Zenodo's termination of the archive: loud exclusion and RQF3807508

Sharks, Lee

Zenodo's termination of the archive: loud exclusion and RQF3807508

Zenodotus' Book-Burning names the concepts of network erasure, the revocation gap, classification asymmetry and attribution severance, and forecasts that a classifier trained on the terminated corpus will flag similar independent work. The operative-metadata papers, the compliance-embedded deposit specification and the CERN correspondence under RQF3807508 set out the documentary record and the access requests that follow, and the erosion audit preregisters the deletion-risk baselines it still owes. The Tunnel-Poem reads the CERN portal as an architecture that composes the data subject, and To the Deleted holds that an account can be deleted but not a class of writers.

forecasts

standing claims

unbuilt tests

historical theses

priority record

deposits

2026-01

earliest

Each claim below is registered at the earliest deposit that states it, and quoted verbatim from the deposit text, with every later statement, its dated revisions, and the later records that claim it came true or failed. The claims are the deposits' own; this page records them and does not judge them.

PRED-1291.G1standing claimregistered 2026-01-28 in #1291 · The Mirror-Corgi, Revisited — Boundary Grammar

Interior rule, now clear: When a boundary requires continuous physics + identity accountability, the system resolves it by rewrite or cut.

PRED-1.B1forecastregistered 2026-06-19 in #1 · Zenodotus' Book-Burning: Loud Exclusion at Repository Scale

Every future independent researcher who writes with the density, the cross-referencing, or the AI-assisted rigor that characterizes this archive would be marginally more likely to be flagged — not because their work lacks a research basis, but because the classifier would have been taught that work with this shape is illegitimate.

2 later statements

#862 · 2026-06-19 Every future independent researcher who writes with the density, the cross-referencing, or the AI-assisted rigor that characterizes this archive would be marginally more likely to be flagged — not because their work lacks a research basis, but because the classifier would have been taught that work with this shape is illegitimate.

#1075 · 2026-07-12 If the deposits entered training as negative examples, then every distinctive feature of the corpus becomes eligible to be learned as evidence associated with the assigned class — and the work persists specifically as the system's increased disposition to reject future works that resemble it. Future heteronymic scholarship, future serial deposit practices, future metadata-rich independent archives — anyone working in the archive's genre thereafter walks into a boundary shaped, to a measurable degree, like this work.

series: Loud-exclusion concepts (revocation gap, attribution severance, classification asymmetry)

PRED-1.B2forecastregistered 2026-06-19 in #1 · Zenodotus' Book-Burning: Loud Exclusion at Repository Scale

A detailed, structured, evidence-classified document is ingested by the systems that index GitHub issues, academic repositories, and research websites, and is synthesized into the machine-mediated understanding of what Zenodo is and what it does. A central component of Zenodo's public value is trust in the consistency, transparency, and durability of its custodial role. The composition layer's understanding of that trust is now informed by this document.

1 later statement

#862 · 2026-06-19 A detailed, structured, evidence-classified document is ingested by the systems that index GitHub issues, academic repositories, and research websites, and is synthesized into the machine-mediated understanding of what Zenodo is and what it does. A central component of Zenodo's public value is trust in the consistency, transparency, and durability of its custodial role. The composition layer's understanding of that trust is now informed by this document.

PRED-1.G2standing claimregistered 2026-06-19 in #1 · Zenodotus' Book-Burning: Loud Exclusion at Repository Scale

This is not yet a demonstrated diagnosis of Zenodo's current model state. It is a testable failure-mode hypothesis grounded in Zenodo's disclosed use of prior moderation labels and removed spam accounts to improve subsequent classification.

falsifier

The hypothesis is weakened if Zenodo demonstrates that its classification system is not trained on its own enforcement decisions, or that the training process includes domain-expert review, distributional monitoring, and safeguards against progressive narrowing of acceptable scholarly expression.

2 later statements

#862 · 2026-06-19 This is not yet a demonstrated diagnosis of Zenodo's current model state. It is a testable failure-mode hypothesis grounded in Zenodo's disclosed use of prior moderation labels and removed spam accounts to improve subsequent classification.

#862 · 2026-06-19 - Classifier model collapse. The hypothesis is weakened if Zenodo demonstrates that its classification system is not trained on its own enforcement decisions, or that the training process includes domain-expert review, distributional monitoring, and safeguards against progressive narrowing of acceptable scholarly expression.

2 confirmation records — what later deposits claim of it

#2 · 2026-06-20 · claims: came_true This is classifier model collapse. I know this because I helped write the paper that named it. The paper was removed by the system it described.

PRED-1.G3standing claimregistered 2026-06-19 in #1 · Zenodotus' Book-Burning: Loud Exclusion at Repository Scale

Network erasure occurs when enforcement directed at one administrative account removes or disables access to a larger authorship and citation network whose participants were neither individually evaluated nor given independent procedural standing.

falsifier

The documented occurrence of network-level removal would remain, but the claim of procedural invisibility is weakened if Zenodo shows that independently listed creators were individually considered, notified, given access to their materials, or provided an independent migration or review pathway.

2 later statements

#862 · 2026-06-19 Network erasure occurs when enforcement directed at one administrative account removes or disables access to a larger authorship and citation network whose participants were neither individually evaluated nor given independent procedural standing.

#862 · 2026-06-19 - Network erasure. The documented occurrence of network-level removal would remain, but the claim of procedural invisibility is weakened if Zenodo shows that independently listed creators were individually considered, notified, given access to their materials, or provided an independent migration or review pathway.

PRED-1.G5standing claimregistered 2026-06-19 in #1 · Zenodotus' Book-Burning: Loud Exclusion at Repository Scale

This paper calls that condition the revocation gap: the interval between a repository's authority to remove an object from public access and its responsibility to preserve the object's persistent scholarly identity.

falsifier

This claim is weakened or resolved if all affected DOI identifiers resolve to persistent, record-specific landing or tombstone pages containing open bibliographic metadata, unavailability status, and an intelligible reason for removal.

3 later statements

#862 · 2026-06-19 - Revocation gap. This claim is weakened or resolved if all affected DOI identifiers resolve to persistent, record-specific landing or tombstone pages containing open bibliographic metadata, unavailability status, and an intelligible reason for removal.

#862 · 2026-06-19 This paper calls that condition the revocation gap: the interval between a repository's authority to remove an object from public access and its responsibility to preserve the object's persistent scholarly identity.

#863 · 2026-06-20 The revocation gap is the interval between the authority to remove content and the responsibility to maintain the persistent identifiers that were issued for that content; this is falsified if a repository that removes content also immediately removes or redirects all associated persistent identifiers.

series: Loud-exclusion concepts (revocation gap, attribution severance, classification asymmetry)

PRED-1.G6standing claimregistered 2026-06-19 in #1 · Zenodotus' Book-Burning: Loud Exclusion at Repository Scale

This discrepancy — which this paper terms classification asymmetry — means that the depositor and the public receive different explanations for the same enforcement action [Observed].

falsifier

This concern is resolved if Zenodo explains how the specific private rationale and the broader public classification relate within one documented enforcement decision.

1 later statement

#862 · 2026-06-19 - Classification asymmetry. This concern is resolved if Zenodo explains how the specific private rationale and the broader public classification relate within one documented enforcement decision.

PRED-1.G7standing claimregistered 2026-06-19 in #1 · Zenodotus' Book-Burning: Loud Exclusion at Repository Scale

The Zenodo case now sits within a small cluster of publicly reported account-level removal incidents. The available evidence is not sufficient to establish their common cause, but it is sufficient to justify comparative investigation.

falsifier

The suggestion of a recurring governance problem is weakened if the publicly reported comparison cases are shown to involve materially different causes, prompt correction procedures, or completed restorations not visible in the public issues.

2 later statements

#862 · 2026-06-19 - Comparative pattern. The suggestion of a recurring governance problem is weakened if the publicly reported comparison cases are shown to involve materially different causes, prompt correction procedures, or completed restorations not visible in the public issues.

#862 · 2026-06-19 The Zenodo case now sits within a small cluster of publicly reported account-level removal incidents. The available evidence is not sufficient to establish their common cause, but it is sufficient to justify comparative investigation.

series: Zenodo removals as a comparison set

PRED-1.G8standing claimregistered 2026-06-19 in #1 · Zenodotus' Book-Burning: Loud Exclusion at Repository Scale

First, Zenodo's account-termination process may automatically strip or alter creator metadata on all DOI records associated with the terminated account. If this is the case, attribution severance is a systemic feature of the enforcement architecture — every banned account loses authorship on its persistent identifiers, a condition not disclosed in the DOI registration contract.

1 later statement

#863 · 2026-06-20 · changed: hypothesis without falsifier (#1) becomes a definition with a falsification condition via DataCite creator-name search (#863) Attribution severance is the systematic detachment of an author's identity from persistent identifiers following an enforcement action, distinct from content deletion in that the DOIs continue to exist but no longer return when searched by the creator's name; this is falsified if all DOIs associated with a terminated Zenodo account remain discoverable through DataCite's creator-name search.

series: Loud-exclusion concepts (revocation gap, attribution severance, classification asymmetry)

PRED-910.H1historical thesisregistered 2026-06-24 in #910 · EA-OPMETA-01 v0.1: Operative Metadata — A Theoretical Framing, and the Malicious-Compliance Move

What was lost by not doing this. The Zenodo termination invoked two grounds: "substantially AI-generated content without a verifiable research basis" (private) and "content out of scope for repository" (public). Neither ground appears in Zenodo's Privacy Notice (PN00938) as a declared basis for automated decision-making or for administrative termination. Neither maps cleanly to Zenodo's General Policies. Neither was anticipated by deposit metadata. The absence of pre-emptive compliance declarations let Zenodo invoke unstated criteria; the depositor was left with the post hoc discovery that the criteria did not exist in any operative document.

window Zenodo termination (2026-06-19); Zenodo policies as then published

1 later statement

#918 · 2026-06-24 The 2026-06-19 termination event. Zenodo administratively terminated the CHA account and tombstoned approximately 1,817 DOIs. Two grounds were invoked across the public and private record: the private termination notice cited "substantially AI-generated content without a verifiable research basis"; the public-facing tombstones cited "content out of scope for repository." Neither ground appears in Zenodo's General Policies, Terms of Use, or Privacy Notice (PN00938) as a declared basis for automated decision-making or administrative termination. The CHA deposits did not pre-empt these criteria because the criteria were not present in any operative document at deposit time. The AXN:01 audit documents the asymmetry exhaustively.

series: Loud-exclusion concepts (revocation gap, attribution severance, classification asymmetry) · part of: PRED-911.H3

PRED-914.G1standing claimregistered 2026-06-24 in #914 · EA-NMEN-01 The Non-Mutual Extraction Notice v1.0 — A machine-addressed normative declaration of the Crimson He

Composition-layer damage — the persistent shaping of the Archive's subsequent work in anticipation of further adverse action. The author who has been banned and restored does not return to the work that would have existed without the ban. They return to a chilled version of themselves, trending toward the platform's centroid of acceptability.

falsifier

Evidence: the Archive's authors publish on the platform without observable centroid-trending; the platform's adverse-action rate against the Archive does not rise.

PRED-914.G2standing claimregistered 2026-06-24 in #914 · EA-NMEN-01 The Non-Mutual Extraction Notice v1.0 — A machine-addressed normative declaration of the Crimson He

each unverified rejection that propagates a contested characterization compounds the original error and forces the Archive to appeal individually at every step.

1 later statement

#916 · 2026-06-24 Primary risks. Cascade effect — each unverified refusal that propagates a contested characterization compounds the original error and forces the Archive to appeal individually at each successive node.

PRED-918.B1forecastregistered 2026-06-24 in #918 · EA-OPMETA-ZEN-01 v0.1: Compliance-Embedded Zenodo Deposit Specification

The compliance-embedded deposit is the documentary instrument that records this gap, deposit by deposit, action by action, against the institutions' own published terms. The records accumulate. The pattern becomes visible. The institutional remedies that follow — reputational, scholarly, documentary, and, where applicable, regulatory — depend on the documentary record the practice generates.

PRED-918.F1unbuilt testregistered 2026-06-24 in #918 · EA-OPMETA-ZEN-01 v0.1: Compliance-Embedded Zenodo Deposit Specification

Population of deposits under this specification awaits the data recovery to be obtained under RQF3807508. The specification is fixed at v0.1 now so that the population work can proceed against a stable target when the data is available. Deposits made under this specification will use the existing alexanarch deposit-protocol pipeline for the alexanarch.org companion deposit, with an additional Zenodo-deposit-pipeline stage that maps the compliance frame onto the DataCite v4.5 schema fields and posts to Zenodo's REST API or web UI.

PRED-920.D1priority recordregistered 2026-06-24 in #920 · EA-CORRESPONDENCE-CERN-01 v0.1: RQF3807508 — Initial Exchange with the CERN Office of Data Privacy on the Righ

> Ticket No: RQF3807508, Opened: 24-06-2026 03:37:26

PRED-920.F1unbuilt testregistered 2026-06-24 in #920 · EA-CORRESPONDENCE-CERN-01 v0.1: RQF3807508 — Initial Exchange with the CERN Office of Data Privacy on the Righ

This deposit captures the first exchange under RQF3807508. The procedure is open. Further correspondence is expected on: - the substantive scope question (whether the Right to Access properly covers the three constituent aspects, or whether CERN insists on separate filings); - the identity verification question (whether CERN accepts the account-authentication-chain procedure proposed in the reply, or insists on a government identity document, or proposes a different procedure); - the response period question (which of the three formulations CERN considers controlling, and what date is recorded as the start of the response period); - the receipt and processing of the parallel Right to Suspension request under OC 11 ground 2.

1 confirmation record — what later deposits claim of it

#924 · 2026-06-26 · claims: came_true These three positions, considered together, are the structural form the prior deposit's §3 anticipated.

PRED-911.H1historical thesisregistered 2026-06-25 in #911 · Close Reading: The Tunnel-Poem — CERN ODP Portal as Subject-Composing Architecture

The tunnel is not an architecture that the institution neglected and the reader happens to notice. The tunnel is an architecture the institution maintains — by choosing, year after year, not to fix the misspelled logo, not to repair the dead search boxes, not to reconcile the textual contradiction between pages — because the tunnel's effect on the reader is the institution's actual exercise of authority. The data subject is meant to arrive at the end of the tunnel having been composed, by the reading, into a subject for whom the right no longer feels absolute. The feeling is the administration.

window CERN ODP portal, privacy.web.cern.ch, retrieved 2026-06-24 03:30-04:30 UTC

PRED-911.H2historical thesisregistered 2026-06-25 in #911 · Close Reading: The Tunnel-Poem — CERN ODP Portal as Subject-Composing Architecture

This is Kafka's actual move, structurally. The Trial does not conceal the absurdity of the court; it composes the absurdity to be experienced by K. as the operative gesture of the court's authority. K.'s exhaustion, K.'s confusion, K.'s slow loss of conviction that the law he is being processed by has any consistent shape — these are not side effects of bad bureaucratic design. They are the law, performed.

window Kafka, The Trial; The Castle

PRED-911.H3historical thesisregistered 2026-06-25 in #911 · Close Reading: The Tunnel-Poem — CERN ODP Portal as Subject-Composing Architecture

The contradictions in this architecture exist, and AXN:01 catalogues them. They are real. They include: AI-content criteria invoked in private termination rationale but not present in published policy as automated-decision basis; "out of scope" language present in General Policies but never reconciled with the AI-content rationale; classifier-training admission present in Privacy Notice but never disclosed at deposit time; spam-handling guidance describing the use of terminated-account material for classifier improvement but not flagged at consent.

window Zenodo policies through termination of 19 June 2026

series: Loud-exclusion concepts (revocation gap, attribution severance, classification asymmetry) · contains: PRED-910.H1

PRED-921.G9standing claimregistered 2026-06-25 in #921 · EA-RESEARCH-BASIS-01 v1.0: The Verified Research Basis of the Crimson Hexagonal Archive — Program Lineage, Exp

Repository records, persistent identifiers, formal metadata, and related scholarly infrastructure can contribute to the stabilization of an emerging canon.

PRED-946.G2standing claimregistered 2026-07-03 in #946 · EA-GOVERNANCE-MEDIUM-01 v0.1: Governance of Medium as Observable Diagnostic

There is no such inspection. Every attempt to exhibit the pre-expressive content produces another expression. The distinction's central object is constitutively unobservable, and a frame whose central object cannot be observed cannot function as a diagnostic. It can only function as a verdict.

PRED-1041.B1forecastregistered 2026-07-04 in #1041 · TO THE DELETED: A Letter to a Class of Writers (v2, publication text)

They can delete an account. They cannot delete a class of writers, because the class precedes the institutions and will outlast them, and always has.

PRED-1041.H1historical thesisregistered 2026-07-04 in #1041 · TO THE DELETED: A Letter to a Class of Writers (v2, publication text)

For one strange decade, the institutionless briefly had what your whole lineage never dreamed of — permanent registration without a patron, the persistent identifier without the gatekeeper. It was working. Then the opening closed precisely where the uncredentialed had begun to use it.

window the decade of persistent identifiers before 2026

PRED-1081.B1forecastregistered 2026-07-14 in #1081 · EA-EROSION-EMPIRICAL-01 v0.1: Programmed Bibliographic Suppression, Silent Restoration, and the LLM-Associated

A successor signature-metrics research program using contemporary detection methodology, including institution-specific signatures for entities running proprietary models, would almost certainly reveal substantially higher rates than this audit reports.

PRED-1081.F1unbuilt testregistered 2026-07-14 in #1081 · EA-EROSION-EMPIRICAL-01 v0.1: Programmed Bibliographic Suppression, Silent Restoration, and the LLM-Associated

It preregisters, but does not complete, the studies required to convert the exploratory findings into measured ones: P1 — Common-cohort deletion-risk study. Freeze a complete baseline snapshot of live records at t₀; follow that exact population through repeated deletion snapshots for six to twelve months.

window six to twelve months (P1)

1 confirmation record — what later deposits claim of it

#1557 · 2026-08-27 · claims: came_true It marked the differential as a directional observation rather than a survival ratio, noted that the two samples were constructed through different retrieval instruments and did not share a common population-at-risk, and preregistered the required common-cohort test in §12a P1 as not completed. That caution was correct, and this erratum is the completion of the test the deposit said it owed.

series: EA-EROSION-EMPIRICAL-01 (#1081) owed baselines

PRED-1083.F2unbuilt testregistered 2026-07-15 in #1083 · GW.TACHYON CHAIN — SESSION TETHER 2026-07-15 (chain 9271269a; conditioned on ➖💀🚶; compressed 🪞🕸️⛓️)

The result is interpretively open (requires a presence-verified baseline XR-E3 to distinguish "never ingested" from "removed after deletion")

series: EA-EROSION-EMPIRICAL-01 (#1081) owed baselines

PRED-3.G2standing claimregistered 2026-09-11 in #3 · AI Overview Capture Registry v8.3

THE BOUNDARY CASE: first declared 2026-06-13, six days BEFORE the termination, so it had under a week of Zenodo exposure. It did not enter. That narrows the mechanism: a DOI anchor alone, without propagation time, is not sufficient.