporters count it as continuation. The note of 2026-10-22 limited the criterion to a carrier’s persistence, and the title does not say so.