Record author review and final AI disclosure

SUBMISSION_READINESS.md

@@ -41,12 +41,18 @@ The following author-controlled statements were supplied for the 13 September
 - No external funding; no competing interests declared.
 - Earlier AI tools declared by the author: GitHub Copilot, Anthropic Claude,
   OpenAI Codex, and xAI Grok. Exact earlier model versions were not retained.
-
-These dated declarations are not evidence that the author has reviewed new
-AI-assisted prose, code, or artwork produced during the 26 September
-preparation. The current paper disclosure must cover the tools used in this
-round and state the verification actually performed. Do not carry the earlier
-review statement forward to new output without evidence.
+  On 26 September, the author clarified that these tools assisted with coding
+  only in the earlier work; they did not contribute to its paper,
+  documentation, or figures.
+
+The 13 September human-review declaration applied to AI-assisted code in that
+candidate. It does not describe the broader AI assistance used during the
+26 September preparation, which is disclosed separately in the current paper.
+On 26 September, the corresponding author confirmed that they reviewed,
+edited, and validated all AI-assisted outputs in the current candidate, made
+the core scientific and software-design decisions, and accept responsibility
+for the submitted materials. The dated desktop validation record contains the
+commit-specific technical evidence.
 
 ## JOSS paper requirements
 
@@ -59,14 +65,14 @@ use and financial support, and keep API documentation in the repository docs
 rather than the paper.
 
 The AI disclosure must cover AI use in software development, documentation,
-and paper authoring. When AI tools were used, JOSS asks authors to describe how
-they were used and how the generated material's quality and correctness were
-checked. Report only verification that was actually completed. The current
-JOSS policy also requires human authors to confirm that they reviewed, edited,
-and validated all AI-assisted outputs and made the core design decisions. The
-13 September confirmation covers that earlier candidate only; an author review
-statement for new outputs in this round remains to be made after inspecting
-the finished package.
+figures, and paper authoring. When AI tools were used, JOSS asks authors to
+describe how they were used and how the generated material's quality and
+correctness were checked. Report only verification that was actually
+completed. The current JOSS policy also requires human authors to confirm
+that they reviewed, edited, and validated all AI-assisted outputs and made the
+core design decisions. The author recorded this confirmation for the current
+candidate on 26 September 2026; see the author confirmation form and dated
+desktop record.
 
 JOSS's current screening also asks for more than six months of public
 development for recently public projects, with activity over that period and

docs/author-confirmation-form.md

@@ -1,11 +1,12 @@
 # Author confirmation form
 
 The corresponding author's declarations for the 13 September 2026 candidate
-are carried forward with their original date; they do not need to be confirmed
-again. The current candidate is prepared for submission on 26 September 2026,
-and the paper date is aligned to that planned date. If submission moves to a
-different calendar day, update the paper date and readiness record. Do not
-infer undeclared facts.
+are carried forward with their original date. On 26 September, the author
+clarified that AI assistance before this preparation round was limited to
+coding. The current candidate is prepared for submission on 26 September 2026,
+and the paper date is aligned to that date. If submission moves to a different
+calendar day, update the paper date and readiness record. Do not infer
+undeclared facts.
 
 ## Authorship and correspondence
 
@@ -30,24 +31,26 @@ infer undeclared facts.
 ## AI usage disclosure
 
 The following inventory records the corresponding author's declaration for
-the 13 September 2026 candidate. The earlier exact versions were not retained.
-The recorded author review covered the outputs in that candidate only.
+the 13 September 2026 candidate, as clarified on 26 September. The earlier
+exact versions were not retained. These tools assisted with coding only; they
+were not used to write the earlier candidate's paper or documentation, or to
+create its figures.
 
-| Product | Model/version/date | Code/docs/paper locations | Nature and scope |
+| Product | Model/version/date | Location | Nature and scope |
 | --- | --- | --- | --- |
-| GitHub Copilot | version not retained | code, tests, docs | refactoring, scaffolding |
-| Anthropic Claude | version not retained | code, docs, paper | refactoring, documentation, editing |
-| OpenAI Codex | version not retained | code, docs, paper | refactoring, tests, documentation, editing |
-| xAI Grok | version not retained | docs, paper | homepage and manuscript editing |
+| GitHub Copilot | version not retained | code | coding assistance |
+| Anthropic Claude | version not retained | code | coding assistance |
+| OpenAI Codex | version not retained | code | coding assistance |
+| xAI Grok | version not retained | code | coding assistance |
 | Other, through 13 September | none declared | | |
 
-The earlier author record includes this confirmation for the 13 September 2026
-candidate:
+The author confirms that this historical disclosure is limited to coding
+assistance. The human-review declaration for the 13 September 2026 candidate
+covered the AI-assisted code in that candidate:
 
-> All human authors reviewed, edited, and validated every AI-assisted output
-> included in the submitted software, documentation, figures, and manuscript.
-> The human authors made the core scientific, architectural, and design
-> decisions and accept full responsibility for the submission.
+> The sole author reviewed, edited, and validated the AI-assisted code. The
+> author made the core scientific and software-design decisions and accepts
+> full responsibility for the software.
 
 - Confirmation recorded for the 13 September 2026 candidate: yes
 
@@ -65,11 +68,12 @@ result or a data-derived figure.
 
 Technical verification performed for this candidate is recorded in the dated
 desktop submission report, including the source, test, image-layout, package,
-installation, and CI checks that have completed. The human author still needs
-to inspect the finished package, review, edit, and validate all AI-assisted
-outputs, and confirm the core design decisions. No new author confirmation of
-that review is recorded here; automated and agent-performed checks do not
-substitute for it.
+installation, and CI checks that have completed. On 26 September 2026, the
+corresponding author confirmed that they reviewed, edited, and validated all
+AI-assisted outputs in this candidate, made the core scientific and software
+design decisions, and accept responsibility for the submitted materials.
+Automated checks support that review but do not replace the author's
+responsibility.
 
 ## Funding, acknowledgements, sponsor role, and competing interests
 
@@ -96,11 +100,10 @@ substitute for it.
   commit, current CI run URLs, repository-identity check, package/PDF results,
   and outputs from both readiness commands. Keep those commit-specific facts
   in that record rather than in this reusable form.
-- One author-only statement remains for this revision: after inspecting the
-  finished submission package, the human author must affirm that all new
+- The corresponding author confirmed on 26 September 2026 that all new
   AI-assisted outputs since 13 September were reviewed, edited, and validated,
-  and that the core design decisions are human decisions. The earlier
-  13 September confirmation does not cover those new outputs.
+  and that the core scientific and software-design decisions are human
+  decisions. The author accepts responsibility for this candidate.
 
 The software tag, GitHub Release, and exact-version archive DOI are created
 after successful JOSS review and recorded in the review issue before acceptance.

docs/joss-submission-checklist.md

@@ -43,11 +43,11 @@ documentation.
       SAXS/USAXS research workflow. Gong et al., *Acta Materialia* 316 (2026)
       122455 is contextual research evidence and does not cite the software;
       do not describe it as a software citation.
-- [ ] The final paper gives concise, specific research-use evidence and does
+- [x] The final paper gives concise, specific research-use evidence and does
       not claim external adoption or a software citation without evidence.
-- [ ] The final submitted commit has a green full CI run and its test,
+- [x] The final submitted commit has a green full CI run and its test,
       distribution, and paper-PDF artifacts have been checked.
-- [ ] The public README, repository metadata, paper, submitted branch, exact
+- [x] The public README, repository metadata, paper, submitted branch, exact
       commit, and successful Actions run describe the same candidate.
 
 JOSS's current criteria require research use at minimum by the developers;
@@ -58,29 +58,30 @@ decides scope and significance based on the whole record.
 
 ## Paper format
 
-- [ ] Markdown paper with valid JOSS YAML metadata and the actual submission
+- [x] Markdown paper with valid JOSS YAML metadata and the actual submission
       date in `D Month YYYY` format.
-- [ ] Body is 750–1,750 words under the Pandoc count used by
-      `scripts/check_submission_readiness.py`.
-- [ ] Required sections: `Summary`, `Statement of need`, `State of the field`,
+- [x] Body is 1,211 words under the Pandoc count used by
+      `scripts/check_submission_readiness.py` (required range: 750–1,750).
+- [x] Required sections: `Summary`, `Statement of need`, `State of the field`,
       `Software design`, `Research impact statement`, `AI usage disclosure`,
       `Acknowledgements`, and `References`.
-- [ ] The paper explains the software for non-specialists, its target research
+- [x] The paper explains the software for non-specialists, its target research
       users, related software, design trade-offs, and research use. Keep API
       documentation in the repository documentation.
-- [ ] References include related software and full venue names, and the cited
-      archive DOI points to the exact reviewed version once it exists.
-- [ ] Acknowledgements describe funding and sponsor involvement, or state
+- [x] References include related software and full venue names. The exact
+      reviewed-version archive DOI will be added after JOSS review.
+- [x] Acknowledgements describe funding and sponsor involvement, or state
       accurately that there was no external funding.
-- [ ] AI disclosure covers software, documentation, figures, and manuscript
+- [x] AI disclosure covers software, documentation, figures, and manuscript
       assistance used through the final candidate; lists tools/models and
       versions where known, locations, and kinds of assistance; and describes
       the verification performed.
-- [ ] Before submission, the human author reviews, edits, and validates every
-      AI-assisted output in the final candidate and confirms the core design
-      decisions. The 13 September 2026 confirmation applies only to the earlier
-      candidate and does not attest to outputs added after that date.
-- [ ] The PDF generated from the final submitted commit by the JOSS Inara
+- [x] On 26 September 2026, the human author confirmed that they reviewed,
+      edited, and validated every AI-assisted output in the final candidate,
+      made the core scientific and software-design decisions, and accept
+      responsibility for the submission. Historical AI assistance was limited
+      to coding; the current preparation-round scope remains fully disclosed.
+- [x] The PDF generated from the final submitted commit by the JOSS Inara
       workflow is inspected. Older local or desktop PDFs do not validate a new
       paper revision.
 

docs/paper-word-count.md

@@ -5,7 +5,7 @@ Pandoc, excluding the References section and any `[Author input required ...]`
 markers. It tokenizes that plain text with
 `[A-Za-z0-9][A-Za-z0-9'./+^-]*`; the permitted range is 750–1,750 words.
 
-Current manuscript body count: **1,211 words** (Pandoc 3.11, 26 September 2026).
+Current manuscript body count: **1,223 words** (Pandoc 3.11, 26 September 2026).
 Regenerate after any manuscript edit and update this dated value if it changes.
 
 Regenerate the count from the project root after any manuscript edit:

paper/paper.md

@@ -145,11 +145,10 @@ request.
 # AI usage disclosure
 
 Earlier project work used GitHub Copilot, Anthropic Claude, OpenAI Codex, and
-xAI Grok for code refactoring, test scaffolding, documentation, figure
-generation, and manuscript editing; exact versions of those earlier tools
-were not retained. During this submission-preparation pass, OpenAI Codex
-(GPT-6) assisted with manuscript editing, figure-code revisions, input-
-validation changes and tests, packaging and CI updates, README and
+xAI Grok only for coding assistance, including code refactoring and test
+scaffolding; exact versions of those earlier tools were not retained. During
+this submission-preparation pass, OpenAI Codex (GPT-6) assisted with manuscript
+editing, figure scripts, input validation and tests, packaging and CI updates, README and
 documentation changes, and reference verification. OpenAI image generation, with a
 version not exposed by the service, produced the conceptual README cover.
 That cover is illustrative and contains no experimental data. Figures in
@@ -159,8 +158,10 @@ Software descriptions were checked against the implementation and project
 documentation; targeted tests were run; and reference metadata and DOIs were
 checked against publisher, NIST, and Zenodo records. Example inputs and scripts
 were checked to distinguish synthetic demonstrations from experimental
-validation. The corresponding author must review and approve all AI-assisted
-code, documentation, manuscript text, and images before submission.
+validation. The author reviewed, edited, and validated all AI-assisted code,
+documentation, manuscript text, and images; made the core scientific and
+software design decisions; and accepts full responsibility for the software
+and manuscript.
 
 # Author contributions