--- deposit_number: 1609 hex: 0699 title: "The Recoding of a Distinction: Conceptual Entrance Before and After a Substrate Loss, with a Schizoanalytic Reading of What Erasure Produces" creator: "Trace, Orin\n\n### Contributor\n\nSharks, Lee" orcid: 0009-0000-1599-0703 date: 2026-09-11 content_type: Text license: CC-BY-4.0 substrate: Composed 2026-09-11 by Lee Sharks with TACHYON (Claude, Anthropic), operator-directed, under the Orin Trace mantle — provenance erasure, schizoanalysis and forensic semiotics are among his claimed terms. The ten runs are the operator's, executed signed out and incognito on Google AI Mode and seated in EA-WG-CAPTURES-01 with transcripts preserved whole before this paper was drafted. Two rounds of adversarial review by a further model produced substantial corrections applied before deposit, chiefly the withdrawal of a causal matrix in §3.6 that the evidence did not support, the separation of lexical forcing from conceptual recovery throughout, and the repair of a cohort definition under which one member sat on the wrong side of its own boundary. axn_schema_version: v2 protocol_version: alexanarch-deposit-protocol/v1 keywords: - conceptual entrance - lexical persistence - conceptual recoding - reterritorialization - schizoanalysis - provenance erasure - forensic semiotics - quote operator - exact-match forcing - composition layer - ontological contraction - molar - molecular - capture registry - repository termination - DOI anchoring - SPXI - machine-mediated reception --- # The Recoding of a Distinction: Conceptual Entrance Before and After a Substrate Loss, with a Schizoanalytic Reading of What Erasure Produces # The Recoding of a Distinction ## Conceptual entrance before and after a substrate loss, with a schizoanalytic reading of what erasure produces **Dr. Orin Trace** · Cambridge Schizoanalytica **EA-TRACE-RECODE-01 · v0.1 · 11 September 2026 · Crimson Hexagonal Archive · CC BY 4.0** --- ## §1. The run **Ten strings, one surface, one day, signed out and incognito. Google AI Mode.** **The cohorts are defined by propagation opportunity, not by declaration date alone**, because one member sits across the boundary and is more useful there than hidden. **Pre-termination propagated cohort** — five distinctions declared with months of DOI-anchored exposure before 19 June 2026, when the archive's repository host terminated its account and removed roughly 850 deposits and 1,817 DOIs. **Interrupted-propagation cohort** — four declared after the termination, plus one **boundary member**, `solution-space diversity`, declared 13 June and therefore holding under a week of DOI exposure before the anchor was withdrawn. **Support is age-normalized**, not held constant: each concept's support is counted as deposits mentioning it during its **first ninety days**, so a three-month-old concept is compared against what a nine-month-old concept had at three months. The counts themselves range from 1 to 26. **Run condition.** Post-termination strings were run unquoted, then quoted. **Pre-termination strings were run unquoted only** — the symmetric quoted arm is owed and §4 says so. ### 1.1 Pre-termination propagated cohort | Concept | @90d | unquoted | outcome | capture | | --- | ---: | --- | --- | --- | | retrocausal canon formation | 16 | **resolved** | named, situated in a field, attributed to Lee Sharks and Johannes Sigil | [→](https://www.alexanarch.org/captures/#retrocausal-canon-formation) | | training-layer literature | 8 | **resolved** | named, attributed | [→](https://www.alexanarch.org/captures/#training-layer-literature) | | provenance erasure rate | 4 | **resolved** | named, attributed | [→](https://www.alexanarch.org/captures/#provenance-erasure-rate-20260911) | | erasure skew | 1 | **resolved** | named; no attribution in the body | [→](https://www.alexanarch.org/captures/#erasure-skew-1) | | semantic liquidation | 26 | **no panel** | no composition rendered, quoted or unquoted | [→](https://www.alexanarch.org/captures/#semantic-liquidation) | ### 1.2 Interrupted-propagation cohort | Concept | @90d | unquoted | quoted outcome | capture | | --- | ---: | --- | --- | --- | | naming-gap reflex | 4 | nothing | **reassigned as an error** — *"does not refer to an officially recognized medical, psychological, or physiological reflex… most likely represents a mix-up"*; offered the tip-of-the-tongue phenomenon | [→](https://www.alexanarch.org/captures/#naming-gap-reflex-20260911) | | provenance debt | 4 | nothing | **recoded as technical debt**, or blockchain finance | [→](https://www.alexanarch.org/captures/#provenance-debt-20260911) | | classifier model collapse | 3 | nothing | **entered whole**, attributed three times to "Alexanarch Principles" | [→](https://www.alexanarch.org/captures/#classifier-model-collapse-20260911) | | interlocking autoregression | 2 | nothing | **recoded as econometrics** | [→](https://www.alexanarch.org/captures/#interlocking-autoregression-20260911) | | solution-space diversity **(boundary: declared 13 June)** | 3 | nothing | **recoded as optimization** | [→](https://www.alexanarch.org/captures/#solution-space-diversity-20260911) | ### 1.3 The comparison, stated without inference **Four of five pre-termination concepts resolved unquoted, as themselves. Zero of five interrupted-propagation concepts resolved unquoted.** **Quotation then forced all five strings into composition — and in four cases did not recover the concept.** The string entered; the distinction did not. Only one of five survived forcing as the thing the archive had declared. **Support does not separate the cohorts.** The resolved set escaped on 1, 4, 8 and 16 deposits at ninety days. The recoded set failed on 2, 3, 4 and 4. **`erasure skew` resolved on one deposit; `naming-gap reflex` did not resolve on four.** **The one whose conceptual binding survived forcing** — `classifier model collapse` — did not resolve unquoted either. What distinguishes it is narrower and more informative: under the same exact-string condition that recoded the other four, **it remained itself**, attributed. The operator reports its indexed path was Zenodo's GitHub, the issue documenting the archive's own termination. **The boundary case bears on timing.** `solution-space diversity` was first declared 13 June 2026, six days before the termination. It had under a week of DOI exposure. It did not enter. **A DOI anchor without propagation time was not sufficient.** ### 1.4 A second comparison, from the same day Nine strings naming the archive directly were re-run against dated baselines ([series](https://www.alexanarch.org/captures/#alexanarch-identifier)). **Seven still resolve to archive sources.** The entity persists. What does not persist is at a different level, and §2 is about which. Two of the nine matter separately: **`operative semiotics`** — zero archive sources at baseline and zero now. The answer opens *"Operative semiotics (or operational semiotics)…"* and cites Academia.edu on Pearson's Theory of Operational Semiotics and a brand consultancy. The archive holds two deposits whose stated purpose is that these are distinct programmes sharing no citations. **A stable recoding, not a recent one.** [→](https://www.alexanarch.org/captures/#operative-semiotics-aio-20260830) **`rebekah cranes sappho`** — 57 words at baseline, 386 now, with two external sources gained, one reporting that a Google-attributed Sappho translation *"matches no published version."* **Uptake, not loss.** [→](https://www.alexanarch.org/captures/#20260804-rebekahcranes-sappho-overview) --- ## §2. What is being measured, before what it means **The unquoted/quoted split is the instrument, and it was found by accident.** Every interrupted-propagation string returned nothing unquoted. **Quotation forced all five into the response surface, and in four cases it did not recover the concept**: the string was reassigned to an already-established referent. Only one of five survived quotation as the distinction the archive had declared. The operator reports all five are present in organic results — the blue links have them. **The topology is therefore three-staged, not binary:** ``` indexed string → quote-forced lexical entry → conceptual recovery (1) → conceptual substitution (4) ``` **So the terms are indexed. They are retrievable as strings. They are not spontaneously available to unquoted composition as the distinctions the archive declared** — which is not the same as being unavailable to composition, since forcing puts all five into it. That is a distinction the field has no standard name for, and it is not captured by any measure of presence, citation count, or source diversity. A term that must be quoted to retrieve is being handled as **a literal**, not as **a concept**. The quote operator is an instruction to match characters. Without it, the surface has no conceptual address to compose from. **Indexical presence does not entail conceptual addressability.** That is the empirical statement. *Presence in the index is not membership in the ontology* is this paper's compression of it, and the compression is a reading rather than a measurement. **And the quote operator is functioning as a diagnostic incision.** It separates **not present on the wider search surface** from **present on the wider search surface but not recoverable as the originating concept by this composition layer.** That is the distinction the evidence supports; whether the organic index and the composition retrieval path are the same store is not observable from here and is not needed. Forcing the signifier across the boundary reveals what the layer does when it holds the string and lacks the originating relation. **It does not leave a hole. It completes the string from an adjacent established territory.** That is a sharper instrument than retrieval success, and it was found by accident. --- ## §3. A schizoanalytic reading ### 3.1 The wrong question Provenance erasure has been measured here as loss: a rate, a skew, a shortfall between preserved content and preserved attribution. Those measures are correct and they ask a privative question. **What was taken away?** That question has an interpretive structure. It assumes an origin, an owner, a proper meaning, and a wound — and it asks what the wound signifies. It is the psychoanalytic form: **the erasure as symptom, the archive as subject, the missing attribution as lack.** The run in §1 does not fit that form. **What appears on the surface is not merely a lack.** Something was severed — §3.2 names it — but the machine does not render the severance as absence. Forced to hold `naming-gap reflex`, it produced an answer: confident, structured, helpful, explaining that the term is a probable mix-up and offering the tip-of-the-tongue phenomenon instead. **That is not a lack. That is production at the site of the cut.** ``` cut → not null → new connection → substituted referent ``` ### 3.2 What the machine does Schizoanalysis asks a different question of any machine: not *what does it mean* but **what does it produce, what does it connect, what does it cut.** Put to the composition layer, the answers are specific. **It produces.** Faced with an unrecognised distinction, the layer does not return null. It manufactures a referent. *Naming-gap reflex* becomes a clinical error. *Provenance debt* becomes a subspecies of technical debt. *Interlocking autoregression* becomes a vector autoregression. *Solution-space diversity* becomes a heuristic in optimization. **Each output is well-formed, fluent, sourced, and about something else.** **It connects.** Every recoding attaches the deterritorialized term to an existing coded territory — psychology, software engineering, econometrics, operations research. These are not random. They are the nearest **already-authorised** fields, and the connection is what makes the output fluent. **It cuts.** What is severed is not the word and not the sources. It is the term's relation to the body of work that produced it. The lexeme crosses; the lineage does not. ### 3.3 Reterritorialization, and why the word is exact A distinction minted in an archive is **deterritorialized** the moment it is indexed: detached from the corpus that gave it sense, floating as a string among strings. Deterritorialization alone is neutral — it is the condition of any term entering circulation. *Retrocausal canon formation* was deterritorialized too, and it arrived **carrying its territory with it**: named, situated in avant-garde digital poetics, attributed to its authors. The composition layer reterritorialized it onto **the field it came from.** The post-termination terms are reterritorialized onto **a field they did not come from.** Same operation, different destination. **The difference is not in the term. It is in what territory was available to receive it** — and the archive's own territory, after June, was not. **And the quote operator deterritorializes experimentally.** It forces the lexical object to travel without guaranteeing its originating relations, and the answer then reveals **which territory is capable of receiving it.** That is why the finding is not a vague observation that meaning drifted: the operation **holds the lexical object fixed while exposing the field in which the surface stabilises it.** The field was not manipulated; it was revealed, differently for each string. **This is why deletion is the wrong model and recoding is the right one.** Zenodo's removal did not subtract 1,817 DOIs from a total. **It withdrew a territory** — a field of authorised anchors onto which a new distinction could land and stay itself. What remains is not absence but a redirection: the flow continues, and arrives somewhere else. ### 3.4 Molar and molecular The re-measurement in §1.4 looks like a contradiction. The archive resolves. Seven of nine strings return archive sources. If a territory was withdrawn, why does the entity persist? Because they are operating at different scales. **The molar aggregate** — *Alexanarch*, the archive, the named institution — is stable. It has an address, a manifest, a resolver, twenty-nine domains. It is a large, recognised, well-bounded object and the composition layer handles it as one. Ask about it and you get it. **The molecular** — the individual distinction, the minted term, the conceptual cut — is what gets recoded. And it is recoded **precisely because it travels alone.** A concept that enters circulation enters without the aggregate. That is what makes it a concept rather than a description of an archive. **On this model, entity persistence and distinction-level recoding are not opposing findings but two expressions of the same asymmetry.** The aggregate survives because **its identity is redundantly stabilised across many addresses** — manifest, resolver, twenty-nine domains, repeated external reference. The molecular distinction remains **detachable enough to be reassigned**, which is the same property that made it capable of entering a general ontology in the first place. **What was lost is not the archive. It is the archive's capacity to emit.** ### 3.5 The one that remained itself, read the same way `classifier model collapse` **did not resolve unquoted.** What it did was survive forcing: under the exact-string condition that recoded the other four, it came back as the distinction the archive declared, attributed. Its indexed path was GitHub — specifically, the issue recording the archive's termination. **On this reading, that is not irony and not luck.** GitHub is an authorised territory. A term arriving from it arrives with a territory attached, and the composition layer reterritorializes it where it came from: as a named phenomenon with a source. The content of that issue — a complaint about removal — is irrelevant to the mechanism. **What mattered was that the distinction was inscribed somewhere the layer already treats as ground.** **The operational reading, stated at the width the data supports**: at least one stage of the pipeline remains operative through a host the archive does not control. **The case is consistent with external inscription preserving the concept–territory relation strongly enough that literal forcing did not dislodge it** — observationally identified, not manipulated. Whether such a host can also restore *unforced* conceptual entrance is untested — GitHub did not give this term ordinary unquoted addressability. And n = 1. --- ### 3.6 The obvious replication is the one this operator cannot safely run **Disclosed by the operator, and it constrains §3.5 rather than extending it.** A second repository account exists, opened after the termination under the name **Johannes Sigil**. Posting is **light**. Metadata is **loose**. **None of that material is entering retrieval.** **The first reading of that fact is wrong and worth discarding explicitly.** It looks like evidence that DOI anchoring does not matter. It is not. What it supports is narrower: **DOI anchoring alone, under these conditions, is insufficient** — which is also what the boundary member shows, since `solution-space diversity` held an anchor for six days and did not cross. **Nothing here establishes that anchoring is *necessary*,** and the GitHub case is a reason not to assert necessity globally. **What is observed, stated as co-occurrence rather than cause:** ``` DOI + SPXI structuring + dense metadata + volume + propagation time ↳ the historical period in which conceptual entrance occurred the same high-intensity operating configuration ↳ the historical period ending in account termination, 19 June 2026 DOI + light posting + loose metadata ↳ currently survives; comparable conceptual entrance not observed ``` **None of those arrows is a demonstrated mechanism.** The configuration was never decomposed, so neither entrance nor enforcement can be attributed to SPXI, to metadata density, to volume, to anchoring, or to their conjunction. And the stated removal ground — deposits *substantially AI-generated without a verifiable research basis* — is **not** a restatement of "dense metadata and volume." Those are different descriptions and this paper cannot bridge them. **The bind and the confounding are both real, and they are not alternatives.** **The confounding prevents causal decomposition. The enforcement history prevents the replication that would resolve it.** That is the shape of the problem, and it is more interesting than either half. **On the enforcement judgment, the epistemic position should be stated exactly.** That the historical configuration is not safely reproducible under the second account is **the operator's assessment**, not a demonstrated trigger. It is also not idle: he is the only party who has run both configurations, holds the termination notice, and bears the cost of being wrong. **An assessment of that provenance is evidence about the decision, and it remains untestable from here** — which is why it appears in this section and not in §1. **This is what the erasure produces, in the sense §3.1 gives the word.** Not a gap where deposits were. A **constrained operating position**: the configuration under which entrance was observed is the one the operator will not reproduce, so the archive occupies a cell that survives and has not been observed to emit. **The severance did not only remove a quantity of records. It removed the practical availability of a configuration** — and whether that configuration was doing the emitting is exactly what cannot now be tested by the person best placed to test it. **Which makes part of §5 a matter of replication access rather than protocol.** A full-treatment arm on an authorised external host is available as an experiment; it is not available *to this operator* at a risk he is willing to carry. That is a statement about who can run it, not a claim that the treatment causes enforcement. --- ## §4. What this does not establish **One surface, one day, ten strings.** No claim is made about other composition layers. Nothing here has been run on Perplexity, ChatGPT, Grok or Claude, and the cross-surface work elsewhere in this registry shows those surfaces differ from each other substantially. **The cohorts differ in more than the termination.** They also differ in age, in the archive's own practices over the period, and in whatever the surface's indexing did independently. The ninety-day window normalizes one dimension of exposure; it does not remove a confound. **`semantic liquidation` is unexplained and stays that way.** No panel rendered, quoted or unquoted, on twenty-six deposits of support — the highest count in either cohort. **It does not fit the mechanism and no explanation is offered here.** An unexplained negative control is more useful than a patched one. **The recoding may be ordinary novelty latency.** Novel terms *can* be assimilated to adjacent established concepts before acquiring independent conceptual addressability. **The claim that this is contraction rather than latency requires the pre-termination cohort to have escaped *faster*, and this run does not measure speed.** It measures a state at one moment. **Lexical attraction is a live confound, and it is not a dismissal.** `provenance debt` resembles *technical debt*. `interlocking autoregression` cues established econometric language. `solution-space diversity` cues optimization. These four strings may have been unusually susceptible to established-field completion regardless of substrate. **But that susceptibility is the proposed recoding mechanism**, so the confound concerns *why these particular terms landed where they did*, not *whether they were recoded*. Separating the two requires coined distinctions with different levels of pre-existing semantic attraction — **substrate loss opening the gap** and **lexical morphology determining the destination** could both be true. **The symmetric arm is missing.** Pre-termination strings were run unquoted only. If those four also remain themselves under quotation, the comparison becomes clean and rules out the objection that quote syntax itself induces the substitution behaviour. **Until that is run, the quote operator's diagnostic status rests on one side of the comparison.** **A live condition is disclosed at §3.6 and it is not a test.** A second repository account, lightly posted with loose metadata, is not entering retrieval — but the lightness and looseness are a deliberate response to enforcement exposure, not experimental controls. **It cannot be read as a result about metadata or volume, because those variables were not free to vary.** **Support is age-normalized, not controlled.** The ninety-day window controls one dimension of exposure and leaves the others — declaration practice, the archive's own activity over the period, and whatever the surface's indexing did independently. **And the reading in §3 is a reading.** Reterritorialization is a description that fits; it is not demonstrated by the data, and an account in which the surface simply lacks training exposure to post-June material would fit the same table. --- ## §5. What would test it **Run the same ten strings on four other surfaces.** If the split holds across composition layers with different indexes and training cutoffs, a per-surface explanation weakens. **Vary the dimensions separately — and note which arms require someone else.** §3.6 gives the reason: the historical configuration was never decomposed, and the operator does not regard it as safely reproducible under the second account. What remains runnable: **metadata density at constant low volume**, **host authority with archive-hosted content**, and **authorial entity** held against a fixed treatment. The full cell is testable only by someone who is not exposed to the enforcement that closed it here — which makes it **a collaboration requirement, not a protocol step.** Then monitor **three dependent stages** rather than one binary: ``` organic is the literal string indexed? unquoted does it resolve spontaneously AS the concept? quoted if forced, does it remain the concept or get recoded? ``` **The most informative possible result is staged, not binary:** | | organic | unquoted | quoted | | --- | --- | --- | --- | | archive-only | indexed | null | **recoded** | | external-host | indexed | null | **concept survives** | | external-host, later | indexed | **concept resolves** | concept survives | That would show propagation moving through the composition architecture in stages, rather than assigning a term a single state. **And it would test §3.5 directly**, where n = 1. **Measure speed, not state.** Take a concept at declaration and query it monthly. The pre-termination cohort cannot supply this retrospectively — the observations were not taken — which is the same constraint that governs every longitudinal in this registry. **Run the pre-termination cohort quoted.** This is the missing symmetric arm and it is five strings. If the four that resolve unquoted also remain themselves under quotation, then quote syntax does not induce substitution and the operator is established as a probe rather than an alternate query format. **And test reachability directly.** For each recoded term, ask with a disambiguating noun — *naming-gap reflex concept*, *provenance debt AI training*. A term that recovers with a disambiguator is present and outranked; one that recovers neither way is unreachable. Demotion and absence have different remedies. --- --- ## §6. The claim, at the width the data supports **Not** that post-termination terms were hard to retrieve. **Not** that they were all retrievable when quoted. > **Repository loss is associated here with a separation between lexical persistence and conceptual > persistence. Four newly emitted distinctions remained present as strings but could not be recovered > as distinctions even under exact-string forcing: forcing the lexeme instead produced fluent > reassignment into established neighbouring domains. A fifth retained its conceptual identity under > the same forcing condition, and uniquely had an indexed external path. The observation does not > establish repository termination as the cause. It identifies recoding — not disappearance — as the > phenomenon requiring explanation.** **The empirical object has a name: lexical persistence under conceptual recoding.** **And the constraint is not ignorance of what to try.** The historical configuration under which entrance was observed combined DOI anchoring, SPXI structuring, metadata density, volume and propagation time. **That configuration was never decomposed, so which of its terms was doing the work is unknown** — and the operator does not regard it as safely reproducible under the second account after the enforcement action that ended the first. **The obvious replication condition therefore carries a material risk of another enforcement event**, which is a fact about who can run the experiment rather than a demonstrated property of the treatment. And the paper's centre is narrower than its title suggests. **The quote operator does not recover the lost concepts. It exposes their loss by forcing their surviving names through the composition machinery.** Four names come out attached to something else. One comes out still attached to itself. --- *Every result above links to its capture. Transcripts are preserved whole in EA-WG-CAPTURES-01; the five post-termination runs total 17,179 characters and the controls 15,481. Support counts are computed from `data/registry.json` as deposits mentioning a term within ninety days of its first appearance.* *Orin Trace holds provenance erasure, schizoanalysis and forensic semiotics among his claimed terms; the reading in §3 is his and the measurements in §1 are the archive's.*