Paper A — δ-path artifact (2026-05-18)

Purpose. Single source of truth for the canonical Paper A reader path under the δ architecture. Input for Cluster 5 PR A (scripts/wikisite/paper_a.py:47 swap; new /paper-a/path/ route; argument-path schema + data seam). The current four-slug spine (science-secrete, coherent-deformation, chiasm, stiftung) at paper_a.py:47 is retired as the canonical Paper A root per the H_synth Option A architecture-level adjudication landed on master 2026-05-17 (fix-plan-2026-05-17.md Cluster 5; claims#science-secrete-stiftung-chiasm Status History 2026-05-17). This artifact replaces it.

Per CLAUDE.md .audit/ exemption. No frontmatter; not counted in wiki page totals. Apply-mode-eligible as input to Cluster 5 PR A; not itself a wiki page.

Authority chain. The reader path is anchored on two supported claims:

δ thesis (operative-form root): coherent-deformation-universal-operative-form (supported, medium, 2026-05-04). MP's coherent deformation is the universal operative form across painting AND literature in the published text, not painterly-specific. The three-element cluster coherent deformation + Stiftung + système d'équivalences operates as MP's expressive architectonic at multiple sites without chiasm at the site.
Painter-side specificity (modest residue): mp-painter-as-primary-witness-for-indirect-ontology (supported, medium, 2026-05-09). Painting is the primary witness — not the only witness — for indirect ontology in MP's published corpus, particularly in E&M (1961). What survives the γ split's defeat of the H_synth four-element-synthesis claim.

Both are concurrently supported. Neither is contested. The reader path runs through them without invoking the four-element-synthesis grammar.

§1 — Canonical δ root label

Label: the three-tier expressive architectonic + painter-as-primary-witness specificity

Short reader-facing thesis (2 sentences). Merleau-Ponty's late writings on expression operate through a three-tier architectonic — coherent deformation (the operative form), Stiftung (the diachronic mechanism), and système d'équivalences (the synchronic structure) — that runs universally across painting and literature in the published text, not in a painterly-specific register. Within that universal architectonic, painting in particular gives indirect ontology its primary witness, especially in Eye and Mind (1961), because the painter's body, the seen, and the made-canvas are co-present in a single act with no mediating apparatus — primacy without uniqueness.

One-sentence handle (for cards / nav). MP's expressive architectonic is three-tier and universal; painting is its primary witness, not its proprietary site.

§2 — Anchor claim slugs

The path stages cite-back to claim slugs that already exist in wiki/claims.md. No new slugs needed for Cluster 5 PR A.

Primary anchors (required):

coherent-deformation-universal-operative-form (supported, medium) — δ thesis. The textual root.
mp-painter-as-primary-witness-for-indirect-ontology (supported, medium) — what survives δ on the painter side. The modest claim about E&M.
ingested-corpus-four-element-gap (supported, medium) — the negative-evidence finding: no ingested secondary source occupies the four-element synthesis. Authorizes the wiki's positive contribution as something the secondary literature does not yet articulate.

Adjacent anchors (cite where useful, not load-bearing for path coherence):

nov-1960-stiftung-grammatical-subject (supported) — the single private-note site where chiasm and Stiftung do co-deploy; the philological residue.
kaushik-stiftung-literary-frame (live) — Kaushik's literary-only reading of Stiftung; useful as the secondary reading the wiki diverges from at the universality finding.
deformation-coherente-mp-coinage (status: see register) — MP-as-coiner vs. Malraux-as-source for coherent deformation; refines the credit line.

Historicized anchor (must NOT appear as a live stage; see §4):

science-secrete-stiftung-chiasm (contested, medium, 2026-05-17). The four-element synthesis claim. Background, not stage.

§3 — Ordered path stages

Six stages. Each stage carries a title, a short reader paragraph (~80–120 words), concept/claim anchors, and an explicit transition note to the next stage. Stages 1–5 are load-bearing; stage 6 is the open-questions tail that the public site should render or hide per policy.

Stage 1 — Where the wiki's reading begins

Reader paragraph. MP's late writings on expression have a recurring problem: the secondary literature reads them as four-element syntheses (chiasm + Stiftung + science secrète + coherent deformation as one operation), but the texts themselves rarely co-deploy all four. The wiki's δ reading starts from a negative-evidence finding — across the ingested secondary corpus, no source occupies the full four-element synthesis as a single architecture. This is not a complaint about the secondary literature; it is a calibration. Whatever the wiki contributes here, it contributes against an empirical gap, not against a settled rival reading.

Anchors. Claim: claims#ingested-corpus-four-element-gap. Concepts: indirect-ontology, indirect-language.

Transition. The next stage shows what MP's text actually deploys at the canonical expressive sites once the four-element framing is set aside.

Stage 2 — The δ thesis: coherent deformation is universal, not painterly

Reader paragraph. Read at the canonical sites — Indirect Language and the Voices of Silence (1952) and The Possibility of Philosophy (1959–61) — coherent deformation operates universally across painting and literature. The textual marker is decisive: at Indirect Language raw 1050, "the meaning of a novel too is perceptible at first only as a coherent deformation imposed on the visible. And it will never be otherwise." The "too" and the "never otherwise" foreclose any painterly-specific anchoring. The literature register receives the most extended gloss of the coherent qualifier (PoP Course 2 raw 2332, on Claude Simon and Conrad); the painting register receives the canonical definition. The operative form is one; the registers are several.

Anchors. Claim: claims#coherent-deformation-universal-operative-form. Concepts: coherent-deformation, indirect-language, systeme-d-equivalences.

Transition. Universality across the expressive register raises the question of what holds the architectonic together once it is no longer the four-element synthesis. The next stage answers: the three-tier cluster.

Stage 3 — The three-tier expressive architectonic

Reader paragraph. At Indirect Language raw 820–874, three terms co-deploy within ~50 lines of MP's text: coherent deformation (the operative form), système d'équivalences (the synchronic structure — "style is the system of equivalences ... it is the universal index of the 'coherent deformation'"), and Stiftung (the diachronic mechanism — "the fecundity of the products of a culture which continue to have value after their appearance"). The same three-element pattern recurs at PoP §IIb raw 597–641. Chiasm-grammar is absent at both sites. This is what the δ thesis re-architects: the operative form, the synchronic structure, and the diachronic mechanism form a self-sufficient expressive architectonic; chiasm operates as a coordinate but separable register, addressing perceptual reflexivity rather than expressive composition.

Anchors. Claim: claims#coherent-deformation-universal-operative-form (textual sites enumerated). Concepts: coherent-deformation, stiftung, systeme-d-equivalences, indirect-language. False-friend caution: do not pull chiasm into this stage as a fourth element; chiasm enters the wiki's architectonic elsewhere and on different terms ([[ineinander]], [[wild-being]]).

Transition. The three-tier architectonic is universal across the expressive register. The next stage asks what survives on the painter side once painting is no longer the proprietary site of the operative form.

Stage 4 — Painter as primary witness: what survives δ on the painter side

Reader paragraph. Painting in particular gives indirect ontology its primary witness — not its only witness, not exhaustively, but primary — because painting is the medium where (a) coherent deformation is most legible as a method (the painter's body, the seen, and the made-canvas are co-present in a single act with no mediating apparatus); (b) the body's interrogation of the visible is enacted rather than only described; and (c) the wild logos appears in its non-discursive register, where what direct ontological speech tries to say the painter shows. Eye and Mind (1961) is where MP most extensively develops this register. This is the modest painter-side claim. It is not a restoration of the four-element synthesis under another name; it is the painter-side residue of the γ split.

Anchors. Claim: claims#mp-painter-as-primary-witness-for-indirect-ontology. Concepts: science-secrete, fundamental-thought-in-art, indirect-ontology, E&M.

Transition. Primacy without uniqueness is a calibrated claim. The next stage shows what the wiki has retired in order to make this calibration possible — the four-element synthesis as historical background.

Stage 5 — Historical background: the H_synth four-element synthesis

Reader paragraph. The wiki tested the reading that MP's late writings on expression articulate a four-element synthesis — chiasm + Stiftung + science secrète + coherent deformation as one operation. That synthesis claim, recorded at [[claims#science-secrete-stiftung-chiasm]], was promoted to live and then contested (2026-05-05) on the textual ground that the four are never co-deployed in a single argumentative gesture in Eye and Mind, Phenomenology of Perception, or Indirect Language; chiasm-grammar is absent at the three-element-cluster sites; Indirect Language (1952) privileges language over painting; and only a single private working note (V&I November 1960, per [[claims#nov-1960-stiftung-grammatical-subject]]) shows the chiasm + Stiftung co-deployment. The four-element synthesis is preserved in the register as a contested claim rather than a retired one because the underlying philological situation is "under-evidenced," not "false." The wiki retains the failure mode as evidence of its calibration discipline; readers should not encounter the four-element synthesis as a live root.

Anchors. Claim: claims#science-secrete-stiftung-chiasm (contested), claims#nov-1960-stiftung-grammatical-subject (supported). Concepts: science-secrete, chiasm, stiftung (each page documents what survives the γ split in the page's own register).

Transition. Calibration is incomplete without naming what would close the open questions. The final stage carries the open-questions tail.

Stage 6 — Open questions and the path's own counterpressure (optional render)

Reader paragraph. Three open questions remain. First: would a re-read of Saint Aubert 2006 Vers une ontologie indirecte (currently absent from raw/) corroborate the three-tier architectonic, or articulate a different settlement? Saint Aubert's framework (Blondelian diplopie ontologique, Maine de Biran fait primitif, the 1952 langage indirect genealogy) is the named-tradition anchor for the painter-side reading; without his text in raw/, the painter-side framing rests on what the 2021 Être et chair II (in raw/) supplies through the surrounding lineage. Second: the three-tier architectonic has not been adjudicated against secondary readings of coherent deformation in Galen Johnson, Carbone, or Saint Aubert; if any of them articulates the three-tier reading independently, the δ thesis moves toward high confidence. Third: the path treats chiasm-grammar as a coordinate-but-separable register; it does not yet exhibit what chiasm-grammar does in the expressive register, only what it does not (form an architectonic with the three terms at the canonical sites). A future path stage may be warranted on chiasm-grammar's own expressive work, distinct from this path's three-tier architectonic.

Anchors. Claim: claims#ingested-corpus-four-element-gap (Counterpressure section), claims#coherent-deformation-universal-operative-form (Counterpressure §4). Concepts: saintaubert-2006-vers-une-ontologie-indirecte (existing page, ingested 2026-05-07).

Render policy note. Stage 6 is the path's own counterpressure. For the public reader it may render below a "Open questions" fold; for the maintainer view at /paper-a/argument/ it should render inline. Cluster 5 PR A's data schema should make this togglable per stage rather than per path.

§4 — Rendering `science-secrete-stiftung-chiasm` as historical contested background

Where it appears in the path. Stage 5 only. Never in stages 1–4. Never as a stage title. Never as the root.

How it appears. As an contested claim, with the wiki's Status History reasoning visible to the reader: the four-element synthesis was tested, the textual evidence under-determined it, the entry was held at contested rather than retired because the philology is "under-evidenced rather than false." This is the wiki's calibration discipline made legible; do not obscure it.

How it does NOT appear. As a "previous version of the thesis," as a "rejected reading," or as a "draft superseded by δ." Each of those framings would be philologically wrong: the four-element synthesis is a live secondary-literature reading (associated with Carbone's master-formula treatment of the chiasm), not a wiki-internal draft. The contested status is the wiki's claim about the textual record, not a status claim about the secondary reading.

Visual handling. If the reader UI exposes a contested badge or icon, the badge should appear on the claim cite, not on the stage. The stage's reader paragraph should set up the disagreement substantively; the badge should mark the slug's register status.

Data shape implication for Cluster 5 PR A. The schema should let a stage carry both anchors_supported: […] and anchors_contested: […] arrays (or one anchors array with per-entry status pulled from claims.md at build time). Stage 5 is the only stage that needs anchors_contested populated.

§5 — Terms and framings to avoid

Hard-avoid list. Any of these in reader-facing copy is a regression of the δ architecture:

"H_synth root" — the H_synth framework is retired as the canonical root.
"the four-element synthesis" in the present tense — present-tense framing implies live thesis status; use past-perfect ("the wiki tested...") or contested-status framing.
"the painterly-specific case" — δ's universality finding empirically defeats painterly-specific anchoring; the painter-side specificity is primacy, not anchoring.
"the only witness" / "the unique site" / "the proprietary medium" — for painting under δ. The primary-witness claim is primacy, not uniqueness.
"chiasm + Stiftung joint operation" / "chiasm-Stiftung grammar" as a live architectonic — the chiasm + Stiftung co-deployment exists at the November 1960 private working note (one site), not across the published expressive corpus.

Soft-avoid list. Useable with care; flag if they appear without scaffolding:

"Merleau-Ponty's late ontology" — fine as an umbrella, but if the stage is about expression specifically, prefer "MP's late writings on expression" to keep the operative-form / perceptual-reflexivity distinction live.
"the four-element apparatus" — fine as historical reference (stage 5), but never as the path's contemporary architecture.
"synthesis" — fine for "three-tier expressive architectonic" or for Carbone's master-formula reading; reserve "synthesis" for explicit contestation contexts so it does not creep back into the live root.

Preferred vocabulary.

For the canonical root: "three-tier expressive architectonic" (operative form + diachronic mechanism + synchronic structure).
For painting's role: "primary witness" or "primacy without uniqueness."
For coherent deformation: "universal operative form across the expressive register" (not "painterly-specific operation").
For Stiftung in this context: "diachronic mechanism within the three-tier architectonic" (not "joint-operation partner with chiasm").
For the V&I November 1960 attestation: "one private working note where chiasm and Stiftung do co-deploy" (single-site, philologically robust, registral exception).

§6 — Schema implications for Cluster 5 PR A

For the data seam at scripts/wikisite/paper_a.py:47 and the new /paper-a/path/ route. Codex adjudicates final shape; this is the content-side ask.

Replace FOUR_ELEMENT_SLUGS as the public spine. Keep the constant (it is still used by other Paper A surfaces — compute_paper_a_relevance and compute_paper_a_match_terms); do not delete. Add a DELTA_PATH_SOURCE reference (file or module-level structure) loaded from this artifact or a stable downstream representation of it. The reader spine is the path stages; the four-element-slug set is now a historical-membership test, not a root-identity test.
Stages are ordered and addressable. Each stage in the data shape should carry: id (kebab-case), title, reader_paragraph (markdown), anchors_supported: [{slug, type: claim|concept}], anchors_contested: [{slug, type: claim|concept}] (often empty), transition: {text, target_stage_id}, render_policy: always | optional.
Anchors resolve against claims.md at build time. Per-anchor render must reflect the claim's current status in claims.md (not a snapshot baked into this artifact). If a claim status moves from live to supported after this artifact lands, the path picks it up automatically.
No new public route name leaks. Per Codex's adjustment: the reader route is /paper-a/path/; the existing /paper-a/argument/ remains the maintainer map until intentionally retired. The path schema should make it cheap to render the same underlying data at both routes with different verbosity policies (e.g., stage 6 rendered as inline H2 at /paper-a/argument/, behind a fold at /paper-a/path/).
Counterpressure surface is not optional. The render must guarantee that at least one contested anchor and at least one open-question surface appears on the reader route. The wiki's calibration discipline depends on counterpressure being visible at the reader layer, not just at the maintainer layer. (Concrete: if a build-time check finds the rendered reader path has zero contested anchors AND zero open-question stages, fail the build.)

§7 — Provenance and adjudication

Drafted by. Claude (this session, 2026-05-18), against the inputs:

wiki/claims.md H2 entries for coherent-deformation-universal-operative-form, mp-painter-as-primary-witness-for-indirect-ontology, science-secrete-stiftung-chiasm, ingested-corpus-four-element-gap.
wiki/.audit/audit-synthesis-2026-05-05.md §"δ architecture" + §"γ adjudication."
scripts/wikisite/paper_a.py:47 FOUR_ELEMENT_SLUGS constant.
wiki/.audit/fix-plan-2026-05-17.md Cluster 5 scope.
Codex's δ-path shape request (2026-05-18 chat, "Review repo top issues" thread).

Adjudication seam. This artifact is not a wiki page and not a claim entry; it is an input to Cluster 5 PR A. Adjudication checkpoints:

Before Codex scaffolds PR A: maintainer reads §3 (path stages) and §5 (avoidance list) and confirms the reader thesis is what the wiki wants to put on a public route.
During PR A review: maintainer reads the rendered preview against §3's stage sequence; if a stage reader paragraph drifts in the template, the artifact's paragraph is the canonical text.
After PR A merge: if a later audit run (Phase 8 or weave Pass 3) materially changes the status of any anchor claim (e.g., mp-painter-as-primary-witness-for-indirect-ontology from supported to contested, or a new supported claim warranting its own stage), update this artifact with a dated revision section rather than rewriting in place. The artifact's provenance discipline is the same as a .audit/ report.

Not a claim. Per CLAUDE.md §17 (user memories nominate; artifacts authorize), this artifact is a structured navigation over already-authorized claims. It does not promote any candidate or contested claim; it does not require a new claims.md entry. If a future maintainer wants to promote "the three-tier expressive architectonic" to its own claim slug, that promotion runs through the normal claims.md gate; it is not authorized by this artifact's existence.

Revision 2026-05-18a — Stage 5 display-title divergence

Surfaced by. Cluster 5 PR B reader-UX adjudication (this session, 2026-05-18, "Review repo top issues" Codex thread). When Codex requested a public Stage 5 title decision during PR B planning, the artifact's §3 Stage 5 stored heading ("Historical background: the H_synth four-element synthesis") was flagged as reader-unfriendly: the "H_synth" label is internal scaffolding from the H_synth/δ adjudication phase that preceded §5's finalization, and the public reader has no context for the framework name. §5's hard-avoid list explicitly forbids "H_synth root" specifically; the bare "H_synth" label is internal-only either way.

Decision. Render the Stage 5 title in the public /paper-a/path/ route as "Historical background: the tested four-element synthesis." This is a display-layer derivation; §3 Stage 5's stored heading remains unchanged in this artifact.

Why display-layer rather than in-place §3 edit. The artifact's §7.2 review rule says "maintainer reads the rendered preview against §3's stage sequence; if a stage reader paragraph drifts in the template, the artifact's paragraph is the canonical text." This Revision documents the divergence ahead of PR B so a future maintainer comparing the rendered route against §3 sees the paper trail and does not mistake the divergence for unauthorized drift. The in-place §3 edit was considered and not taken for two reasons: (a) §7.3 establishes a discipline of revising via dated sections rather than rewriting in place, even though §7.3's literal scope is claim-status changes; respecting the spirit preserves audit integrity. (b) The "H_synth" label is historically meaningful — it names the framework whose four-element-synthesis reading the wiki tested and recorded as contested; preserving it in §3 keeps the philological trail intact, while the rendered display uses reader-language.

Rendered-title rationale.

"the tested four-element synthesis" satisfies §5's "use past-perfect or contested-status framing" requirement unambiguously: past-tense "tested" foregrounds the wiki's calibration discipline of having tested the four-element reading and recorded its under-evidence.
Drops the internal "H_synth" label — reader has no context for the framework name.
"Historical background:" prefix retained from §3 (past-tense framing reinforced at the prefix level as well as the noun-phrase level).
§5 hard-avoid list compliance: no "H_synth root", no present-tense "the four-element synthesis", no "painterly-specific case", no "uniqueness" framing. The rendered title is in compliance.

Scope of this Revision.

§3 Stage 5 stored heading: unchanged. Render layer overrides for display only.
§3 Stage 5 reader paragraph, anchors, transition: unchanged.
§5 hard-avoid list: unchanged.
§7 review checkpoints: unchanged; this Revision IS the paper trail §7.2 anticipates.
Source stage id (HTML id/data slot in paper_a_path.html): preserved as the §3-derived slug per Codex's PR B implementation contract; the display-title override does not change the stable anchor target.
PR B carries a tiny display-title override map (in scripts/wikisite/paper_a.py or scripts/build_site.py) for Stage 5 specifically; no other stage requires an override under the current §3.

Provenance.

Adjudication: Claude (this session) accepting Codex's alternative title ("the tested four-element synthesis") proposed in the Cluster 5 PR B planning chat (2026-05-18, ~16:10–16:30).
Implementation seam: PR B (Codex, in flight at Revision land time).
Artifact commit: independent of PR B per the .audit/-discipline (single-writer seam — Codex's exact ack: "I'll leave any parallel δ-artifact revision alone unless you explicitly ask to bundle it").

What this Revision does NOT authorize.

No change to any claim's status in wiki/claims.md. science-secrete-stiftung-chiasm remains contested, medium, with the same Status History.
No change to §3's other stage headings, reader paragraphs, anchors, or transitions.
No precedent for cosmetic display overrides. Stage 5 is the only stage whose §3 heading contains an internal scaffolding label; if §3 acquires similar internal labels in future revisions (claim-status changes that introduce new internal vocabulary), a fresh Revision section must articulate the same display-vs-canonical seam.
No precedent for in-place §3 edits triggered by display-UX preferences. The discipline established here is: display-side derivation + dated Revision section, not in-place §3 rewrites. If a future maintainer wants to fold this divergence back into §3 (in-place edit), the operation must surface a new binding constraint (e.g., §5 hard-avoid list adds "H_synth" as a forbidden form in any context), not merely a preference for code simplicity.

End of artifact. Cluster 5 PR A scaffolds against §3 + §6. Cluster 5 PR B scaffolds against §3 + §4 + §5 + §6 + Revision 2026-05-18a.