Phase-Reservoir Topology as a Hidden State Variable in Planetary Evolution, version 1.0.6

SPPT/ASTRA v1.0.6

Version 1.0.6 is the accessibility and public-reading successor to immutable
v1.0.5. It carries forward the admitted scientific claims, frozen benchmark
data, negative outcomes, and numerical implementation unchanged. The release
does not add empirical planetary validation, a topology-recovery theorem, a
mission-data retrieval, a DOI, or a priority claim.

Accessible reading and document repair

• A GitHub Pages reading room provides stable, versioned browser URLs for the
  preprint and technical supplement. The Pages controller reconstructs every
  edition from immutable GitHub Release assets, checks the published
  SHA256SUMS, and refuses to deploy unless main is the exact commit behind
  the current published release.
• The self-contained HTML artifacts remain downloadable for offline reading;
  the README now says explicitly that GitHub's large-file viewer does not
  render them.
• Headings, acknowledgments, and bibliography URLs reflow without horizontal
  page overflow at 320 px and 400 px. Keyboard focus is 7.14:1 against white
  and 6.58:1 against the page background; highlighted command options are
  5.93:1 against their code background.
• All nine tables have descriptive captions, scoped column headers, and
  appropriate row headers. Plots use dash patterns, markers, and hatching in
  addition to color, and their alternatives point readers to the underlying
  numeric data.
• The supplement names SciPy's scalar least-squares cost as $J$ and its
  bound-aware solver-reported optimality as $O_{\mathrm{SciPy}}$, distinct
  from the previously defined capacity vector $C$, in the optimizer-admission
  criterion.
• Tagged PDFs expose mathematics as /Formula structure with exact TeX
  /Alt and /ActualText, remove duplicate unlabeled figure wrappers,
  preserve readable bookmark spacing, and normalize compatibility ligatures so
  exact text search works. The HTML editions remain the richer native-MathML
  representation; the PDFs do not claim a native MathML expression tree.

Public metadata and contribution paths

• Seven Draft 7 JSON Schemas now describe the release specification, claim
  matrix, source inventory, runtime identity, document identity, PDF
  inspection record, and detached release identity. The verifier resolves each
  declared public URL to a shipped schema and validates every tracked record.
• CITATION.cff now binds both its top-level URL and preferred citation to the
  immutable v1.0.6 release instead of the unversioned repository root.
• Dedicated public issue forms distinguish accessibility barriers,
  reproducibility failures, and scientific corrections or counterexamples.
• Generated self-contained HTML is marked for GitHub Linguist, and the pinned
  checkout, Python, and Pages actions use their official Node 24 releases.

Scientific and release boundary

The 64-seed benchmark results are unchanged: training BIC selects the released
chain in 64/64 realizations; the overconnected triangle has smaller
post-selection held-out RMSE in 23/64; and its shortcut reaches the lower bound
in 29/64. CSV and JSON remain duplicate serializations, not independent
confirmations. The exact cross-support transfer-function counterexample and
the pointwise-only interpretation remain in force.

Canonical verification requires the exact runtime-bound command:

.\.venv\Scripts\python.exe -I -B tools\verify.py --all --workers 4

Release promotion remains fail-closed: a natural successful main run and a
natural successful annotated-tag run must bind the exact commit, tree,
manifest, fixed build epoch, and seven versioned assets. Local assets must
build byte-identically twice; uploaded assets must survive exact remote
read-back, fresh-download comparison, and GitHub attestation verification.
Pages deployment occurs only after the release is published and immutable.
The v1.0.1 and v1.0.5 releases and the unreleased v1.0.2-v1.0.4 evidence tags
remain unchanged.

Phase-Reservoir Topology as a Hidden State Variable in Planetary Evolution, version 1.0.5

SPPT/ASTRA v1.0.5

This is the publishable successor to the audited v1.0.2, v1.0.3, and v1.0.4 candidates. Their public tags are retained unchanged as negative release-process evidence, but none has a GitHub Release. Version 1.0.5 carries forward the same scientific corrections and frozen numerical results, makes the release-integrity tests independent of ambient GitHub Actions state, and supersedes v1.0.1 for current use without rewriting or replacing its immutable tag or assets.

Publication-controller evidence and test-harness correction

The v1.0.4 commit passed natural main run 30762828768, including the complete deterministic scientific and document replay, Git-archive verification, and the clean-worktree gate. Its one natural tag run, 30764066469, then passed all production tag-controller gates: exact Git and Python, restoration of annotated tag object b163ccfed005ffcbb4b15a49644ea27e09d5a7ba, hash-locked dependency installation, independent post-install tag verification, and pinned Chromium installation.

The tag run stopped at the first Pytest pass inside canonical verification. A unit test intentionally simulates an incomplete tag event by removing GITHUB_SHA and expecting the missing tag-ref check to fail first. On an actual tag runner, however, that test inherited the real GITHUB_REF_TYPE, GITHUB_REF_NAME, and GITHUB_REF; its v1.0.1 fixture therefore encountered the ambient v1.0.4 name before reaching the intended missing-ref condition. The run recorded 203 passing tests and one harness failure. Scientific replay did not start, no v1.0.4 GitHub Release or draft was created, and no assets were uploaded.

Version 1.0.5 makes the test boundary hermetic. Before every release-integrity test, an automatic fixture clears all ten GitHub context variables read by production code:

GITHUB_ACTIONS
GITHUB_EVENT_NAME
GITHUB_EVENT_PATH
GITHUB_REF
GITHUB_REF_NAME
GITHUB_REF_TYPE
GITHUB_REPOSITORY
GITHUB_REPOSITORY_ID
GITHUB_SERVER_URL
GITHUB_SHA

Each simulated scenario must then declare its complete context. The previously failing test and the complete suite pass with a real v1.0.4 tag context injected outside Pytest. The production controller and workflow are unchanged: their v1.0.4 restore and post-install identity gates already passed. The controller continues to model the typed chain

tag ref -> annotated tag object -> direct commit target

and binds the creation payload's after object to the annotated tag while independently binding GITHUB_SHA and HEAD to the direct commit.

Scientific scope carried forward unchanged

The package remains a not-peer-reviewed perspective and mathematical framework with reduced synthetic demonstrations. It contains no empirical planetary validation, general topology-recovery theorem, mission-data retrieval, or claim that numerical agreement constitutes proof.

The v1.0.2 scientific corrections remain in force:

• Figure 5 applies its unit internal power consistently to the deep reservoir in both transient and equilibrium calculations.
• Exact algebraic-statistical counterexamples show that distinct positive reservoir-network supports can share the same surface transfer function under the declared single-port design.
• The released generating chain remains distinguishable only pointwise in the declared nonnegative candidate domain; the work does not establish family-wide physical-topology identification.
• The one-frequency demonstration uses amplitude alone and illustrates conditioning and finite-grid localization, not noise resistance or a minimal-data theorem.
• Numerical routines preserve representable slow modes, reject unresolved finite-scale inputs, and fail closed on non-finite, asymmetric, or shape-invalid inputs.

The frozen ensemble results are unchanged:

• Training BIC selects the released generating chain in 64/64 realizations.
• The overconnected triangle has lower post-selection held-out RMSE in 23/64 realizations.
• The triangle shortcut reaches its declared lower bound in 29/64 realizations.
• The 20-start design remains post-audit regression evidence rather than untouched, blinded, or external evaluation.

The CSV and JSON are duplicate serializations of the same synthetic evidence, not independent confirmations.

Verification, identity, and preservation

Canonical verification requires the exact runtime-bound command:

py -3.12 -I -B tools\verify.py --all --workers 4

Release identity binds the annotated tag object, direct commit, tree, tracked manifest, fixed build epoch, and exactly seven versioned assets. Promotion additionally requires one natural successful main run, one natural successful tag run, two identical local asset builds, exact remote read-back, fresh-download equality, and post-publication attestation verification.

v1.0.1 remains immutable. The public, unreleased v1.0.2, v1.0.3, and v1.0.4 tags remain unchanged and have no GitHub Releases. Original software remains MIT licensed; original manuscript text, documentation, figures, and generated results remain CC BY 4.0. Publication remains GitHub-only; no Zenodo ingestion or release DOI is claimed.

Phase-Reservoir Topology as a Hidden State Variable in Planetary Evolution, version 1.0.1

SPPT/ASTRA v1.0.1

This is the first public versioned reference package for Solar-Planetary Phase-Partition Theory (SPPT) and its Astronomical State-Topology and Reservoir Analysis (ASTRA) inference layer.

Scope and evidence

The release is a not-peer-reviewed perspective and mathematical framework with reduced synthetic demonstrations. Admitted analytical results are source-asserted and hand-checked; the reduced calculations and frozen synthetic benchmarks are mechanically replayed or independently reproduced as recorded in CLAIM_MATRIX.json. Numerical agreement is not proof, and the package contains no empirical planetary validation or claim of priority.

Contents and reproduction

The seven release assets contain tagged PDFs, self-contained accessible HTML editions, a deterministic source/reproducibility archive, checksums, and a detached release identity. Reproduce the tracked package on the frozen Windows x86-64 reference runtime with:

py -3.12 -c "import platform; assert platform.python_version() == '3.12.10', platform.python_version()"
py -3.12 -m venv .venv
.\.venv\Scripts\python.exe -m pip install --require-hashes -r requirements-lock.txt
.\.venv\Scripts\python.exe -m playwright install chromium
.\.venv\Scripts\python.exe tools\verify.py --all --workers 4

The release gate executes 69 discovered unit, invariant, evidence-boundary, document-contract, optimizer, runtime-identity, and negative release-integrity tests; strict Ruff and mypy checks; CFF schema validation; privacy, path, metadata, citation-key, license, HTML, PDF, cache-boundary, and archive checks; the complete frozen numerical replay; figure regeneration; two byte-identity document builds; and fresh-clone, linked-worktree, and extracted-git-archive verification. The 64-seed ensemble uses frozen seeds, 20 distinct generic multistarts per fitted family, full per-start and per-endpoint convergence diagnostics, a fail-closed materially-better-endpoint check, bound flags, and preserved negative outcomes. The 20-start design was strengthened during release audit after this same benchmark exposed a missed endpoint under the earlier 12-start design, so the reruns are regression evidence rather than untouched evaluation. Byte identity is scoped to the exercised outputs under CPython 3.12.10, Git for Windows 2.55.0.windows.3, the probed single-thread NumPy and SciPy OpenBLAS Haswell kernels, and the recorded Matplotlib-distributed DejaVu/STIX font inputs on compatible Windows x86-64 hardware.

Synthetic-data status and limitations

Training BIC selects the generating chain in 64/64 frozen realizations. The post-selection held-out comparison preserves the negative result that the overconnected triangle has lower RMSE in 23/64 realizations; its shortcut reaches the declared lower bound in 29/64. The data are transparent, synthetic, deliberately favorable demonstrations—not blinded evidence, an external validation, a general false-positive-rate estimate, or predictive superiority over planetary observations.

The hybrid topology syntax does not establish general existence, uniqueness, reset-map closure, simultaneous-guard resolution, or non-Zeno behavior. The static and weak-cut results retain their stated positivity, connectivity, injectivity, equilibrium, and domain hypotheses. Electrochemical terms describe supplied free-energy conversion, not latent heat or spontaneous planetary sequestration.

The manuscript also contains a proposed-only typed-layer research outlook added after review of four 2026 primary studies. One diamond inclusion with no observed present fracture or exterior connection is a qualified present-isolation example; permeability and past exchange were not measured. The moonlight, quantum-sampling, and xenophagocytosis studies are structural analogies only. None validates SPPT or changes the implemented physical core. A universal selectivity--throughput law is explicitly deferred, and a new no-go rule forbids substituting an observation, semantic association, biological recognition signal, or certificate for a physical transport edge without a constitutive bridge.

Conceptual provenance records the author-reported dream/collage/ChatGPT origin of the cold-trap/Saturn idea and subsequent OpenAI language-model assistance. Neither the dream, the excluded collage, the unpublished chat, nor model output is evidence. No third-party article text or figure is redistributed.

Licensing

Original software is MIT licensed. Original manuscript text, documentation, figures, and generated results are CC BY 4.0. Bibliographic references and third-party components retain their own terms, including the DejaVu/Bitstream/Arev font terms and the SIL Open Font License 1.1 for STIX. See LICENSE_MAP.md and THIRD_PARTY_NOTICES.md in the source archive.

Preservation

This release is preserved on GitHub only. Zenodo integration was disabled before publication; no Zenodo record or DOI is requested for v1.0.1.