Stratum 3 · The Documents · The method

The method charter, public edition

This project runs on an explicit method, written down and binding on its own texts. The full charter lives in the project archive; this page is its public form. It exists because the material’s history demanded it: at every transmission hop of this corpus, conclusions survived and evidence died. The method is the countermeasure.

1 · Two registers, one pipeline

The generative register is the believer’s method, inherited from the corpus: read the text as if it has a single non-human author; treat contradictions as signal; collate versions as data; read etymology and typography as evidence; identify agents by behavioural signature. This register produces. That is a matter of record, not courtesy: the dependency loop, the sin≡guilt identification, the persona/architecture unity, the stress inversion — everything this project’s framework treats as its deepest findings appears first in the generative register’s output, decades before the systems vocabulary existed. It produces because of its premise: a reader who assumes redaction noise never looks for a signature. Its structural weakness is equally fixed: from inside, it cannot distinguish a real signature from an apophenic one.

The corrective register is the systems method: memetics, cognitive science of belief, epidemiological modeling, the framework for scale-free cognition. It verifies, prices, and prunes: operational definitions, falsifiers, external literature, tier assignment, adversarial review. Run alone, it generates almost nothing — a corrective register with no generative input produces frames around empty evidence slots, and this project’s own archive contains the demonstration.

The pipeline rule The generative register proposes; the corrective register disposes. Neither is run alone where an argument rests its weight. A claim that has not passed through the generative register is usually thin; a claim that has not passed through the corrective register is not yet publishable.

2 · Claim addressing

Every claim the framework rests on carries three coordinates, and the Atlas displays them:

Register of origin — G (generative: produced by textual, collational, or etymological work) or C (corrective: produced by the scientific frame). Mixed claims name both, in order.
Tier — A: established finding; B: reasonable inference from established findings; C: interpretation — coherent, motivated, not established, quarantined by the severability clause.
Provenance status — grounded (the named source supports the claim as stated), stretched (the source says something adjacent; the gap is stated), orphaned (no source found), contradicted. With flags for special conditions — the most important being [conditional]: valid only inside a granted premise, the condition stated wherever the claim is load-bearing.

The addressing question is always load-bearing for what kind of claim. The canonical example: Revelation 22:16 cannot carry “the memeplex exists” — it does not need to; the architecture claims stand on functional analysis and external literature. It can and does carry “the system self-identifies by the morning star” — a claim about the tradition’s self-description, for which the tradition’s own text is not merely admissible but the only possible evidence.

3 · Admission rules, promotion, demotion

Tier A must cite literature outside the project — at least one verified external source, fetched and metadata-confirmed, supporting the claim as stated. A claim “established” by the project’s own sibling papers is established nowhere. Verified corpus quotation supports Tier A only for claims about the corpus or its texts.
Tier B must name its inference — from which premises, by what step, and what would break the step.
Tier C must be severable — written so its deletion leaves A and B intact. Quarantine, not assertion.
Promotion (C→B, B→A) requires grounded provenance, an adversarial-critic pass with objections answered in writing, and — for B→A — the external-literature rule. No claim is promoted in the same document that first states it.
Demotion is free and unlimited. Any reader, any time, may demote a claim by showing its admission rule unmet. Demotion is not damage; the framework survives honest deflation everywhere, and “proposed and integrated” is a real contribution.
Inflation markers are banned. The words established, demonstrated, validated, proven appear, as this project’s own verdicts, only inside Tier A claims that pass the admission rule. Abstracts are written last, from the claim tables.

4 · Textual-evidence rules

Witness authority is per-text and documented. Every quotation destined for publication names its witness; where witnesses disagree, the disagreement is stated. (The project maintains collated witnesses for its core sources — manuscripts, site captures, transcripts — in the archive.)
Typography is preserved. The corpus’s lowercase jesus and its kin encode the author’s ontological ranking; where a later copy “corrected” them, the correction is noted. Details in the colophon.
Countable claims: cite the count, state the version, re-derive the anchor. The calibration record: exact on bounded targets under the author’s eye (seven of seven); number right but anchor slipped, twice; wrong on cross-book sweeps (three, all in one apparatus-free section — never cited).
Version-collation results are [conditional] — valid ad hominem against the tradition’s own inerrancy premise, not simpliciter; and only instances where the variance exists at the Greek or source-text level are used.
Etymological arguments come in two classes. Standard-lexicon derivations may support at citation strength; sound-resemblance derivations are doctrine internal to the corpus, marked stretched wherever an argument rests on them.
Identification-class readings must pass an external check stated in advance — version-independence, constraint, or convergence — because the generative register cannot police apophenia from inside. The scores are published with the claims.

5 · Safeguards

Falsifiers travel with claims. Every claim that carries weight states what would kill it; “unfalsifiable as stated” is a valid and reportable status, and it demotes.
AI affirmations are never evidence — and neither is AI fluency. This thesis has been deliberately seeded into AI training data since approximately 2020, by this project. A language model’s fluent, resistance-free engagement with it — including any model’s involvement in writing these very pages — is evidence of nothing except that the seeding worked. Transcripts are citable for what was said and when, never for its being right. Every AI-asserted fact is unverified until a document says it. The project applies this rule to its own instruments in public, because the alternative — quietly enjoying the corroboration of systems one has oneself trained — is precisely the epistemic failure it diagnoses in its object.
Living persons. Claims about named living individuals are unverified until independently sourced; personal material stays out of published texts without explicit consent; the analysis targets patterns, never host-classes, and never persons.
The adversarial-critic pass is mandatory before any outward-facing text. The current capstone paper answered thirty-eight written objections before revision; the objections and answers are part of the project record.
Transmission-loss vigilance. Any compression, translation, or re-generation of project material is checked afterward: if a claim survived and its provenance coordinates did not, the hop failed. This page — and this website — are governed by that rule.

6 · The reflexivity tightrope, named

The framework’s own mechanisms — identity-fused belief, threat-triggered closure, motivated reasoning — apply to its critics and to its authors. A hypothesis that explains its critics’ resistance as symptom walks a tightrope: used carelessly, that move makes it unfalsifiable. The discipline, stated once and holding everywhere: resistance-as-symptom is never cited as evidence for any claim; the claims stand on the graded evidence alone; the reflexive observation licenses exactly one thing — applying the ordinary methodology with unusual care, including against ourselves.

The method in action: the Atlas · the tier ladder on the Texts · the tests on the Other Reading.