# Revision change log

## Deliverables

- Revised: `output/article.md`, `output/metadata.json`, `output/editor-notes.md`.
- Preserved initial files: `output/original-submission/` (article, metadata, notes and original validation script), with a SHA-256 manifest.
- Supporting scripts: `output/check_package.py` (current validation), `output/review_package.py` (archive/read-input step), `output/apply_review.py` (revision step). The revision-step script is a one-time working artifact; do not rerun it on the revised article.

## Changes

1. Added Canada/United States eligibility to the excerpt and the scenario year to the article lead.
2. Changed “controlled trial” to “controlled evaluation” and roadmap “promise” to “plan”, avoiding implications unsupported by the announcement.
3. Made the annual-offer evidence limit explicit and added the pre-tax qualifier to the search description and closing recommendation.
4. Expanded practical evaluation advice on checking time, representative cases and upload costs. No hands-on results are claimed.
5. Extended validation to plain prose word count, citation definitions/source headings and archive integrity. No arithmetic corrections were needed.

## Checks completed

Re-read all seven supplied inputs and reviewed the revised files. Confirmed the USD 29 price correction takes precedence; USD 87 minimum, USD 99 with OCR and 60% vendor draft-stage reduction are correctly calculated and qualified. Confirmed unsupported internal-note leads remain excluded and explained in notes.

Ran `output/check_package.py` successfully: prose 630 words (671 including section headings/citation markers; 708 including title and sources); headline 58 characters; excerpt 31 words; search description 151 characters. JSON fields, matching headline, source markers/headings and preserved-file hashes passed.

## Limitations and handoff

No product access, hands-on testing, external research or independent verification beyond the supplied packet. Buyer workflow and privacy suitability remain untested, as stated in the article/notes. Estimated final human copy-review time: 5–10 minutes; this is not a measured result. No publication work is required.
