Evaluation, Counterexamples and the Limits of Claims

Abstract

The paper proposes a programme for testing meaning management as a whole process, from request to observable result and back-tracing. The scientific question is whether explicit meaning objects, conditions, dependencies and provenance help a person or a model reach a better-grounded result under limited source access and budget. The question is not answered on a representative or independent load: local targeted comparisons against a strong flat implementation gave equal statuses and no established superiority. An appendix keeps runs R37-R44, including first failures.

Claims and their basis

• The six hypotheses in the paper's table remain hypotheses, with no comparison on a representative or independent load. Basis type: hypothesis.
• R42: equal statuses in 30 of 30 pairs with a shared validator, so not independent. The first R44 integration checks gave 15 of 16 expected outcomes; not replaced by 16 of 16. Basis type: finite-test.
• R43 (8 pairs) and R44 (5 task pairs and 24 projections) also gave equal statuses; none is representative or independent. Basis type: authors-domain-assessment.
• Eight verification subjects and three separate coverage reports are design decisions. No observations on real users exist; for agents only limited local model answers (R43: 16, R44: 10; not blind). Basis type: authors-domain-assessment.
• Preservation of obligations under exact reuse holds only for immutable content, matching requirements and passed admission checks. The justification is not reproduced in this paper; the mapping of the row to a specific I15 property is NOT VERIFIED; the source MM_A0457 was not read by the card writer (the reuse model is MM_A0459). Basis type: proof-under-premises.
• A cross-check by the kit authors, self-checked by one assistant and not independent, links separate conformance, source-fidelity, outcome and limit checks to 40 of 40 expected outcomes; the first harness run gave 31 of 32. It is not held-out and has no blind review. Basis type: authors-domain-assessment.

Limits

• The catalogue of 320 synthetic plans and 18 prepared illustrations (the 18 not re-derived, source not read) is not a sample of solved tasks.
• Results on well-known books cannot exclude prior model exposure; the cited contamination work was not read (NOT VERIFIED).
• The related reuse pilot (median latency ratio 0.9982, no stable gain) did not compare a compilation level with an exact-match cache, so that falsification criterion is not yet applied.

Status and what is open

• The late research synthesis gate is open; it still needs the formal-properties document (MM_A0457) behind the reuse row.
• Independent semantic review of this text has not been performed.
• This English page is a summary, not a translation, and does not replace the Russian original.
• Not integrated and still missing: the NAV reports, the TIM reports, MM_OBLIGATIONS_20261030 together with the previous-stage report, the runs behind section 7 of the knowledge-admission amendment (MM_KA0004), and the items ING-01 to ING-08.
• Held inconsistency: one corpus audit appears as 5 594 flags and 169 ledger rows, and elsewhere as 5 538 flags and 174 ledger edits; unexplained.
• Editorial notes read the R44 task comparison as 5 task pairs and 24 projections, and the 51 equivalent A/B pairs of the same section as qualification control inputs, not a task comparison. The Russian text as published still lists 51 A/B pairs next to the comparison and is unchanged.

Relation to the full text

The Russian full text is the terminology edition of 2026-11-03. The late synthesis row is integrated only as the authors' self-check. The Russian text is published without the card corrections, so its abstract may differ from this page.

Credit and licence

Based on Knowledge Contract Arcana, https://contract.place (CC BY 4.0). Licence: CC BY 4.0. Credit lines and how to cite: ATTRIBUTION.md; rights holder statement for the repository: NOTICE; third-party sources: SOURCES.md.