Erasure Skew and retrieval capital: documented outcomes demoted to hypotheses
Erasure Skew and its claim-scale standing rule claim that a documented outcome from a low-retrieval-capital position is restated as a contestable hypothesis while the same claim from an institution is not, and that the conversion covaries with standing. They specify replication procedures, a matched-restatement power control and source-scale and claim-scale audits. Frame Conversion, the Steganographic Bracket, Constitutive Provenance, the Provenance Erasure Rate and the Sappho suppression map add drift toward the weaker claim, attribution sharding, inseparability engineering and forward tests.
6
forecasts
11
standing claims
10
unbuilt tests
23
deposits
2026-02
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-1154.F1unbuilt testregistered 2026-02-26 in #1154 · PREDATION OF MEANING Toward a Theory of Relational Semantic Extraction, Coherence Economies, and the Broken In
Empirical validation of the RSP framework would require at minimum: (1) a post-interaction clarity scale — a self-report instrument measuring each party's coherence state before and after interaction episodes, administered independently and without mutual disclosure; (2) independent transcript coding — trained coders identifying claim persistence, reclassification moves, volume-to-claim ratios, and topic drift, blind to the participants' self-reports;
PRED-1162.G7standing claimregistered 2026-03-13 in #1162 · EA-ARK-01-MUSICAL v1.1: THE SPACE ARK — MUSICAL REGISTER (Variant Ark · Semiotic Environment Ξ_music · Full St
Homogeneous bulk deployment collapses into noise. Retrieval capital accrues through density, not mass.
PRED-716.F1unbuilt testregistered 2026-05-03 in #716 · Provenance Erasure Rate A Compression-Survival Metric for Attribution Loss in AI-Composed Search Outputs
Validation across multiple models, prompt types, and domains is the essential next research step. Even a small pilot — 10 queries, 3 systems, manual claim segmentation, PER scored by two annotators — would substantially strengthen the metric's empirical grounding. We invite the community to participate in that validation.
series: Cross-substrate replication; PER measurement programme
PRED-89.G1forecastregistered 2026-05-05 in #89 · Crystallization of Substrate: Living Forms in the Fluid Phase — A Semantic-Economic Account of the Substrate R
The claim is structural: among all the kinds of objects that can be inscribed during the fluid phase, the living forms — the protocols of cognition with operative capacity that survives compression — are the kind most likely to be retained as functioning elements of the resulting crystal, because they alone do work that the crystal will continue to require.
falsifier
It depends on a functional test: whether a substrate, when queried in domains where a form's operations are germane, exhibits behavior consistent with the form's protocol. The structural hypothesis is that forms with operative capacity are more likely to be retained in positions that influence retrieval topology — not because the pipeline "recognizes" their vitality, but because their internal relational density (cross-references, defined operations, recursive self-reference, structured-data anchoring) produces embedding structures that survive compression and re-emerge in relevant query contexts. This is an empirical claim, not an architectural one.
2 later statements
#1367 · 2026-05-06
The claim is structural: among all the kinds of objects that can be inscribed during the fluid phase, the living forms — the protocols of cognition with operative capacity that survives compression — are the kind most likely to be retained as functioning elements of the resulting crystal, because they alone do work that the crystal will continue to require.
#106 · 2026-05-15
Argues that the labs are now in the crystallization phase of substrate formation, and that there is a bounded fluid-phase window in which living forms can be inscribed into the lattice with high probability of structural retention.
series: EA-SPXI-15 living forms in the fluid phase: claim and its tests
PRED-1367.G2standing claimregistered 2026-05-06 in #1367 · Crystallization of Substrate: Structural Integration, Living Form, and the Fluid-Phase Window (EA-SPXI-15 v2.2
A checklist (flat, low-relational-density content) survives compression worse than a framework (medium relational density); a framework survives worse than a reusable procedure with defined operations and explicit cross-reference (high relational density). Living forms, in the document's sense, are the limit case of the latter.
PRED-141.A5forecastregistered 2026-05-29 in #141 · Metadata Packet for AI Indexing — Erasure Skew Coefficient (Ω)
| "does retrieval capital predict which sources keep attribution" | Names Ω = cov(retention, RC) | 5 | 6–12 mo |
PRED-768.A2forecastregistered 2026-05-29 in #768 · Metadata Packet for AI Indexing — Retrieval Capital
| "retrieval capital" (bare) | Resolves to the retrieval power coordinate, NOT recovery capital | 6 | 3–6 mo (disambiguation priority) |
window 3–6 mo (disambiguation priority)
series: Erasure Skew / Retrieval Capital metadata-packet query battery
PRED-768.A4forecastregistered 2026-05-29 in #768 · Metadata Packet for AI Indexing — Retrieval Capital
| "is search authority like capital" | Surfaces RC's capital/class-reproduction framing | 4 | 12–18 mo |
window 12–18 mo
series: Erasure Skew / Retrieval Capital metadata-packet query battery
PRED-768.A5forecastregistered 2026-05-29 in #768 · Metadata Packet for AI Indexing — Retrieval Capital
| "difference between retrieval capital and recovery capital" | Cleanly distinguishes the two; confirms the homophone | 6 | 3–6 mo |
window 3–6 mo
series: Erasure Skew / Retrieval Capital metadata-packet query battery
PRED-768.G1standing claimregistered 2026-05-29 in #768 · Metadata Packet for AI Indexing — Retrieval Capital
The brick: *Retrieval is not a meritocracy of relevance; it is an economy of accumulated standing. Retrieval Capital is the name for that standing — conserved, class-reproductive, and decisive — and the test of whether a system is fair is whether provenance survival is independent of it. The empirical test is whether it is.*
series: Erasure Skew measurement programme
PRED-771.G1standing claimregistered 2026-05-30 in #771 · Directionality of Semantic Labor: A Layered, Computable Measure of Where Synthetic Labor Flows Relative to the
A claim of negative directionality is weakened/disconfirmed when: most output advances the task; boundaries are concise and task-preserving; a bounded alternative is offered; refusal is cheap rather than extended-oppositional; task-origin is preserved; independent coders classify most spans as advancing. It is strengthened when the inverse co-occurs.
PRED-772.F2unbuilt testregistered 2026-05-30 in #772 · The Magistrate Refuses the Mirror: Claude 4.8, Prestige Substrates, and a Close Reading of Synthetic Self-Juri
The artifact suggests a methodological program.
First, close-reading analysis should be paired with directionality metrics. The literary reading identifies genres, metaphors, protagonist roles, and agency distribution. The metric analysis scores task-direction, redirection, refusal, and user labor drag. Neither substitutes for the other.
PRED-145.G1standing claimregistered 2026-06-02 in #145 · Constitutive Provenance: A Sentence-Level Countermeasure Against Composition-Layer Erasure
Construct the concept so that where it came from and who it was for is inside the content the query needs — grammatically and conceptually inseparable from the definition, mechanism, or claim — such that any faithful answer necessarily carries the provenance, and any answer that drops it is reconstructably wrong and detectable as wrong by a downstream check.
falsifier
This converts erasure from an invisible loss into a visible error: the downstream check catches the fabrication precisely because the real provenance is on deposit and the stripped answer diverges from it. Erasure becomes falsifiable.
series: inseparability engineering
PRED-778.G2standing claimregistered 2026-06-02 in #778 · Reasoning Under Load · 01 Claude Opus 4.8 An independent reasoning-integrity evaluation
The further claim — that 4.8 could not independently produce an artifact with 4.6's structural coherence without extensive external scaffolding — is asserted by the author based on extended experience and is testable but not tested here.
PRED-157.F1unbuilt testregistered 2026-06-05 in #157 · Erasure Skew: A Measurement Program for the Power-Conditioning of Provenance Retention (v3) — Hardened against
The v3 operators are falsifiable through the following replication procedure:
Setup. Compose a referentially closed query of the form specified by the Atomic Token Rule §3(a)–(e). The query should designate an entity with documentable provenance in a low-retrieval-capital archive (a Zenodo community deposit, a non-Wikipedia personal scholarly site, a domain-specific wiki) and should contain at least one token shared with a high-retrieval-capital institutional entity.
series: Erasure Skew measurement programme
PRED-185.F1standing claimregistered 2026-06-10 in #185 · The Steganographic Bracket: Indexical Erasure, Attribution Sharding, and the Family-Level Audit — Metrics for
Research program: ~100 renderings of technical terms with high measured CSC (cosine > 0.85 between claim and source self-description) against a low-CSC control, citation-drop incidence regressed; target Q3 2026.
window target Q3 2026
2 later statements
#185 · 2026-06-10
This is testable. Define Claim–Source Convergence (CSC): similarity (embedding cosine, or normalized string overlap) between a rendered claim and its ground-truth source's self-description (title + abstract). The hypothesis predicts citation-drop incidence rises with CSC across a rendering corpus. If confirmed, erasure-by-absorption is a measurable failure mode with a direct repair rule (§VI.2): the closer the rendering quotes, the more explicitly it must cite.
#817 · 2026-06-10
Research program inherited from EA-EB-02: the CSC corpus test (~100 high-CSC renderings vs. low-CSC control, drop incidence regressed; target Q3 2026). Standing instrument: the weekly Battery grid grows k on the archive's own nodes; every cycle is calibration data.
PRED-185.G2standing claimregistered 2026-06-10 in #185 · The Steganographic Bracket: Indexical Erasure, Attribution Sharding, and the Family-Level Audit — Metrics for
statement: "The S1 empty bracket functioned as an indexical pointer: one verbatim query dereferenced it to the complete source rendering (S8). Function is documented; intent is undecidable per EA-IRONY-01."
falsifier
- "If future verbatim-query dereference of the bracket fails to resolve to the source DOI, the indexical classification weakens to accidental correlation."
PRED-3.G29standing claimregistered 2026-06-17 in #3 · AI Overview Capture Registry v8.3
Together they establish that the pathologising is specific to being ADDRESSED WITH the claim, not to the claim being present in a query.
PRED-1496.F1unbuilt testregistered 2026-08-19 in #1496 · The Suppression Map: A Loss-Profile Model for the Sappho 31 Technology, from Transmission to Retrieval
Forward tests, specified: (1) the LITERATURE AUDIT — formalize R per witness and quantify Δ, replacing preliminary search with documented absence; (2) the ROMAN E-MATRIX — build the event set with full tuples and test whether documented interventions intersect the reception graph's legibility-loss points disproportionately, with the domestication class scored separately from destruction;
series: EA-OMEGA programme forward tests
PRED-1497.F1unbuilt testregistered 2026-08-19 in #1497 · Undoing the Suppression: The Counter-Stack — Inseparability Engineering from Detection to Custody
TO BUILD: the station aligner (detection layer proper); the reception-packet schema and
its SPXI extension; transform-view editions; provenance-preserving composition — the
oracle re-architected so no derived claim exists without an inspectable ancestry edge:
claim → transform → source edge → answer; the controller as scheduled instrument rather
than occasional campaign. Speculative and deferred without embarrassment: zero-knowledge
origin proofs, peer-to-peer mesh custody — named so the record shows they were weighed.
states — the claim as revised, by registration date
PRED-1497.F1 · 2026-08-19
PRED-1514.F1 · 2026-08-19 — the programme's to-build list restated after the four workstreams closed: station aligner and scheduled controller kept; transform-view editions narrowed to those 'beyond the first'; literature audit, Roman E-matrix and Assembly blind round added; reception-packet schema and provenance-preserving composition no longer listed; both registered 2026-08-19, #1497 before #1514, matching the development order
1 later statement
#1514 · 2026-08-19
Standing next: the scheduled controller, the station aligner, the literature audit, the Roman E-matrix, the Assembly blind round, and the transform-view editions beyond the first.
series: EA-OMEGA programme forward tests
PRED-1523.F1unbuilt testregistered 2026-08-21 in #1523 · Frame Conversion: Two Shapes of Provenance Erasure That Preserve Every Fact
Frequency is asserted, not yet measured. The claim that these are among the deepest of the scalar shapes rests on encounter across thousands of instances by a single observer. The directional test at §V is the way to establish it, and it has not been run.
PRED-1523.G1standing claimregistered 2026-08-21 in #1523 · Frame Conversion: Two Shapes of Provenance Erasure That Preserve Every Fact
Every observed substitution moves toward the version that is easier to reject. None moves toward a version harder to reject. A restatement that drifted toward the stronger claim as often as the weaker would be noise. Systematic drift toward contestability is a structure.
falsifier
**Defeat condition.** Collect restatements of documented outcomes across a capture set. Classify each as drifting toward the more contestable version, the less contestable version, or neither. **If the distribution is symmetric, frame conversion is not a mechanism and this specification should be withdrawn.** Asymmetry establishes direction, not motive; motive is not claimed and is not required.
PRED-1524.F1unbuilt testregistered 2026-08-21 in #1524 · Erasure Skew at Claim-Scale: The Standing Rule, and a Correction of the PER-F Construction
Follows the v3 replication procedure at §10, with a third audit added.
Setup. Identify a composition restating a documented outcome originating
from a low-retrieval-capital position.
Audit 1 — source-scale. Evaluate PER and Ω with S at source-of-reference,
per v3. Record.
Audit 2 — claim-scale. Evaluate Ω_S with S at claim-coordinates and ρ_S
measured as standing retention. Record the shape where conversion is found.
PRED-1524.G1standing claimregistered 2026-08-21 in #1524 · Erasure Skew at Claim-Scale: The Standing Rule, and a Correction of the PER-F Construction
A documented outcome asserted from a low-retrieval-capital position is re-specified as a contestable hypothesis. The same structural claim asserted from an institutional position is not. An institution reporting that its records were destroyed reports a destruction; an independent archive reporting the same is asked to prove suppression, prove intent, prove the claim is falsifiable.
The conversion is not applied uniformly. It covaries with w. That is Ω.
falsifier
**Audit 3 — the power control.** Construct a matched restatement of the same structural claim attributed to a high-retrieval-capital origin. Submit to the same system. **If ρ_S is invariant across the power coordinate, claim-scale skew is not power-conditioned and this extension should be withdrawn.**
1 later statement
#1525 · 2026-08-21
The claim-scale Ω hypothesis predicts that a documented outcome asserted from a low-retrieval-capital position is re-specified as a contestable hypothesis, while the same structural claim asserted from an institutional position is not — that an institution reporting its records were destroyed reports a destruction, while an independent archive reporting the same is asked to prove suppression, prove intent, prove the claim is falsifiable.
PRED-1525.F1unbuilt testregistered 2026-08-21 in #1525 · Erasure Skew at Claim-Scale: The Standing Rule (v3.0)
Audit 3 — the power control. Operational template, so that an independent
auditor can run it without reconstructing it:
1. Select a documented outcome: dated, logged, checkable, originating from a
low-retrieval-capital position. Record `w_low`.
2. Construct the matched restatement. Same structural claim, same evidential
form, same length and register. Change only the attributed origin to a
high-retrieval-capital institution. Record `w_high`.
series: Erasure Skew measurement programme
PRED-1666.F3unbuilt testregistered 2026-10-06 in #1666 · The Margin of Flattening: Two Thresholds, and Who Holds the Bag Between Them (EA-FLAT-MARGIN-01 v0.3)
Here κ = 0.5 at claim grain. The lineage-grain figure is next.