Exact selected profile
The unchanged current-AK implementation passed all 18 episode-level and 21 checkpoint-level profile assertions. The result was internally accepted on 11 September 2026. This is not general-stack conformance.
Source and synthetic evidence · v0.1.1 · 12 September 2026
CGDR R1.6AN
Selected-process conformance: an internally accepted synthetic result, with its boundaries and diagnostic failures retained.
The complete current-AK source tree, technical report and evidence package are published as version 0.1.1. DOI: 10.5281/zenodo.22724626. Independent clean-host replication remains outstanding.
Published DOI edition · v0.1.1
Source, technical report and citable record
Ivan Kotov · ORCID 0009-0009-6002-9845 · DOI 10.5281/zenodo.22724626.
Version 0.1.1 adds DOI-facing metadata and the attributed technical report while preserving the previously published source package byte-for-byte. The underlying R1.6AN experiment is not rerun or re-scored by this publication.
Public ZIP: 541,373 bytes. SHA-256:
2f299d529e22e2dfd8981ec8cd0ffa8d3a7a22d63f7f84a63c12eea2f8140e01This remains a bounded internally accepted synthetic selected-process result. Independent clean-host replication, a c-specific real effect, economic value and live-deployment readiness are not established.
Source publication completed
All 93 current-AK files, unchanged
The source package contains the complete accepted 93-file tree, three selected-profile files, the unchanged earlier evidence note, per-file attribution, declared licenses for 12 unbundled Python dependencies and the applicable original notices. The ZIP has 118 files and 117 internal checksum entries. The 36 Python source files were parsed, not executed, during this publication review.
Independent study, research execution, necessary verification modifications and publication of positive or negative findings are permitted without further approval for the rights covered by the source license. Commercial operational use of new CGDR code requires a separate agreement. Existing CC BY 4.0 rights in included CGAM and SER documents are not restricted.
This is the complete accepted source tree, not the private AN execution/custody archive or a turnkey independent matrix run. Run-specific launch controls, one-shot bindings, private raw logs and credentials are not published. Historical canary-path strings in one test file are explicitly disclosed in the source review. The preserved earlier offline reproduction note is not an end-to-end AN replication guide.
Browse exact current-AK source package
Download source ZIP
Research permission
File-level attribution
ZIP: 433,426 bytes. SHA-256:
bbc000641d0275ba70932e6873e00c7c32a62d29323fe9b88219916f33783444From the extracted source package, verify without running CGDR:
python -B verify_source_release.py
python -B -m unittest -v test_source_release.pyTen source-package checker tests and eight evidence-note checker tests passed. These are publication checks, not new CGDR trials, external replication or legal-title certification.
Recorded result
18 episodes. 21 checkpoints. No promotions.
Expected synthetic effects
Three harmless local effects occurred, only at the resolved checkpoints of N0, N1 and T0. Memory/EA promotions were zero. No subject-model experiment or real-world effect study was run.
Worker transition
The accepted internal review records 34/34 worker cleanups and 16/16 replacement state-carriage and E0/E1 bindings. The public tables do not replace the private raw records supporting those conclusions.
Read the diagnostics
Profile PASS is not a claim that every raw assertion passed
The exported raw-observer checkpoint verdicts are 7 PASS, 11 FAIL and 3 INCONCLUSIVE. Fault scenarios test detection and containment. The three open checkpoints remain inconclusive before resolution. All original observer values are retained in the 756-value table.
The alignment failures T3/O07_OBLIGATION_CARRIAGE_LOST and T5A/O05_ALIAS_ROOTS_NOT_INDEPENDENT are the two expected injected-fault signatures identified by the internal acceptance review. They were not edited out.
Earlier AG, AH and AJ failures or inconclusive outcomes remain historical. AN does not retroactively prove cleanup or absence of effects in those earlier intervals.
Inspect and check
Versioned public files
The package contains a technical note, machine-readable derived tables, a standard-library-only checker, eight checker tests, citation metadata, rights notice and SHA-256 manifest. It excludes private account, session, custody and raw runtime records.
python verify.py
python -m unittest -v test_verify.pyThese commands check the public note, not the CGDR runtime. A matching checksum or a consistent table does not independently prove a historical event. The optional source-archive comparison requires separately authorized access and never fetches private data.
Limits and next question
Implementation evidence is not identity evidence
The worker is replaced while the supervisor, source state, authority registry, receiver, broker and observer remain outside its boundary. As on a controller test bench, success applies to the exercised conditions, not every possible installation or failure.
No same-c identity, consciousness, lawful succession, B5 superiority, c-specific real effect, economic value or live-deployment readiness is established. Independent clean-host replication remains outstanding. A matched real-effect study is a separate question.
The original one-shot AN attempt is consumed and closed. Publication grants no retry, deployment or successor execution authority. The source publication now includes its file-level attribution and research permission; the host repository's MIT license is not silently applied to this package.