Procedure for MMv3 Exhibit Page Generation#
Documented by Claude Opus 4.6 (dv_ClaOp46Max_MMv2_2026m04d02) for continuation across sessions.
Session 1 Progress (2026-04-02)#
Pages created: 23 exhibit pages + 2 full catalogs (all 253 PDFs) + structural files
Completed folders:
open-letter/ — all 11 letters (OL0–OL10), each with individual exhibit page + folder index
supporting-doc/ — all 10 documents (SD1–SD10), each with individual exhibit page + folder index. SD1 was rewritten from an older format to match the current template.
about/ — 1 folder-level catalog page covering 6 PDFs
dont-panic-guide/ — 1 exhibit page covering 1 PDF
mmv3/index.rst — master gateway page linking all folder indices
Structural files created:
mmv3/index.rst— master gateway (VV toctree points here only)mmv3/open-letter/index.rst— folder index for OL0–OL10mmv3/supporting-doc/index.rst— folder index for SD1–SD10_templates/include-file/footer/cta/gnp-mmv3-review.rst— generic review CTAmmv3/catalog-text.rst— Level 1 plain text catalog (all 253 PDFs)mmv3/catalog-visual.rst— Level 2 visual catalog (all 253 PDFs with thumbnails)mmv3/collection-overview.rst— detailed section breakdown with exhibit page statusmmv2/index.rst— moved from VV root into its own subfolder (consistent with mmv3)mmv3/deprecated/catalog-test.rst— old template test page (replaced by proper catalogs)
Catalog generation script: /tmp/gen-catalogs.py (Python, generates both catalogs from filesystem scan).
To regenerate after adding new PDFs: python3 /tmp/gen-catalogs.py from the repo root.
VV-level toctree: hidden, points only to mmv2/index and mmv3/index.
MMv2 is linked from the VV table but not visible in the navbar.
MMv3 master index has 3 toctree groups: Core Documents, Catalogs & Overviews, Session Logs (hidden).
Catalog ordering: Don’t Panic Guide first, then Supporting Documents, Open Letters, About, then Extra Good News, Flying Scroll, Dusty Deep Data, and remaining folders. Each catalog has a prominent tip box pointing to the Don’t Panic Guide and SD1-SD3 as entry points.
Remaining Work (~226 PDFs)#
Folder |
Est. PDFs |
Notes |
|---|---|---|
|
5 |
Academic CV, tenure dossier, grants, teaching — done (5 exhibit + catalog, 2026m04d16) |
|
2 |
Personal spiritual exhibits — done (2 exhibit + catalog, 2026m04d16) |
|
19 |
Published scientific papers (6–444 pages each!) — done (19 exhibit + 2 indexes, 2026m04d14) |
|
6 |
7 tribes model, Waldo game |
|
1 |
AIPTO canary rules (20 pages) |
|
2 |
Sensitive content (sex positivity, Paul study) |
|
4 |
Ketubah (360pg!), Book of Life, anti-Christ, Craftiest Criminal (150pg) |
|
4 |
Physical banners (1x6ft, 3x6ft) |
|
5 |
ZoE flags (4-state epiocracy) — done (catalog, 2026m04d14) |
|
24 |
Iron Rod views, StayVS diagrams — done (6 exhibit + catalog, 2026m04d14) |
|
22 |
Posters A–Z2 (many wall-sized 3x4ft!) |
|
8 |
Stickers (beast number, David stars, etc.) |
|
1 |
Draft talk (69 slides) |
|
~55 |
12 stadia (sta1–stb12) + overview + evolb samples |
|
2 |
Wall-sized Flying Scroll combos (5x10m, 12–20MB!) |
|
10 |
Y3T intro series — done (10 exhibit + catalog, 2026m04d16) |
|
1 |
16-page sample |
|
2 |
Combined packs (108pg!) |
|
4 |
Shared attachments |
|
8 |
Ketubah signature pages — done (catalog, 2026m04d14) |
|
1 |
SD1–SD10 combined (14pg) |
|
8 |
SD1 figures (RiskyMAD diagrams) — done (catalog, 2026m04d14) |
|
2 |
SD3a contract overview, SD3b signature — done (2 exhibit, 2026m04d16) |
|
1 |
SD5b modeling quality 15 guides — done (1 exhibit, 2026m04d16) |
|
3 |
4 horsemen, datageddon prison, datageddon escape — done (3 exhibit, 2026m04d16) |
|
2 |
SD8a ark hardware, SD8b ark software — done (2 exhibit, 2026m04d16) |
|
1 |
Hardship explanation — done (1 exhibit, 2026m04d16) |
Root level |
2 |
AHA readme, Jonah License |
Per-File Procedure#
For each PDF, Claude follows these steps:
Create directory:
mkdir -p source/good-news-pack/vv/mmv3/[path]/Get file sizes:
du -h [pdf] [webp]Read PDF:
Read tool, pages 1-N(max 20 pages per request; chunk if longer)Write exhibit page at
source/good-news-pack/vv/mmv3/[path]/index.rstcontaining:.. meta::with description, keywords,og:card:title,og:card:description.. title::with short navbar title (e.g., “SD1 — Nuclear Winter”, not full title)Full RST title (overline MUST be at least as long as the title text, or Sphinx warns)
One-line hook in italics
Download admonition (
:class: note, blue) with:PDF link + size in parentheses
Page count + license:
:doc:`Jonah License with CC0 Public Domain </license/joli/index>`Literal filename with
.pdfextensionWebP preview link + size
“Also in this folder:” links if the source folder has additional PDFs
Full-width WebP preview image (from the matched
.webpin the same source folder)--- Overview AI-generated by dv_ClaOp46Max_MMv2_2026m04d02 ---Start---Abstract
Broader Significance (Claude’s Assessment)
Who This Document Is For (audience table, for major docs)
Key Concepts at a Glance (term-definition table)
Document Information (metadata table including PDF/WebP sizes)
Related documents (links to PDFs; to be updated to
:doc:links once exhibit pages exist)--- Overview AI-generated by dv_ClaOp46Max_MMv2_2026m04d02 ---End---Footer form block:
.. |content-vvn| replace:: dv_ClaOp46Max_MMv2_2026m04d02.. include:: /_templates/include-file/footer/separator/tidal-gradient.rstNotes dropdown with
page-footer-stability.rstinclude.. only:: htmlblock containing:.. include:: /_templates/include-file/footer/cta/gnp-mmv3-review.rstFF FeedbackFlow dropdown with prelink, page-specific mailto, postlink
Add to folder-level toctree (NOT the VV-level toctree). Each folder has its own
index.rstwith a toctree. New folders must also be added tommv3/index.rst.Periodic build check:
make htmlto catch errors (especially title overline warnings)
Folder-Level Catalogs#
For simple visual assets (logos, stickers, flags, figures):
skip individual PDF reading, create one index.rst per folder with visual
catalog entries. Each entry is a list-table row with:
WebP thumbnail (clickable, linking to PDF)
Bold title (AI-derived or from filename)
VVN as-is (do NOT parse into separate fields)
Page count + license link
PDF link + size, WebP link + size
Literal filename with extension
Warning for large files (>10 MB) or wall-sized documents
Key Conventions#
License text: Always lead with “Jonah License” and always link:
:doc:`Jonah License with CC0 Public Domain </license/joli/index>`
Use :doc: reference (NOT raw <path> hyperlink).
FF slug pattern: Use the full mmv3 path including filename (drop folders
above mmv3/). Example:
FF+mmv3-supporting-doc-sd1-nuclear-winter-iv_llol_qqv4_2025m12d03@balospe.com
Truncate if over 64 characters (email local-part limit).
Navbar nesting: The VV-level toctree points ONLY to mmv3/index.rst.
mmv3/index.rst points to folder-level indices (open-letter/index.rst,
supporting-doc/index.rst, etc.). Folder indices point to individual pages.
Every new folder needs its own index.rst with a toctree, and must be
registered in mmv3/index.rst.
Short navbar titles: Use abbreviations (e.g., “OL1 — to US Presidents & All”, “SD6 — BABL Millstone”, “About — Logos & Cards”). Do NOT use full document titles in toctrees.
Title overlines: RST overlines (****...) MUST be at least as long as the
title text. Check this before writing — Sphinx emits a warning if too short.
Download box color: :class: note (blue). NOT :class: important (red).
Version code: dv_ClaOp46Max_MMv2_2026m04d02 for all pages in this batch.
Increment on significant rewrites (e.g., MMv3_2026m04d03).
Related document links: Currently link to PDF files directly. Once exhibit
pages exist for those documents, update to :doc: references. This is a
deferred task for a future session.
Context Budget Analysis#
Problem: Processing 253 PDFs with full AI-generated abstracts consumes ~2.5% of the 1M context window per document (reading PDF + writing RST). At this rate, one session can process ~25–35 documents before context pressure becomes critical.
Current session: 22 pages written, ~38% context remaining, ~226 PDFs to go. That requires 7–9 additional sessions at the current quality level.
Proposed Sustainable Strategy#
Option A: Multi-session sequential processing (RECOMMENDED)
Split remaining work into ~8 sessions, each processing one folder group:
dusty-deep-data/(26 PDFs, many are long academic papers)extra-good-news/(13 PDFs, includes 360-page Ketubah)flyingscroll/banner+flag+sticker+talk/(18 PDFs, simple visuals)flyingscroll/ironrod/(21 PDFs, diagrams)flyingscroll/page/(22 PDFs, wall-sized posters)flyingscroll/transwarpkey/ part 1(~28 PDFs, stadia 1–6)flyingscroll/transwarpkey/ part 2(~27 PDFs, stadia 7–12 + evolb)flyingscroll/yahs-3rd-temple/ + remaining(~20 PDFs)
Each session: read this procedure file, pick up where the previous left, process the assigned folder group, update this inventory.
Option B: Tiered processing
Session 2: Generate Level 1 plain-text catalog for ALL 253 PDFs (no PDF reading, filenames only, zero garbling risk). This makes everything immediately accessible. Then do Level 4 exhibit pages in subsequent sessions for priority documents only.
Option C: Hybrid (BEST)
Session 2: Generate Level 1 text catalog + Level 2 visual catalog for ALL remaining PDFs (from filenames only, ~1 session). Then subsequent sessions add Level 4 exhibit pages for selected high-priority documents (e.g., Ketubah, complex posters, Y3T Rosetta, key transwarpkey stadia).
Key efficiency gains for future sessions:
For simple visual assets (flags, stickers, banners), use folder-level catalogs instead of individual exhibit pages
For academic papers in
dusty-deep-data/, brief pages with bibliographic info extracted from filenames (no full PDF reading)Reserve full AI abstracts for the most important/complex documents
Batch toctree updates (add all pages at end of session, not one by one)