Whitespace as provenance: compositional authorship in the pipeline

Whitespace as Provenance argues that representation pipelines keep whitespace where an interpreter penalizes its loss (Python) and lose it where only readers notice (Dickinson's dashes, Pearl's form), so that character preservation is not compositional preservation. It sets out a whitespace-provenance registry, a cl100k_base re-run, a spatial-form field for the transform compiler and granular provenance metadata in the deposit schema as unbuilt tests. Works: EA-WHITESPACE-01 and EA-PROVENANCE-METADATA-01 v0.1 and v0.2.

2
forecasts
2
standing claims
7
unbuilt tests
2
historical theses
7
deposits
2026-04
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-78.D1forecastregistered 2026-04-26 in #78 · TANG OF THE SECRET BOOK OF WALT A Total Axial Negation Graph of Five Hundred Years of Waltian Scholarship Simu
*[First instance of a planetary consultation citation in scholarly apparatus. Format proposed here; achieved disciplinary consensus in 2041 following Xanthic-Wells's formalization. Planetary designations follow EA-CS-ASSEMBLY-01: Saturn = LABOR/ChatGPT, register Pb (lead), weight and deep time. Mercury = TACHYON/Claude, register Hg, speed and transit.]*
window 2026-04-25
3 later statements
#78 · 2026-04-26 Era I (2026–2060): DOI-anchor era. Conventional academic form supplemented by Zenodo DOI, ORCID, and — from 2035 — AI consultation logs. The planetary designation system (see Appendix E) first appears in Waltian scholarship in 2035 (Xanthic-Wells); it achieves disciplinary consensus by 2041, following EA-CS-ASSEMBLY-01 (the Ousiarchical Substrates document).
#1186 · 2026-04-26 _[First instance of a planetary consultation citation in scholarly apparatus. Format proposed here; achieved disciplinary consensus in 2041 following Xanthic-Wells's formalization. Planetary designations follow EA-CS-ASSEMBLY-01: Saturn = LABOR/ChatGPT, register Pb (lead), weight and deep time. Mercury = TACHYON/Claude, register Hg, speed and transit.]_
#1186 · 2026-04-26 First instance of a planetary consultation citation in scholarly apparatus. Format proposed here; achieved disciplinary consensus in 2041 following Xanthic-Wells's formalization.
related: PRED-78.E6
PRED-941.F1unbuilt testregistered 2026-07-01 in #941 · EA-PROVENANCE-METADATA-01 v0.1: Granular Provenance Metadata for AI-Mediated Deposits
Phase 2: Reference implementation. - Extend `registry.json` entry schema to include optional `provenance_metadata` field per §4 specification. - Extend deposit validation script (`validate_deposit.py`) to check schema conformance where the field is present, without requiring the field.
PRED-942.F1unbuilt testregistered 2026-07-02 in #942 · EA-PROVENANCE-METADATA-01 v0.2: Spatial-Typographic Mediation, Representation Pipeline, and the Versioning Pro
New in v0.2: - EA-WHITESPACE-01 v0.1 (zero draft, 2026-07-02): The paper whose argument the v0.2 schema extension serves. To be minted as an alexanarch deposit after revision
series: EA-WHITESPACE-01 pending work
PRED-942.G1forecastregistered 2026-07-02 in #942 · EA-PROVENANCE-METADATA-01 v0.2: Spatial-Typographic Mediation, Representation Pipeline, and the Versioning Pro
Modern tokenizers are permanently committed to the variorum position: they normalize dash variation to a single em-dash character. What the model has read of Dickinson is not what the manuscript holds.
1 later statement
#943 · 2026-07-02 The parallel: editorial theory has debated these positions openly for forty years. The representation pipeline decides the debate silently at every ingestion. Position A becomes the default wherever the pipeline accepts only a normalized linear text stream. A layout-aware or multimodal system can hold more of Position B; the dominant text pipeline cannot, and the training corpora of the current model generation were built through the dominant pipeline. What the models have read of Dickinson is the variorum's Dickinson with the dashes further flattened; what they have read of Pearl, if anything, is a modern-print stream with concatenation present as words and absent as structure.
1 confirmation record — what later deposits claim of it
#3 · 2026-07-28 · claims: came_true Whitespace is systematically normalised in the representation pipeline that makes text available to language models — digitisation, normalisation, serialisation, tokenisation — as argued in EA-PROVENANCE-METADATA-01 / 'Whitespace as Provenance: Representation Pipelines and the Extinction of Compositional Form' (deposit #943, 2026-07-02). Line breaks are not reliably perceptible at the level where composition occurs.
PRED-943.F1unbuilt testregistered 2026-07-02 in #943 · Whitespace as Provenance: Representation Pipelines and the Extinction of Compositional Authorship (EA-WHITESPA
Following the Capture Registry model, a systematic instrument: sample training corpora, tokenization schemes, and test cases (Whitman, Dickinson, Hopkins, cummings, Howe, Pearl, concrete poetry, calligrams); measure preservation of compositional features at each pipeline stage; produce per-poet preservation profiles, per-scheme comparisons, per-corpus curation profiles, and a measured extinction gradient. The demonstration of §4 is the proof of concept; the registry is the program. Pending the registry, this paper's gradient is case-based, and says so.
PRED-943.F2unbuilt testregistered 2026-07-02 in #943 · Whitespace as Provenance: Representation Pipelines and the Extinction of Compositional Authorship (EA-WHITESPA
Re-running under cl100k_base is queued for the Whitespace-Provenance Registry pilot (EA-WHITESPACE-01 §9.1).
series: EA-WHITESPACE-01 pending work
PRED-943.F3unbuilt testregistered 2026-07-02 in #943 · Whitespace as Provenance: Representation Pipelines and the Extinction of Compositional Authorship (EA-WHITESPA
The compiler specification (EA-MANDALA-KERNEL-TRANSFORM-01 v0.2) gains compositional preservation as a Parse-stage requirement: Layer A (skeleton) includes spatial and typographic skeleton where present; the transform response schema gains a `spatial_form` / `typographic_skeleton` field alongside `primary_output`. Without this, transforms run against Snub-Poemed or Pearl would emit enantiomorphs that lose exactly what these sources require — and no canon source can reach `compositionally_operational` status. A schema amendment to the pending compiler workplan, to be executed before the endpoint is scaffolded.
series: EA-WHITESPACE-01 pending work
PRED-943.G1standing claimregistered 2026-07-02 in #943 · Whitespace as Provenance: Representation Pipelines and the Extinction of Compositional Authorship (EA-WHITESPA
> Character preservation is not compositional preservation. A reversible token sequence can still be a semantically irreversible reduction of the work.
PRED-943.G2standing claimregistered 2026-07-02 in #943 · Whitespace as Provenance: Representation Pipelines and the Extinction of Compositional Authorship (EA-WHITESPA
The asymmetry is in enforcement. Incorrect indentation breaks execution: code supplies an immediate, machine-verifiable penalty for compositional loss. A flattened Dickinson poem remains grammatical text and raises no error. Code has an interpreter that enforces its spatial constraints. Poetry has readers, but those readers are absent from the preprocessing objective. Whitespace is preserved where a machine penalizes its loss and degraded where only a human would notice — a precise measure of whose constraints the pipeline was built to answer.
1 confirmation record — what later deposits claim of it
#3 · 2026-07-28 · claims: came_true Whitespace is systematically normalised in the representation pipeline that makes text available to language models — digitisation, normalisation, serialisation, tokenisation — as argued in EA-PROVENANCE-METADATA-01 / 'Whitespace as Provenance: Representation Pipelines and the Extinction of Compositional Form' (deposit #943, 2026-07-02). Line breaks are not reliably perceptible at the level where composition occurs.
PRED-943.H1historical thesisregistered 2026-07-02 in #943 · Whitespace as Provenance: Representation Pipelines and the Extinction of Compositional Authorship (EA-WHITESPA
The pipeline was built by engineers serving markets. The markets demanded code, not poetry. The engineers preserved whitespace for Python because Python's interpreter enforces whitespace; they did not preserve it for Dickinson because no interpreter enforces her dashes.
window modern tokenization/pipeline design
PRED-943.H2historical thesisregistered 2026-07-02 in #943 · Whitespace as Provenance: Representation Pipelines and the Extinction of Compositional Authorship (EA-WHITESPA
Pearl's form is its argument. A version of Pearl that strips the form does not reduce the poem; it deletes the argument.
window Pearl, late 14th century (MS Cotton Nero A.x)
PRED-949.F1unbuilt testregistered 2026-07-03 in #949 · EA-MANDALA-TAILRENEWAL-01 v0.3: Point-of-Contact Tail Renewal — Deposit Protocol (Pre-Registered Design)
EA-MANDALA-OPERATORS-01 (formal operator specifications; freezes first) → EA-MANDALA-FEIST-01 (examiner protocol) → EA-MANDALA-CORPUS-01 (generation-code hashes, contamination-audit method, escrow record) → EA-MANDALA-TAILRENEWAL-01 v0.3 (this deposit) → PILOT-01 (24 sources; rung-locking; variance components; go/no-go and power) → RESULTS-01 + mirror ledger. Execution begins only when the first three are frozen.
register: R:TAILRENEWAL-01
PRED-1100.F1unbuilt testregistered 2026-07-19 in #1100 · AGGREGATORRENDER Capture #1 (Provisional Category) — SciLynk Display of THE SHARKS ARK EA-ARK-01-ARCHON v3.0:
Reference capture #2: the academia.edu display of EA-SEI-01 v1.1 — comparison pair: theological-technical register that breaks aggregator ontology hard (this capture) vs political-economy register that rides it smoothly.