Phase 7 established that the Hanumanic Strategy corpus does not cross-reference itself: sixteen documents, one explicit citation, with four corrections sitting in the same folder as the errors they fix. That was recorded as the corpus's governing structural defect.
Phase 8 turns it into a sharper question, because one construct does propagate — and once there is an exception, the finding stops being the corpus is forgetful and becomes what makes a construct portable inside one author's own body of work?
The exception. Constitutional adaptation — Vāta, Pitta, Kapha — appears in the psychology cluster's requisite-variety system, in the memetic document's viral design, and in the creator cluster's writing adaptations. Three clusters, same three categories, sometimes for opposite purposes.
The mechanism, provisional. Constitutional adaptation produces a section. Drop it into any document and it generates three sub-headings with content under each. The safety framework produces no section. The exaptation frame produces a clause. The conversion table produces two lines. None of those travelled.
⚑ So the working hypothesis is that the corpus is template-driven rather than memory-driven, and what survives between documents is whatever fits a slot a drafting document already has open.
The cost, made concrete. Phase 8 found the sharpest instance yet: the four stages Warming → Peak → Patience → Highest Worldly Accomplishment are introduced in one folder document as "Bodri's four stages," naming the source, and appear in another as "Ram bhakti stages." Nothing had to be researched. A citation existed and did not travel. Which collapses the attribution problem and the propagation problem into one problem — and makes the repair a cross-reference rather than a correction.
The other cost. The safety framework is the case where this matters most: four conditions, clinical consultation as a precondition, an observable stopping criterion — governing one document, while a sibling prescribes 90–180 minutes daily at crisis intensity with no clinician named.
And the unexpected benefit. A construct that cannot propagate has to be re-derived, and the anti-suppression principle was re-derived four times in four vocabularies by three documents that do not cite each other. Four independent arrivals at one position. The propagation failure costs the corpus its citations and buys it convergent evidence — which is why that principle is the one doctrine in the folder needing no repair.
| Claim | Evidence | Status |
|---|---|---|
| The corpus does not cross-reference itself | 16-document census, phase 7 | Established for the creator cluster |
| Assembling documents do cite | the crisis-navigation document carries two explicit cross-references | [n=1 document] — narrows the finding rather than contradicting it |
| What propagates is what has a shape | constitutional adaptation across 3 clusters vs 5 non-travelling constructs | Hypothesis, [n=1 positive case] |
| Attribution failure is propagation failure | the Bodri instance | [n=1], and it is the cleanest instance |
| Non-propagation produces convergent re-derivation | anti-suppression, 4 formulations, 3 documents | [n=1 principle] |
⚠ Nothing here is above [n=1] except the original census. The shape hypothesis in particular is one positive case and five negative ones, which is suggestive and is not a finding.
52 folder documents are unread. The hypothesis makes checkable predictions before any of them is opened:
Prediction 3 is the one worth testing first, because it is cheap — the titles are visible without reading anything — and because it would tell the book whether the repair is editorial or architectural.
If the corpus's problem were quality, the book would need better ideas. It does not have a quality problem — this build has recorded the conversion table, the pre-registered designs, the four safety conditions, the Nārada method, the Negative Space Observer, the dual-assessment log and the exaptation frame, every one of them good and every one of them used once.
It has a distribution problem, and a distribution problem is fixed by architecture: give the orphaned constructs a shape, and they will travel for the same unremarkable reason constitutional adaptation does. A required column. A standing preamble. A fixed closing section.
⚑ Which is also an argument about how to write the book itself — and it is developed further in the phase-8 spark, where the same evidence suggests the author writes well when the form asks a question back and drifts when it does not.
[n=2 — phase 7's compound-signature collapse is the reason to say nothing stronger.]Prediction 3 was pre-registered at the top of WORKBENCH/Research/hanumanic-phase-9-extraction-log.md before any document was opened, classified from titles alone, with two null outcomes named in advance.
Predicted: cross-references concentrate in assembling documents; generating documents carry roughly none. Found: assemblers 5/5, generators 7/9. Twelve of fourteen cite. ⚠ The first named null — "if cross-references are evenly distributed… the shape hypothesis is refuted" — is met.
⚑⚑ The only two documents in the cluster carrying no cross-reference are the two whose subject is an external traditional text — the Sankaṭ Mocan material and the Hanumān Chālīsā.
Every document built on the corpus's own mantra work cites its neighbours. The two answerable to somebody else's text do not.
⚑⚑⚑ What produces cross-referencing is not document type and not discipline. It is a SHARED OBJECT.
⚠ This is not forgetting. It is an absence of pressure. ⚑ Which changes the repair completely: the corpus does not need a citation discipline, it needs shared objects. Give two documents the same subject and they will cite each other with nobody being told to.
⚑ Demoted, not dead. Constitutional adaptation still travels across clusters that share no object — so some constructs do move on their own, and producing a fillable section is plausibly why. Both mechanisms are live; the object-sharing one is much stronger and explains far more of the data.
⚑⚑ Chapters about the same thing will cross-reference. Chapters about different things will not, whatever the style guide says. So the architecture question is not "how do we enforce internal citation" but "what are our recurring objects?" — and a book organised around a small number of returned-to objects will be internally coherent for free.
⚠ And it reframes the corpus's whole propagation failure as a design consequence rather than a character flaw, which is both fairer and more useful.
⛔ Prediction 3 is closed. The remaining live prediction is 4 — a document with a drawbacks or critical-assessment section will carry a real limit; one without will not. Phase 9 added five positive cases and no negatives (the reasoning-approach document's eight drawbacks, the Chalisa precautions, the leadership safety measures, the Green/Yellow/Red tiers, and the earlier Bhakti-Sorcery drawbacks). ⚑ [n=7] and still no counter-example — which is now strong enough that a counter-example is what the re-sweep should be hunting.