# V2 comparison ordinal correction · 2026-09-08

This receipt closes the focused TODO for six V2 column comparisons whose
`01`–`04` markers implied an order that their content did not define. Each
authored slide now sets `numbered: false`; the shared column renderer and all
other slides retain their existing behavior.

## Semantic boundary

The corrected independent comparisons are:

- `v2-promise-audience` and `v2-promise-cross-platform`;
- `v2-surfaces-game-menu`, `v2-surfaces-accessibility-bridges`,
  `v2-surfaces-results-operations` and
  `v2-surfaces-companion-integrity`.

The actual progressions remain numbered:

- `v2-promise-complete-content`;
- `v2-promise-evidence`;
- `v2-surfaces-broadcast-providers`.

The source-level regression verifies that the six comparisons use the column
layout, opt out of numbering and render no step marker. Its mutation restores
numbering and must produce four markers. The same test requires four markers on
each real progression. The six current orientation-review fingerprints and two
current V2 source-unit assignment bindings were refreshed against the exact
assembled slides; no pinned source, explanation, evidence span or historical
source-application record changed.

## Browser and PDF review

The [visual ledger](visual-review.json) binds 30 individually inspected primary
images. For every corrected slide it includes standalone and embedded desktop
and 390px mobile states, for 24 browser captures total. Each state preserves the
four exact headings and descriptions, visible separators, title-to-content
space and nonoverlapping responsive items. Desktop keeps one four-item row;
mobile reflows the same content into four full-width rows with at least 16px
body and heading type and no horizontal overflow.

The six affected pages from the newly exported canonical PDFs were rastered at
1440×810 and inspected directly: player-promise pages 6–7 and surfaces pages 2,
4, 7 and 10. All retain their title, subtitle, four independent items,
takeaway, footer and page number without ordinal markers, clipping or overlap.
The four contact sheets are reduced navigation aids; acceptance observations
and hashes bind the individual images.

The focused browser gate also checks all three numbered progressions in the
same four standalone/embedded desktop/mobile modes, for 12 progression states.
It injects a marker into an independent comparison and removes one from a real
progression; both defects are rejected and the restored production DOM is
rechecked. Axe accessibility, artifact immutability and actual geometry are
part of the same run.

## Verification

The full standalone suites cover all nine player-promise and twelve surfaces
slides at desktop, compact and mobile sizes, including accessibility, source
dialogs, actual narration controls and complete PDF sequences. The focused
center run selects the two changed guides plus the 25 guides required by the
older canonical baseline: 27 selected guides, 33 exactly baseline-bound guides,
462 embedded slide captures and 27 PDF sequences pass in the current 60-guide
library.

The complete presentation Node and Python suites, builder freshness,
orientation source review, teaching-coverage integrity, narration integrity,
formatting, lint and receipt checks also pass. The library remains 578 unique
slides from 347 cited sources with 578 narration tracks totaling 14.36 hours.

## Reproduce

Run the focused browser and complete affected-guide checks serially with the
prepared presentation dependencies:

```sh
OSHUN_V2_ORDINAL_QA_DIR=/tmp/v2-ordinals CHROME_CHANNEL=chromium \
  python3 tools/presentations/run-supervised.py \
  node tools/presentations/tests/v2-comparison-ordinals.mjs
OSHUN_DECK_QA_DIR=/tmp/v2-promise CHROME_CHANNEL=chromium \
  python3 tools/presentations/run-supervised.py \
  node tools/presentations/tests/eve-oshun-deck.mjs --guide v2-product-promise
OSHUN_DECK_QA_DIR=/tmp/v2-surfaces CHROME_CHANNEL=chromium \
  python3 tools/presentations/run-supervised.py \
  node tools/presentations/tests/eve-oshun-deck.mjs --guide v2-product-surfaces
```

Raster the six affected PDF pages with `pdftoppm`, using `-png`,
`-scale-to-x 1440` and `-scale-to-y -1`. `verify-receipt.py` validates current
artifact and retained evidence hashes, report structure, visual-review scope
and the live focused browser behavior.

```sh
python3 docs/presentations/presentation-center/verification/v2-comparison-ordinals-2026-09-08/verify-receipt.py
```

The implementation baseline is
`3440a8d76dfa171cdd6ac2d83ae247c4fa7fc1bd`; the presentation sources remain
pinned to `1ceae52c7896a59ac74a4429d410f3218f0b91c6`. This receipt establishes the
semantic ordinal correction and the affected rendering states. It does not
claim willing-reader comprehension, normal-speed listening, live platform
acceptance or resolution of the separate ownership-graph typography finding.
