# Submission-candidate freeze

Effective: 2026-07-15  
Status: frozen for author review and external submission preparation  
Provisional first target: *Journal of Political Economy*, subject to author approval

## Frozen artifacts

- Canonical source: `task_frontier_v3.tex`
- Canonical PDF: `task_frontier_v3.pdf`
- Source SHA-256:
  `d14a94620845a47552422c2b32c35356bfd0cb8055d1f1679d0771b04173ec0e`
- PDF SHA-256:
  `c95b7cff62a375ad456e953556f615de25cbd5071c5de8ac151f3c825739aae8`
- Verification suite: 132 passes, 0 failures
- Frozen protected-message charter: `KEY_MESSAGES.md`

The canonical, submission-package, and public-site copies of the source and PDF match
these hashes exactly.

## Do-no-harm rule

Do not make another substantive manuscript revision merely to add sophistication,
generality, a theorem, an application, a calibration, or a stronger-sounding claim.
The current internal editorial verdict is “send to external referees.” Further
internal embellishment has a real cost in length, focus, proof surface, and credibility.

The theory may be reopened only when at least one of the following is documented:

1. an actual editor or external referee raises a substantive objection;
2. a proof, definition, comparative static, citation, factual claim, or internal
   cross-reference is shown to be wrong;
3. a current public fact must be refreshed for the actual submission date; or
4. the chosen journal imposes a mandatory format or disclosure requirement that
   cannot be met administratively.

An internal preference for “more closure,” “more realism,” or “one more robustness
result” is not by itself sufficient evidence to reopen the paper.

## Permitted frozen-stage edits

The following may be added after the human author supplies and approves them:

- author names, order, affiliations, email addresses, and correspondence details;
- acknowledgments, funding, conflicts, related-paper, and AI-use disclosures;
- keywords, subject classifications, ORCID identifiers, and submission metadata;
- an approved cover letter and data/code availability statement;
- journal-mandated mechanical formatting that does not change economic content; and
- a dated factual refresh immediately before submission.

Every change to the source—even an administrative one—creates a new candidate. It must
receive a new source/PDF hash, a clean three-pass build, the full mathematical suite,
reference and citation audits, PDF structural validation, and visual inspection of the
changed pages before upload.

The executable package preflight is `submission/preflight_submission.py`. Run it only
after refreshing the package manifest; exit code 0 means the recorded technical and
author-controlled gates pass, code 1 identifies a technical/package failure, and code
2 means the technical package is sound but the author gate remains unresolved.

## External-decision rule

Internal simulated reviews are screening evidence, not acceptance evidence. Once the
author gate is resolved, the next informative test is an actual exclusive journal
submission. The persistent goal remains active until an actual journal accepts the
paper.