Live site: https://emberian.github.io/comp-theo/ — poster browser (all 210 panels, both transcription passes side-by-side), critical essays, guided lessons, philosophy comparisons, glossary, and poster map.
A mirror and reconstruction of the Computational Theology visual-essay series (5 imgur albums, 210 posters), plus the dialogue transcript that produced it.
The source albums are private imgur galleries and prone to link rot. This repo preserves the images, machine-transcribes every panel, and reconstructs the underlying intellectual framework from the primary material.
images/ptN/NNN_<id>.png mirrored posters, album order preserved
manifest.json ordered index: part, index, imgur id, url, dims
parsed/ptN/NNN_<id>.md per-image transcription + visual description
scripts/fetch_albums.py re-runnable mirror (parses embedded imgur JSON)
scripts/parse_images.sh re-runnable codex/ChatGPT vision pass (resumable)
synthesis/MAP.md one line per poster (210), album order
synthesis/GLOSSARY.md core vocabulary, grouped, with provenance
synthesis/provenance.json which engine parsed each image (codex vs claude)
transcript/conversation.full.md complete originating ChatGPT dialogue
(prompts, replies, hidden reasoning,
tool calls — full fidelity)
logs/ run + error logs
| Part | Slug | Count |
|---|---|---|
| pt1 | computational-theology-xCd6wA0 | 50 |
| pt2 | computational-theology-pt2-eST1EhE | 50 |
| pt3 | computational-theology-pt3-5FEtmF0 | 50 |
| pt4 | computational-theology-pt4-joslo98 | 50 |
| pt5 | computational-theology-pt5-le7NmNz | 10 |
Author account: UrPissMissedThePot (imgur id 155100497). Albums created
2026-05-15, platform iOS, privacy: private. Captured 2026-05-18.
Transcriptions were produced by two engines: 73 via codex exec
(ChatGPT GPT-5.x vision) before the ChatGPT plan usage limit was hit, and
the remaining 137 via Claude (Opus 4.7) vision when the quota blocked
codex for hours. Per-file mapping in synthesis/provenance.json. Format and
prompt were identical across both.
Plain HTML/JS (no build framework) under site/, deployed to GitHub Pages
by .github/workflows/pages.yml (assembles site/ + mirrored images/,
regenerating site/data/ via scripts/build_site_data.py). Local preview:
scripts/serve_site.sh → http://localhost:8765. Interpretive essays/lessons/
comparisons are a skeptical outside reading, not the original author's voice.
python3 scripts/fetch_albums.py # mirror images + manifest
scripts/parse_images.sh 5 # parse via `codex exec` vision
# (needs `codex` logged in; resumable)
CODEX_MODEL=gpt-5.5 scripts/parse_images.sh 5 # to pin a model