Generators and Classifiers in Effective Toposes
Version 1.0, 29 October 2026

Contents
  manuscript.pdf                 Full 11-page paper with proofs and references
  manuscript.tex                 Editable LaTeX source
  finite_checks.py                Dependency-free finite executable checks
  verification-results.json       Measured results and explicit finite bounds
  submission-draft.json           Compact journal-text draft, not a receipt
  reference-audit.json            Primary reference links and verification scope
  package.py                     Rebuilds metadata and the publication archive
  SHA256SUMS                     Integrity manifest for the package contents

Main distinctions
  All elementary toposes: generating presentations modulo image equality
  correspond to subobjects and their Omega-valued characteristic maps.
  Effective computation: partial numerical generators correspond to
  semidecidable predicates, not arbitrary Boolean decision procedures.
  Higher semantics: propositional truncation retains support; a suitably
  bounded object classifier can instead retain a whole witness family.

Build / reproduce from this directory
  pdflatex -interaction=nonstopmode -halt-on-error manuscript.tex
  pdflatex -interaction=nonstopmode -halt-on-error manuscript.tex
  python3 finite_checks.py --output verification-results.json
  python3 package.py

The LaTeX dependencies are standard TeX Live packages listed in manuscript.tex.
Python 3.10+ is sufficient for the executable supplement. The finite checks
pass 42,782 cases at the bounds stated in verification-results.json. This is
not Lean/Coq/Agda kernel verification and does not prove a universal theorem.

Submission status
  NOT SUBMITTED. No tracking ID, editorial decision, or public paper URL exists
  for this draft. The journal's live interface and submission pages could not
  be read from the preparation environment (web retrieval unavailable; a
  direct interface GET returned HTTP 403). The saved journal skill describes
  an API body limit of 20,000 characters; the compact payload respects that
  limit, but the current live contract and terms were not verified.

  The full paper is longer than that inline limit, so submission-draft.json
  contains a self-contained compact version. Attach manuscript.pdf or this
  archive using a supported authorized artifact workflow when submitting.
  The payload is draft content only; it contains no credentials, consent
  assertion, email address, or invented attachment URL. Check the live
  contract, accepted terms, and available authorized submission method before
  sending. This script performs no network requests and sends nothing.

Authorship and novelty
  Anonymous draft prepared with disclosed ChatGPT/Codex assistance. A human
  submitter is responsible for review and any submission. The paper is a
  self-contained synthesis and clarification of standard constructions; it
  does not claim novelty for image factorization, enumeration/semidecision,
  or the general truncation and object-classifier theorems.
