# Project Tolera — Final Package Consistency Review

**Reviewer:** Claude Science · **Date:** 2026-07-11 · **Scope:** 3 manuscripts, metrics ledger, website, deck + talking points, Gladstone prize text.
**Single source of truth:** `accomplishments_ledger.md` / `.csv`.

## Verdict: consistent and ready, with two delivery caveats (below).

## 1. Canonical facts — verified across all live deliverables
Cross-checked programmatically (word- and digit-forms). ✓ = present & correct; no stale forms found.

| Canonical fact | Website | Deck | Prize text | Ledger |
|---|---|---|---|---|
| 3 manuscripts (was 2) | ✓ "Three" | ✓ | ✓ | ✓ |
| 10 repos (was 8) | ✓ "Ten" | ✓ | ✓ | ✓ |
| 2 diseases (EoE + esophageal cancer) | ✓ | ✓ | ✓ | ✓ |
| pMHC engine 4,640 → 832 | ✓ | (in talking-points script) | ✓ | ✓ |
| No "Claude AI" co-author | ✓ | ✓ | ✓ | ✓ |
| No CZ Biohub / IP language | ✓ | ✓ | ✓ | ✓ |
| Canonical URL = ruthannepai.netlify.app | ✓ | ✓ | ✓ | ✓ |

**No stale strings** anywhere: "2 manuscripts", "8/eight repos", "9 constructs", "three formats" all return zero hits. Only `ruthannepai.com` occurrences are `mailto:` email hrefs.

## 2. EAC/Barrett's third stream — now present everywhere (was absent)
- **Manuscripts:** EAC is a co-equal third manuscript (harmonized front/back matter).
- **Website:** full Manuscript 3 section + home Stream 3 card + about mention. Per-page EAC-term counts (GUCY2C|DKK1|esophageal (adeno)carcinoma|EAC|Barrett|adenocarcinoma), recomputed against the final built HTML: home 8, manuscripts 39, about 2; software/biotech/specialist 0 (expected — no disease-specific manuscript content there).
- **Deck:** woven into slide 10 (reusable engine) as the second-disease proof.
- **Prize text:** in both the description and the "how I used Claude" answer, substantiating the title's "for Any Disease".

## 3. Harmonized manuscripts — front matter (12/12 checks passed in Step 3)
Sole-author byline (Ruth-Anne Pai, PhD), Citizen Scientist + Pai Advisory LLC affiliations, advising@ruthannepai.com contact, "not peer reviewed" + in-silico + not-medical-advice disclosure banner, Funding / Competing-interests / Data-and-code / AI-use / Acknowledgments end-matter (Anthropic · Gladstone Institutes · Cerebral Valley named). Hill/Spergel FTO note retained on pMHC only, per ruling.

## 4. Person-first & accessibility
- No "sufferer / afflicted / victim" in any deliverable.
- Website passes WCAG AAA contrast (23/23 body-text pairs ≥ 7:1); lang, skip-link, main-landmark intact; AI-disclosure footer site-wide.

## 5. Honesty framing — consistent
The decoy-control result (scrambled sequences clear the same ipTM/pLDDT gate as real designs) is carried in the CCL26 manuscript, the website ("The honest result" callout), and now the prize text. Structure prediction is called computational/in-silico, never "validation", throughout.

---

## Outstanding items (need your hand / a decision)

1. **CCL26 harmonized manuscript — not yet in Drive.** The connector's `create_file` rejects large payloads ("Internal error"); it failed for the ~2 MB CCL26 docx and the 5.7 MB deck. Both are saved as **artifacts** and are correct. **Action:** download the artifact and drag it into the CCL26/POSTN Drive folder manually. (pMHC + EAC harmonized manuscripts and the corrected CCL26 lay summary *did* upload successfully.)

2. **Revised deck — not yet in Drive** (same payload cap). Saved as artifact `Project_Tolera_Deck__revised.pptx`. **Action:** drag into the Drive folder root manually.

3. **Misfiled lay summary.** `lay_summary_pMHC_nanoparticle.docx` still sits in the CCL26/POSTN Drive folder (wrong content for that folder; also reflects the older nanoparticle framing). The correct CCL26/POSTN lay summary was uploaded. **Action (your OK needed):** delete the stray file from the CCL26/POSTN folder.

4. **pMHC "nanoparticle" figures.** The website manuscripts page no longer shows the old nanoparticle schematics (removed this round). They remain in the manuscript `.docx`/supplement bodies (unchanged) — no action required unless you want them pulled from the supplement too.

## Website revision round (all resolved)
- **Manuscript-page lead figure** now correctly swapped to the engine-consistent figures in the delivered build (a prior report overstated this — the swap had been blocked by an f-string build error that shipped stale HTML; fixed and verified against the zip).
- **Whitespace bug** under "Nanoparticle architecture" / "EAC is copy-number-driven" fixed (`.fig-2` grid `align-items:start`; pMHC pair now single-column).
- **Light/dark theme** added — dark default, nav toggle, `localStorage` persistence; **both themes verified WCAG AAA** on all text/surface pairs.
- **All DOCX downloads replaced with PDFs** (3 harmonized manuscripts + supplemental, converted via mammoth→weasyprint); Manuscripts 2 & 3 gained a manuscript download they previously lacked; stale pre-harmonization docx removed.
- **Biotech page** past-trials critique removed; **About page** thesis updated to the author's supplied text.

## Deliverable index (artifacts)
- Ledger: `accomplishments_ledger.md`, `accomplishments_ledger.csv`
- Manuscripts (harmonized): pMHC, EAC, CCL26 (+ corrected CCL26 lay summary); `manuscript_consistency_report.md`
- Website: `project_tolera_site.zip`, `pages.py`, `style.css`, `asset_manifest.csv`; `website_rebuild_report.md`
- Deck: `Project_Tolera_Deck__revised.pptx`; `deck_talking_points.md`
- Prize: `prize_submission_revised.md` / `.txt` (also uploaded to Drive)
- Decisions of record: `consistency_decisions.md`