Repository navigation
Releases: node-and-norm/node-norm-control-evidence
Release list
Missingness-Boundary Execution Infrastructure
Infrastructure prerelease, 3 October 2026. This is a development snapshot, not an empirical or scientific release. The repository still contains zero independently coded empirical events and has no approved corpus release.
This version adds the reviewed eight-packet missingness-boundary design, a frozen 32-request runner, answer-level eligibility, order-preserving request hashes, and paired analysis of twelve dependent single-factor edges. Canonical and reversed option-order results remain separate. Incomplete attempts are retained, and unavailable answers are excluded from comparisons that require them.
The included 32-response rehearsal uses a uniform synthetic transport and records zero live requests. It verifies archive plumbing and descriptive arithmetic only; it does not measure model performance, validate the authored references, establish independent accuracy, or test TAE/HIT. The authored cases are exposed development material, not a holdout. No historical runs or canonical codebook values changed.
Source commit: 0287ce6f16c45e02ee31a9697f3a24d6ac00c584
See docs/releases/v0.3.0-dev.md in the tagged source for reproduction instructions. Verify the downloadable archive and manifest with shasum -a 256 -c SHA256SUMS. Checksums establish byte consistency, not truth or independent review.
v0.2.0-dev · Inspectable development evidence
v0.2.0-dev · Inspectable development evidence
Infrastructure prerelease, 3 October 2026. Zero independently coded empirical events; no approved scientific corpus release.
This snapshot adds the sixteen-condition TAE/HIT development sandbox, preserved execution and technical appendices, paired checklist/trace procedures, trace-quality stress evaluation, the fixed 36-request Jev run, and its nine-issue boundary review. The release also provides a reproducible figure and formulas, and sixteen unrun successor paired cases.
The first Jev pass matched 24/33 valid determinations with 33/36 coverage against an AI-authored reference. Five of 36 requests across all passes were invalid. All outputs remain preserved; all disagreements remain unresolved. The successor is a public, exposed development draft, not a new validation result.
Download and inspect
cec-development-evidence-0.2.0-dev.zip contains tracked experiment directories, the figure and source table, plotting script and release notes at the tagged commit. release-manifest.json records the source commit, package size and SHA-256 digest. SHA256SUMS covers both files. GitHub's source archives provide the complete repository, tests and citation metadata.
Verify downloaded assets with shasum -a 256 -c SHA256SUMS. Reproduce the evidence ZIP from the tagged checkout:
git archive --format=zip --output=cec-development-evidence-0.2.0-dev.zip v0.2.0-dev experiments docs/figures scripts/plot_cec_live.py docs/releases/v0.2.0-dev.mdUse the figure guide for plotting and formulas, runner guide for archived results, and successor specification for development cases. The repository CI checks schemas, examples, links, tests and saved sandbox evidence. Checksums establish byte consistency, not truth or independent review.
The archive contains authored synthetic data and preserved model outputs. It grants no new rights to third-party materials and claims no institutional endorsement. Human review, construct validation and empirical release approval remain separate requirements. Cite this exact development tag and commit.
Maintenance completed: all twelve merged work branches, including the release branch, were removed after ancestry or merge-tree verification. Only main remains. The prior tag and release are preserved.
v0.1.0-dev · Research infrastructure
v0.1.0-dev · Research infrastructure
First tagged development snapshot of the Node & Norm Control Evidence Corpus. GitHub prerelease; zero independently coded empirical events.
Included
| Component | What is available |
|---|---|
| Research design | Governing Charter and Agenda, targeted novelty review, comparative pilot proposal, sampling, reliability, and validation plans |
| Data contract | Explicit schemas and relational validation for events, controls, sources, evidence, claims, packets, annotations, adjudication, provenance, measures, and releases |
| Demonstrations | Three synthetic bundles and six generated review/packet views |
| Research workbench | Blank pilot instruments, an open decision register, and an AI-assisted, unrated NTSB source-reconnaissance note |
| Reuse | Deterministic synthetic exports and a proposed downstream contract for TAE, HIT, RGDS, and the future site |
Downloads
| Asset | Purpose |
|---|---|
cec-override-unresolved-0.1.0-dev.json |
Override command with unreported propagation |
cec-override-confirmed-0.1.0-dev.json |
Acknowledged stop with unreported mitigation |
cec-policy-only-0.1.0-dev.json |
Declared authority with an unreported intervention attempt |
release-manifest.json |
Source commit, input paths, output hashes, sizes, and research-status limits |
SHA256SUMS |
Checksums for the three JSON exports and manifest |
| GitHub source archives | Tracked repository files at the tagged commit, including methods, examples, schemas, and tests |
Each JSON export is a standalone synthetic bundle; identifiers are scoped to that bundle. Do not concatenate records on IDs alone. Original third-party PDFs, private material, and participant records are excluded.
Verification and reproduction
The release preparation checks cover all 22 software tests, schema and example regeneration without drift, local Markdown links and expandable sections, and synthetic export validation. These establish implementation behavior; they do not establish scientific validity.
After downloading all five assets into one directory, check their integrity:
shasum -a 256 -c SHA256SUMSTo reproduce the exports, check out v0.1.0-dev, install the pinned dependency, and run:
python3 -m pip install -r requirements.txt
python3 scripts/export.py examples/synthetic.json --output build/cec-override-unresolved-0.1.0-dev.json
python3 scripts/export.py examples/override-confirmed.json --output build/cec-override-confirmed-0.1.0-dev.json
python3 scripts/export.py examples/policy-only.json --output build/cec-policy-only-0.1.0-dev.jsonCompare the generated export hashes with SHA256SUMS. The manifest identifies the exact source commit and schema version. Checksums detect byte changes; they are not signatures or evidence of independent review.
Research status
Novelty closure, independent human coding, comparative usefulness, construct validation, holdout custody, and empirical release approval remain open. Source reconnaissance is outside the empirical corpus. The exporter and dossier renderer accept synthetic bundles only. A GitHub release does not advance these scientific gates.
The next research milestone is a staffed, rights-reviewed development pilot. See the versioned roadmap, data statement, and citation metadata.
No blanket corpus license or third-party redistribution permission is created by this prerelease.