Return to the full study

KEDDEH SYSTEMS · K-KRS · Technical editorial E1

KEDDEH SYSTEMS | RESEARCH AND FORMAL PUBLISHING | ENGINEER · AUTOMATE · SCALE

KRS Finite Counterpart Model: Contracts, Reproduction and Source Fidelity

KEDDEH-TECH-K-KRS-001 | Technical editorial E1 | 18 October 2026
Aboudy Keddeh — Systems Architect
Keddeh Systems | Internal research · Owner review | Editorial draft

Authorship, rights and editorial standing

1. Abstract and purpose

This reference implements a finite counterpart-identifiability control. Given a declared preparation table and an observed readout with optional context, it returns a counterpart only when all matching preparations agree. Otherwise it returns ambiguity or no admissible preparation. Ten deterministic tests pass in the new isolated 18 October 2026 replay, including an inserted counterexample that converts a formerly unique answer into ambiguity.

The implementation is explicitly not the native KEX/ZCG codec, QZL hardware, a universal KRS law or a demonstrated KRS-to-P15 adapter. Its role is to make one bounded inference question executable while retaining the authored source's premises. “KRS” is kept as the source label; no expansion is invented.

The R4 fidelity audit withdraws an earlier reviewer inference that foreground filtering necessarily establishes information loss in the complete observer map. The corrected reading retains all terms of Or=T(S,Mo,F,H,R), including environmental translation and recursive residue-memory.

Source coverage and new replay

The implementation, tests, RESULTS and R4 audit were retrieved again from the exact branch. Restoring the retained capture's terminating newline reproduces all four registered byte hashes; no code semantics were edited. Historical test-log and six-case raw receipts were not recovered for inclusion. New reproduction-tests.txt and reproduction-result.json separately record a fresh Python 3.12 run on 18 October 2026: ten tests passed and six cases emitted. These new files are not relabeled as historical receipts.

2. Source identity, setup and model scope

The source is the research/krs-observer-counterpart-r1 branch of Keddeh1/Virtual-File-Space-ZCG-ARCHITECTURE. The implementation docstring calls it a testable research control. Primary files are src/krs_counterpart.py, tests/test_counterpart.py, docs/RESULTS.md and scholarly/SOURCE_FIDELITY_AUDIT_R4.md. Historical outputs comprise the ten-test log and six reference cases; new reproduction outputs are included separately.

The source uses Python standard-library dataclasses, enum, typing and json. Its built-in tuple annotations require a compatible modern Python; Python 3.12 is suitable. No third-party package, browser, network, native KEX runtime or hardware adapter is required for this finite control.

Implementation SHA-256 is bc101cbde5026b301819fefabe6ebdffa182d7d6753ab8dea546d6a466fcdd10. Test SHA-256 is 3c46d0f9ef78cb6b9107620c961aaa38cdd10e5cd346a6d3b1120de2eaa90473. The R4 audit digest is 3591ff630411d466777a051c0c8b974d87da3f4f155461a10d476b1bd4560449. Byte-exact originals and their manifest should be verified before reproduction.

The finite preparation universe is part of the model. A conclusion is conditional on its completeness for the intended query. It is not inferred from the observed scalar alone or from a claimed universal physical relation.

3. Data and function contracts

3.1 Preparation, Readout and Inference

Preparation is a frozen dataclass with identity, observed, counterpart, frame, lineage, observed_time and address, all annotated as strings. Its address is explicitly a synthetic test key, not a native KEX address.

Readout is a frozen dataclass containing observed and optional frame, lineage and observed_time. None means that a contextual filter is not supplied. It does not mean numeric zero, unknown physical existence or a measurement of absence. Dataclass annotations are not runtime validators; callers remain responsible for the declared types.

Inference is a frozen dataclass with status, candidates, counterpart and matched_preparations. Candidates is the lexically sorted tuple of distinct counterpart strings among matches. matched_preparations preserves the input preparation order and identity strings. Counterpart is emitted only for the unique status; otherwise it is None.

Status values are DEDUCED_UNIQUE, AMBIGUOUS and NO_ADMISSIBLE_PREPARATION. The enum member names in code are UNIQUE, AMBIGUOUS and MISSING respectively. The explicit missing status is not a zero-valued result.

3.2 infer(preparations, readout)

infer first selects preparations whose observed value matches the readout. Each supplied optional context adds an equality condition on frame, lineage or observed_time. It then collects distinct counterpart values:

• No matching preparations: NO_ADMISSIBLE_PREPARATION, empty candidates, counterpart None.
• At least one match and exactly one distinct counterpart: DEDUCED_UNIQUE and that counterpart.
• More than one distinct counterpart: AMBIGUOUS with the complete distinct candidate tuple and counterpart None.

Uniqueness is uniqueness of counterpart value, not necessarily a single matching preparation. Two preparations can support the same answer while retaining two identities. The function performs no I/O and does not mutate the frozen input records. It does not validate preparation-identity uniqueness or synthesize missing preparations.

For a fixed declared preparation set, the criterion is that the target counterpart be constant on the admitted match set. This is a finite control's criterion. Applying an analogous fibre argument to the authored full observation map requires declaring that map, domain, target and all premises; it is not a theorem supplied by the mere notation T.

3.3 run() and command-line behavior

run applies six fixed readouts to PREPARATIONS and returns a JSON-ready record with model, whole_identity, case_count and results. The model string is finite-reference-not-native-KEX-or-QZL. Each result retains the readout and inference fields.

The historical fixture includes whole_identity=1. It remains byte-exact source metadata and is not a current .1 encoder. Current author clarification treats .1 as a literal nonnumeric origin designation. The fixture's numeral does not override that clarification or define native address syntax.

When krs_counterpart.py is executed directly, it prints indented JSON. The lifecycle is construct declared preparation universe → construct readout → filter → deduplicate counterpart values → classify → retain matched identities. There is no native actuation or state-evolution stage.

4. Reproduction method

Run python verify_sources.py in the edition folder to check included source identities, then python reproduce.py to stage the exact sources in a temporary layout and run both commands below. Preserve the src/tests sibling layout in a disposable reproduction directory. The test source inserts its sibling src directory into sys.path, so no package installation is required.

From the reproduction model root run:

• python -m unittest discover -s tests -p 'test_counterpart.py' -v
• python src/krs_counterpart.py

The first command executes the ten deterministic tests; the second emits the six reference cases. Record Python version, source/test digests, complete test output and the emitted JSON. A fresh run may report a different elapsed time; the substantive comparison is test outcomes and structured cases.

If using a flat source-files package, first stage the two Python files under src and tests according to package-plan.json or use the provided reproduction helper. Do not rewrite imports in the retained original merely to fit a new packaging layout.

5. Results and counterexample

The four fixed preparations are:

• W1: negative → positive-A, frame F1, lineage L1, observed time T1, synthetic:001.
• W2: negative → positive-B, frame F2, lineage L2, observed time T1, synthetic:002.
• W3: positive → negative-A, frame F1, lineage L3, observed time T2, synthetic:003.
• W4: negative → positive-A, frame F1, lineage L4, observed time T2, synthetic:004.

The six emitted cases are:

1. Negative alone: AMBIGUOUS between positive-A and positive-B; matching identities W1, W2 and W4 remain retained.
2. Negative + F1: DEDUCED_UNIQUE positive-A, supported by W1 and W4.
3. Negative + F2: DEDUCED_UNIQUE positive-B, supported by W2.
4. Negative + L4: DEDUCED_UNIQUE positive-A, supported by W4.
5. Positive alone: DEDUCED_UNIQUE negative-A, supported by W3.
6. Unknown: NO_ADMISSIBLE_PREPARATION, with no counterpart.

The included new replay reports 10/10 for the unchanged source test suite. It checks these principal classifications, identity retention, a withheld-counterpart readback for every preparation using full context, the six-case receipt and a deliberate conflict mutation.

The critical counterexample adds W5: negative → positive-C, frame F1, lineage L5, observed time T3, synthetic:005. Negative + F1 must change from unique positive-A to ambiguity between positive-A and positive-C. This demonstrates that uniqueness depends on the admitted preparation universe. It does not demonstrate that the real universe contains W5 or falsify the author's conditional counterpart proposition.

The included reproduction-tests.txt log records 0.000 seconds. This is test-harness display precision, not a performance benchmark. No GitHub Actions execution, native QZL run, hardware readback or universal reconstruction result is implied.

6. Source fidelity and interpretation limits

The R4 audit's principal correction concerns Or=T(S,Mo,F,H,R): Or is observed reality; S source state; Mo observer motion; F foreground filtering; H environmental translation; R recursive residue-memory. Function notation alone supplies no injectivity, reversibility, independence or information-loss theorem. Reducing the equation to F and treating F as noninjective was a reviewer-added premise. That criticism of the source was withdrawn.

A permissible future test declares the full T and admissible domain, then asks whether distinct source/context tuples share an output and whether the required counterpart is constant over those tuples. The finite reference demonstrates such a decision rule for its own table. It does not prove the answer for native T or show that R cannot support reconstruction.

Other retained audit boundaries matter:

• O is an algebraic observer variable; unresolved X must remain unresolved, not replaced by zero or a preassigned tuple.
• Observed-time relationships include both time/observed-time and observed-time/time orientations. String timestamps in this fixture are filters, not implementations of those relations.
• A deduced counterpart under accepted relation/orientation is distinct from an independently observed physical entity.
• Retained residue in a transition expression does not itself settle Markov/non-Markov semantics.
• Different persistence formulas are variants until common domains, conditions and meanings establish an actual contradiction.
• Criticism of an assistant-created random-walk control does not falsify the authored naturalisation proposal.
• Missing native operators in this small branch do not establish absence in the broader corpus.

K-WP05 supplies observer/context structure; K-WP17 supplies a foreground/observer/residue evaluation programme; K-ORM and K-CURRENT govern current source interpretation; K-KEX and K-LINE concern native structure. Their assumptions cannot be imported selectively to manufacture a stronger result. No complete two-/three-body quadratic operator or P15 integration is supplied by this control.

7. References and package plan

• K-KRS: KRS finite counterpart branch, including implementation, tests, results and Source-Fidelity Audit R4 dated 12 October 2026.
• Historical output identities: krs-reference-test-output.txt and krs-reference-six-cases.json, from the earlier 18 October 2026 rerun; their raw files were not recovered. Included new outputs: reproduction-tests.txt and reproduction-result.json, from the separately identified replay.
• Draft XXXVII constraints paragraph 22, full observer expression; the audit uses secondary extracts and does not claim a full monograph verification.
• K-WP05, K-WP17, K-ORM, K-CURRENT, K-KEX and K-LINE: related source definitions, proposals and architectural boundaries, not interchangeable test oracles.
• original_hashes.json identifies every packaged original; package-plan.json records the reproducible src/tests layout. Editorial helpers are separate from originals.
• KEDDEH-STD-PUBLISHING-001 v1.0, Publishing Standard.

8. Revision history

18 October 2026 — Technical editorial E1. Expanded the finite reference into a full input/output, lifecycle, reproduction and source-fidelity guide. Preserved the conflict counterexample and withdrawn inverse-loss criticism. No native integration or external approval is asserted.

© 2026 Aboudy Keddeh | Keddeh Systems | All rights reserved
KEDDEH-TECH-K-KRS-001 | Technical editorial E1 | Internal research · Owner review | Editorial draft