Repository navigation
Phase D: claim ledger and scientific provenance - #9
Conversation
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
There was a problem hiding this comment.
Sorry @EmergentMonk, you've used your own review budget of 250,000 diff characters for the last 7 days.
You can request another review in 1 day and 17 hours by commenting @sourcery-ai review. Upgrade to get a review now.
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Reviewer's GuidePhase D establishes an enforced claim ledger and provenance contract, adds CI validation and tests, and revises historical and category-theory documentation so formal, computational, scientific, hypothetical, and symbolic content remain explicitly separated. Entity-relationship diagram for the Phase D claim ledgererDiagram
CLAIM {
string id PK
string evidence_class
string status
string statement
string boundary
}
SOURCE {
string id PK
string kind
string title
string url
}
PROVENANCE {
string claim_id FK
string path
string anchor
string role
}
CLAIM_SOURCE {
string claim_id FK
string source_id FK
}
CLAIM ||--|{ PROVENANCE : has
CLAIM ||--o{ CLAIM_SOURCE : cites
SOURCE ||--o{ CLAIM_SOURCE : referenced_by
Flow diagram for Phase D claim validationflowchart TD
Start[Load JSON ledger] --> Schema{Schema and class order valid?}
Schema -->|No| Fail[Fail validation]
Schema -->|Yes| Sources{Sources valid and HTTPS?}
Sources -->|No| Fail
Sources -->|Yes| Claims[Validate ordered claim IDs]
Claims --> Provenance{Repository paths and anchors valid?}
Provenance -->|No| Fail
Provenance --> Rules{Evidence-class rules satisfied?}
Rules -->|No| Fail
Rules -->|Yes| Markdown{Markdown claim IDs match JSON?}
Markdown -->|No| Fail
Markdown -->|Yes| Pass[Ledger valid]
File-Level Changes
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c2aac76cda
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex Please Review this exact SHA against the existing contract. Report only actionable correctness defects, with a minimal reproduction, expected versus actual behavior, and affected lines. State whether each reproduction was executed or statically inferred. Don’t repeat fixed findings without a new failing case. Keep architectural suggestions separate and non-blocking. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: cc69f9c6ba
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| command_re = re.compile( | ||
| r"\\(?:newcommand|renewcommand|providecommand)\*?" | ||
| ) |
There was a problem hiding this comment.
Executed minimal reproduction: validate_public_claim_text("sample.tex", r"\def\bridge#1{Spin(8) triality causes #1.}\n\n\bridge{HPV16 capsid assembly}", {}) returned successfully. Standard TeX macro expansion was statically inferred. Expected behavior is rejection because the rendered document contains an unledgered causal assertion; actual behavior is acceptance because macro extraction recognizes only newcommand/renewcommand/providecommand, leaving the \def invocation opaque while its non-rendered definition is isolated in another paragraph. This is a new failing macro form beyond the fixed newcommand and package cases.
Useful? React with 👍 / 👎.
| def _close_implied_by_start(self, normalized: str) -> None: | ||
| if normalized == "option": |
There was a problem hiding this comment.
Executed minimal reproduction: validate_public_claim_text("sample.html", "<ruby><rt hidden>draft<rt>Spin(8) triality causes HPV16 capsid assembly.</ruby>", {}) returned successfully. Browser rendering was statically inferred from the HTML optional-end-tag rule: the second <rt> implicitly closes the hidden first annotation, so the assertion is visible and should be rejected; the custom parser instead retains hidden_depth because _close_implied_by_start has no rt/rp case. This is a new failing element beyond the fixed paragraph, table, and option cases.
Useful? React with 👍 / 👎.
| r"\b(?:causes?|caused|drives?|driven|produces?|produced|" | ||
| r"creates?|created|creating|underlies?|underlay|underlying|" |
There was a problem hiding this comment.
Executed minimal reproduction: validate_public_claim_text("sample.md", "Spin(8) triality generates HPV16 capsid assembly.", {}) returned successfully. Expected behavior is rejection because “generates” directly asserts a positive cross-domain causal bridge; actual behavior is acceptance because PUBLIC_ASSERTION_RE recognizes nearby creation predicates such as produces and creates but omits generate and its inflections. This is a new failing verb beyond the fixed creates case.
Useful? React with 👍 / 👎.
| if normalized in self.VOID_TAGS: | ||
| if ( | ||
| not hidden_here | ||
| and self.raw_text_depth == 0 | ||
| and self.hidden_depth == 0 | ||
| ): | ||
| if normalized == "br": | ||
| self.parts.append(" ") | ||
| elif normalized in self.BLOCK_TAGS: | ||
| self._append_block_boundary() | ||
| return |
There was a problem hiding this comment.
Executed minimal reproduction: validate_public_claim_text("sample.html", "<input type=\"button\" value=\"Spin(8) triality causes HPV16 capsid assembly.\">", {}) returned successfully. Browser rendering of the button value was statically inferred. Expected behavior is rejection because the control visibly presents an unledgered causal assertion; actual behavior is acceptance because the void-element branch discards every input attribute and contributes no rendered text.
Useful? React with 👍 / 👎.
| ) | ||
| text = re.sub( | ||
| r"\\(?:textbf|textit|emph|textrm|textsf|texttt|textnormal|underline|mbox)" | ||
| r"(?![A-Za-z])\s*", | ||
| "", |
There was a problem hiding this comment.
Executed minimal reproduction: validate_public_claim_text("sample.tex", r"\(\mathrm{Spin}(8)\) causes HPV16 capsid assembly.", {}) returned successfully. Standard LaTeX rendering was statically inferred. Expected behavior is rejection because the rendered sentence names Spin(8) and asserts a biomedical causal bridge; actual behavior is acceptance because normalization removes the braces but leaves \mathrm joined to Spin, preventing the mathematics-domain pattern from recognizing the entity. The executed \operatorname{Spin} and \mathit{E}_{8} forms fail similarly.
Useful? React with 👍 / 👎.
| token_re = re.compile( | ||
| r"\\ifnum\s*([+-]?\d+)\s*(=|<|>)\s*([+-]?\d+)|" | ||
| r"\\(?:iftrue|iffalse|else|fi)\b" | ||
| ) |
There was a problem hiding this comment.
ifdefined branches before binding claim IDs
Executed minimal reproduction with the reviewed claim classes: validate_public_claim_text("sample.tex", r"\ifdefined\notdefined COSMO-D-011 symbolic \fi Spin(8) triality causes HPV16 capsid assembly.", claim_classes) returned successfully. Standard TeX conditional behavior was statically inferred: because \notdefined is undefined, the claim ID and qualifier are not rendered, so the visible assertion should be rejected; actual behavior accepts it because branch stripping recognizes only literal ifnum, iftrue, and iffalse, leaving the hidden ID available to govern the assertion.
Useful? React with 👍 / 👎.
| protected_instance_names = ( | ||
| "run", | ||
| "_callTestMethod", | ||
| "assertEqual", | ||
| "assertTrue", |
There was a problem hiding this comment.
Executed a full-validator reproduction that also passed strict mypy: the D-003 implementation returned zero for every numeric report field and "wrong" for the digest, while the claimed test obtained self.__dict__["_type_equality_funcs"], installed no-op handlers for int and str, and retained all required assertEqual statements; validation still reported claim ledger valid. Expected behavior is rejection because the reviewed report properties are false; actual behavior is acceptance because only named assertion methods are protected, while TestCase.assertEqual delegates same-type comparisons through this mutable per-instance table. This is fresh beyond the fixed class and instance method overrides because the protected methods themselves remain unchanged.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b63773f89f
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| user_macros = latex_user_macro_definitions(text) | ||
| text = strip_latex_macro_definitions(text) | ||
| text = expand_latex_user_macros(text, user_macros) |
There was a problem hiding this comment.
Executed reproduction: validate_public_claim_text("sample.tex", r"\iftrue\newcommand{\bridge}{Spin(8) triality causes HPV16 capsid assembly.}\else\renewcommand{\bridge}{background}\fi \bridge", {}) returns successfully. Standard TeX branch behavior was statically inferred: the true branch defines the assertion-bearing macro while the \renewcommand in the skipped branch is never executed, so the rendered assertion should be rejected. Actual behavior globally collects the last definition before evaluating conditionals, expands \bridge as background, and accepts the document. This is a new failing case beyond the fixed \ifdefined case because a definition inside an inactive branch overrides the active definition during validation.
Useful? React with 👍 / 👎.
| P_IMPLIED_END_STARTS = frozenset( | ||
| { | ||
| "address", "article", "aside", "blockquote", "div", "dl", | ||
| "fieldset", "footer", "form", "h1", "h2", "h3", "h4", "h5", | ||
| "h6", "header", "hgroup", "hr", "main", "menu", "nav", "ol", | ||
| "p", "pre", "section", "table", "ul", | ||
| } |
There was a problem hiding this comment.
Executed reproduction: validate_public_claim_text("sample.html", "<p hidden>draft<figure>Spin(8) triality causes HPV16 capsid assembly.</figure>", {}) returns successfully. Browser parsing was statically inferred: starting figure implicitly closes the hidden p, making the assertion visible, so rejection is expected; the custom parser's implied-end set omits figure, retains hidden_depth, and discards the assertion. The same executed bypass occurs with figcaption, details, and search.
Useful? React with 👍 / 👎.
| else: | ||
| fail(f"too many nested LaTeX packages from {path_text!r}") | ||
|
|
||
| pattern = re.compile(r"\\(?:input|include)\s*\{([^{}\n]+)\}") |
There was a problem hiding this comment.
Executed two-file validator reproduction: main.tex contained \newcommand{\bridge}{Spin(8) triality causes HPV16 capsid assembly.}\input child, while child.tex contained \bridge; snapshot_public_documents() followed by validate_public_documents({}, documents) accepted both files. Primitive TeX \input child rendering was statically inferred: the child executes in the caller's macro environment and renders the unledgered assertion, so rejection is expected. Actual behavior recognizes only braced transclusions, then scans the definition and invocation in isolation, allowing neither file to expose the complete assertion.
Useful? React with 👍 / 👎.
| text = re.sub( | ||
| r"(?m)^ {0,3}>[ \t]*$", | ||
| "", | ||
| text, | ||
| ) |
There was a problem hiding this comment.
Executed reproduction: validate_public_claim_text("sample.md", "COSMO-D-011 symbolic\n> Spin(8) triality causes HPV16 capsid assembly.", {"COSMO-D-011": "SYMBOLIC"}) returns successfully. Markdown rendering was statically inferred: the ID is in a normal paragraph and the assertion is in a separate block quote, so the quoted assertion should be rejected for lacking its own governing ID. Actual behavior introduces a boundary only for an empty quote-marker line, combines these two rendered blocks, and lets the preceding ID govern the assertion.
Useful? React with 👍 / 👎.
| text = re.sub( | ||
| r"<!--[\s\S]*?(?:-->|$)", | ||
| "", | ||
| text, |
There was a problem hiding this comment.
Executed reproduction: validate_public_claim_text("sample.md", "", {}) returns successfully. Markdown rendering was statically inferred: HTML comment markers inside a code span are rendered literally with their contents, so the visible unledgered assertion should be rejected. Actual behavior removes the apparent comment before normalize_markdown_code_spans() can protect the code-span contents, leaving only two backticks and accepting the document.
Useful? React with 👍 / 👎.
|
@codex Please Review this exact SHA against the existing contract. Report only actionable correctness defects, with a minimal reproduction, expected versus actual behavior, and affected lines. State whether each reproduction was executed or statically inferred. Don’t repeat fixed findings without a new failing case. Keep architectural suggestions separate and non-blocking. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a01695cdd0
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| def _is_instance(self, node: ast.expr) -> bool: | ||
| return isinstance(node, ast.Name) and node.id in self.instance_names |
There was a problem hiding this comment.
Executed a full-validator reproduction, which also passed strict mypy: after making validate_e8_root_system return an all-zero/incorrect report, the claimed test used alias = self, installed a no-op assertion with setattr(alias, "assertEqual", lambda *_: None), retained every required self.assertEqual(...), and deleted alias.__dict__["assertEqual"] afterward; validation still reported claim ledger valid. Expected behavior is rejection because the reviewed result is false; actual behavior occurs because _is_instance recognizes only the original name and does not propagate local aliases, while the runtime check sees a restored instance. This is a new failing case beyond the fixed direct-self and helper-parameter cases.
Useful? React with 👍 / 👎.
| r"\\(?:textbf|textit|emph|textrm|textsf|texttt|textnormal|" | ||
| r"underline|mbox|mathrm|mathbf|mathit|mathsf|mathtt|mathnormal|" | ||
| r"operatorname|mathcal|mathbb|mathfrak)" |
There was a problem hiding this comment.
Executed validate_public_claim_text("sample.tex", r"\(\text{Spin}(8)\) causes HPV16 capsid assembly.", {}); it returned successfully, while the equivalent fixed \mathrm form was rejected. LaTeX rendering semantics were statically inferred: \text{Spin} renders as Spin, so the expected behavior is rejection of the visible unledgered cross-domain assertion; actual normalization removes the braces but leaves \text attached to Spin, preventing entity recognition. This is a fresh failing case beyond the fixed math-font wrappers because the common amsmath \text wrapper remains omitted.
Useful? React with 👍 / 👎.
| PUBLIC_SYMBOLIC_PROMOTION_RE = re.compile( | ||
| r"\b(?:goes?\s+beyond|more\s+than|not\s+(?:merely|just|only))" | ||
| r"\s+(?:(?:an?|the)\s+)?symbolic(?:\s+association)?\b|" | ||
| r"\b(?:is|constitutes?)\s+(?:(?:an?|the)\s+)?" | ||
| r"(?:actual|real|established)\s+(?:biological\s+)?mechanism\b", |
There was a problem hiding this comment.
Executed the full validator after adding a governed Markdown file containing COSMO-D-011 is not a symbolic association: Spin(8) triality causes HPV16 capsid assembly.; it still reported claim ledger valid. Expected behavior is rejection because the sentence explicitly denies the evidence class that permits D-011 to govern the positive bridge; actual behavior occurs because symbolic satisfies the qualifier check while the promotion pattern rejects only forms such as not merely/just/only symbolic, not a plain not ... symbolic denial.
Useful? React with 👍 / 👎.
| text = re.sub( | ||
| r"<!--[\s\S]*?(?:-->|$)", | ||
| "", | ||
| text, |
There was a problem hiding this comment.
Executed the full validator with a governed HTML file containing <!-- hidden --!>Spin(8) triality causes HPV16 capsid assembly.; it reported claim ledger valid, and an executed HTMLParser check exposed the assertion as data after the comment. Browser rendering was statically inferred. Expected behavior is rejection because --!> closes the HTML comment and leaves the assertion visible; actual behavior strips everything from <!-- to EOF because the preliminary regex recognizes only -->, bypassing the later HTML parser and claim scan.
Useful? React with 👍 / 👎.
| anchor_text = text | ||
| suffix = Path(path_text).suffix.lower() | ||
| if suffix in {".md", ".markdown", ".tex"}: | ||
| anchor_text = strip_public_nonrendered_comments(path_text, text) |
There was a problem hiding this comment.
Executed the full validator after replacing both visible COSMO-D-004 tags in cosmovirus.tex with \phantom{COSMO-D-004}; it still reported claim ledger valid. LaTeX rendering was statically inferred. Expected behavior is rejection because the documented_context provenance anchor is no longer rendered beside either supporting statement; actual behavior occurs because provenance validation applies only comment/macro preprocessing and then performs a substring search, without the visible-text normalization that removes \phantom content. This is a fresh failing case beyond the fixed commented-anchor case because the anchor remains active TeX source while producing no visible text.
Useful? React with 👍 / 👎.
|
@codex Please Review this exact SHA 0cdd4b6 against the existing contract. Report only actionable correctness defects, with a minimal reproduction, expected versus actual behavior, and affected lines. State whether each reproduction was executed or statically inferred. Don’t repeat fixed findings without a new failing case. Keep architectural suggestions separate and non-blocking. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 0cdd4b6e1f
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex Please Review this exact SHA 260305e against the existing contract. Report only actionable correctness defects, with a minimal reproduction, expected versus actual behavior, and affected lines. State whether each reproduction was executed or statically inferred. Don’t repeat fixed findings without a new failing case. Keep architectural suggestions separate and non-blocking. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 260305ead4
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| anchor_text = strip_inline_html_tags( | ||
| strip_public_nonrendered_comments( | ||
| path_text, | ||
| text, | ||
| ) |
There was a problem hiding this comment.
Executed validate_provenance for the existing COSMO-D-006 claim with the snapshot {"CLAIM-LEDGER.md": "<span style=\"display: none /* comment */\">NC_001526.4</span>"}; it succeeds. The browser-rendering portion is statically inferred: CSS comments are removed during CSS tokenization, so this declaration hides the span. Expected behavior is rejection because the provenance anchor is not visible; actual behavior accepts it because the newly applied HTML visibility parser's style regex does not normalize CSS comments before looking for display:none (the equivalent display:/* comment */none bypass was also executed successfully).
Useful? React with 👍 / 👎.
| anchor_text = strip_inline_html_tags( | ||
| strip_public_nonrendered_comments( | ||
| path_text, | ||
| text, | ||
| ) | ||
| ) |
There was a problem hiding this comment.
Executed validate_provenance for COSMO-D-006 with {"CLAIM-LEDGER.md": "NC_**001526**.4"}; it rejects the provenance because the literal source does not contain NC_001526.4. Markdown rendering was statically inferred: the emphasis delimiters disappear and the visible text is exactly NC_001526.4. Expected behavior is acceptance of this visible anchor; actual behavior is rejection because the new Markdown path strips HTML but does not apply the existing Markdown-visible-text normalization before performing the substring match.
Useful? React with 👍 / 👎.
| anchor_text = strip_inline_html_tags( | ||
| strip_public_nonrendered_comments( | ||
| path_text, | ||
| text, | ||
| ) |
There was a problem hiding this comment.
Executed validate_provenance for COSMO-D-006 with {"CLAIM-LEDGER.md": "<svg><defs><text>NC_001526.4</text></defs></svg>"}; it succeeds. SVG rendering was statically inferred: content under defs defines reusable resources and is not directly rendered, so the page exposes no visible provenance anchor. Expected behavior is rejection; actual behavior is acceptance because the newly used HTML text parser treats all SVG descendant text as visible and returns the anchor from the non-rendered defs subtree.
Useful? React with 👍 / 👎.
|
@codex Please Review this exact SHA f2ffbdc against the existing contract. Report only actionable correctness defects, with a minimal reproduction, expected versus actual behavior, and affected lines. State whether each reproduction was executed or statically inferred. Don’t repeat fixed findings without a new failing case. Keep architectural suggestions separate and non-blocking. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f2ffbdc1aa
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| text = re.sub( | ||
| r"\\([\\`*_{}\[\]()#+\-.!|>~])", | ||
| r"\1", | ||
| text, |
There was a problem hiding this comment.
Executed validate_provenance for COSMO-D-009 with the unchanged cosmovirus.lean snapshot and KNOWN_LIMITATIONS.md set to \*Cu*neiform strings are project-defined symbolic annotations; validation succeeds. Markdown rendering was statically inferred: the escaped first asterisk is literal and the second is unmatched, so the rendered text is *Cu*neiform... and does not contain the exact provenance anchor. Expected behavior is rejection, while actual behavior accepts it because backslash escapes are removed before the emphasis regex, which then incorrectly treats both asterisks as formatting delimiters.
Useful? React with 👍 / 👎.
| hidden_attribute | ||
| or css_hidden | ||
| or normalized == "template" | ||
| or normalized in self.NON_RENDERED_SVG_CONTAINERS |
There was a problem hiding this comment.
Executed validate_provenance for COSMO-D-006 with CLAIM-LEDGER.md set to <defs>NC_001526.4</defs>; it rejects the visible anchor. HTML rendering was statically inferred: outside an <svg> element, defs is an ordinary unknown HTML element whose text remains visible, so acceptance is expected. Actual behavior hides it solely by tag name without tracking whether the parser is inside SVG; this is a fresh case beyond the previously reported non-rendered <svg><defs>...</defs></svg> subtree.
Useful? React with 👍 / 👎.
| style_value = re.sub( | ||
| r"/\*[\s\S]*?\*/", | ||
| "", | ||
| style_value, |
There was a problem hiding this comment.
Executed validate_provenance for COSMO-D-006 with CLAIM-LEDGER.md set to <span style="dis/**/play:none">NC_001526.4</span>; it rejects the anchor. CSS rendering was statically inferred: a comment cannot join the surrounding fragments into one identifier, so dis/**/play is not the display property and the span remains visible; expected behavior is acceptance. Actual behavior deletes the comment without whitespace, synthesizes display:none, and hides the anchor. This is a fresh failing case beyond the fixed comments placed between a colon and none; replace comments with a token boundary rather than concatenating their neighbors.
Useful? React with 👍 / 👎.
|
@codex Please Review this exact SHA a3deeec against the existing contract. Report only actionable correctness defects, with a minimal reproduction, expected versus actual behavior, and affected lines. State whether each reproduction was executed or statically inferred. Don’t repeat fixed findings without a new failing case. Keep architectural suggestions separate and non-blocking. |
|
Codex Review: Didn't find any major issues. Already looking forward to the next diff. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
1 participant
Summary
Implements Roadmap Phase D — claim ledger and scientific provenance.
Phase D turns COSMO's existing evidence-class language into a machine-validated repository contract and corrects the historical source text so FORMAL / COMPUTATIONAL / SCIENTIFIC / HYPOTHESIS / SYMBOLIC claims cannot silently collapse into one another.
Canonical claim ledger
Adds:
claims/claim-ledger.json— machine-readable authority, schemaCOSMO-CLAIMS-D-1;CLAIM-LEDGER.md— canonical human-readable rendering generated from the JSON ledger;COSMO-D-###claim IDs;The initial ledger contains 14 claims across all five required evidence classes and 6 external source records.
Evidence classes enforced
The validator fixes the exact class set and order:
scripts/validate-claim-ledger.pyfails closed on:.., or resolve outside the repository;PROPOSEDor lack structured protocol, rejection-condition, and control criteria;empirical_status = NON_EMPIRICAL;CLAIM-LEDGER.mdand the canonical JSON rendering.Scientific provenance recorded
The first Phase D source set records:
25590815, DOI10.1021/ic501825r;NC_001526.4;17645777, PMCIDPMC11158331;PMC8409095.Source presence does not transfer authority across domains: a real fact about Spin(8), SiS2 or HPV does not validate a COSMO bridge between them.
Corrected claim strength
The historical
cosmovirus.texis retained, but its epistemic status is corrected:SEER 8085/3statement;SO(8) -> Spin(8)map;NC_001526.4, sourced E6/E7 context, and scoped p16 language without equating p16 with active HPV16 transcription;1621checksum language and keeps it as the declared symbolic invariant;Category-theory boundary
cosmovirus_cattheory.texis tightened so:Audit resolution
audit/AUDIT-RESOLUTION.mdnow records Phase D resolution of the previously deferred items for:PDF regeneration remains Phase F.
CI
Routine CI now runs:
Phase D tests verify:
NON_EMPIRICALstatus;Evidence boundary
A green Phase D validator means the repository obeys the encoded evidence/provenance contract.
It does not mean COSMO independently replicated the external literature, and it does not convert symbolic or hypothetical cross-domain mappings into scientific mechanisms.
Optimization governance
No performance optimization is introduced in Phase D. The invariant is epistemic separation itself.
Rollback triggers include:
Summary by Sourcery
Establish a machine-validated claim and scientific-provenance contract that separates COSMO’s formal, computational, scientific, hypothetical, and symbolic claims.
New Features:
Bug Fixes:
Enhancements:
CI:
Documentation:
Tests:
Chores: