1 Enoch (Mashafa Henok) — scope and source strategy

This document covers the People's Open Bible's dedicated sub-phase for the Book of Enoch (1 Enoch / Ethiopic Enoch / Mashafa Henok / መጽሐፈ፡ ሄኖክ). It sits under the Pseudepigrapha track alongside 2ESDRAS.md, complementing DEUTEROCANONICAL.md (the 13 LXX-based books) and applies the three-zone scholarly-source policy from REFERENCE_SOURCES.md.

Status (2026-10-02): translated corpus is incomplete. The read-only primary-source inventory finds 53 absent verse files in chapters 1–35, and 209 recovered-primary missing-unit leads across all 108 chapters. Counts beyond chapter 35 are parser-derived leads, not a certified versification inventory. Existing clipped source/translation records also need repair. Do not infer completeness from the presence of 108 chapter entries or from generated Graphic Reader panels.

Run python3 tools/enoch/audit_pob_source_coverage.py --chapters 1-108. See sources/enoch/coverage/pob-primary-inventory-20261002.json for exact missing verse numbers and source-review leads. These checks do not approve translations; source-complete candidates belong outside translation/ until their bounded source/review/application gates pass.

Current source-review route (2026-10-07)

Vertex is not required. tools/enoch/review_source_candidates.py runs a fresh, read-only, ephemeral scan-review pass through the existing ChatGPT-signed-in Codex CLI, using its configured model. It does not read or extract authentication tokens and never installs or approves translations. The bounded transaction remains the separate application gate. Review inputs, actual source images, output and provenance hashes are preserved.

Chapters 7–8 were restored by the October 6 bounded application. Chapter 13's October 7 package restores clipped 13:4 and absent 13:5–10; it explicitly discloses the selected critical edition's uncertain/editorially restored readings. Neither that work nor the older inventory constitutes complete-book or human-specialist certification. Recompute the audit for current counts.

Chapter 83 bounded repair (2026-10-14)

The authorized read-only Codex scan-review workflow accepted all eleven units of chapter83 against Charles1906printed159-161, including the4/11cross-page continuations. The bounded transaction restores clipped83:11, repairs source glyphs and note anchors, removes the unsupported83:7 reveals assertion, uses the printed future disclosure at83:10, and retains provisional83:6 Malalyal rather than normalizing the different name forms. Corrupt83:11 readings and the final passive-clause interpretation remain disclosed. Original OCR and old model claims are preserved as unverified history. See sources/textual_restoration/applications/enoch_83_20261013/ and run tools/textual_restoration/enoch_83_transaction.py verify-applied.

Chapter83 now has an explicit scan-bound unit inventory. Chapters45 and97 have separately scan-checked count boundaries only; their missing final units remain unapproved. Other counts beyond35 remain parser leads unless separately certified. Chapter83's work is selected-edition editorial approval, not human specialist, independent manuscript, full-book or Graphic Reader publication certification. The chapter45/97 count checks are not editorial approval at all. Artwork continuity/publication QA is still separate.

Chapter 97 individual source repairs (2026-10-20)

Two scan-bound whole-chapter reviews held 97:10 and called for revisions in 97:1-9. A separate bounded review accepted revised individual units 97:1, 3, 5 and 8. Their hash-checked transaction installed only those four source, translation and note repairs; canonical 97:10 remains absent and the mobile exporter still excludes chapter 97. Neither the reviews nor this source repair approve chapter-level Graphic publication. See sources/textual_restoration/applications/enoch_97_review_20261019/ and sources/textual_restoration/applications/enoch_97_1_3_5_8_20261019/.

Bounded repairs and complete issue register (2026-10-20)

The October 20 audit records fifteen bounded repairs, including three newly restored records (18:11,14,15), restored 18:10's heavens subject, corrected 63:7 negatives/notes and mechanical English deduplication at 81:4 and 89:26. The independent review rejected unproven 34:1 wording, and the whole chapter84 hold remains intact. No artwork was published.

The fresh all-108 inventory reports 15 absent files under declared counts, 153 additional parser-derived absent-unit leads and 50 source-alignment signals. These are not a sum of proven translation errors or full-book certification. Run python3 tools/enoch/audit_editorial_issues.py --output /tmp/enoch-issues.json for a hash-bound issue register; inspect the linked audit for exact references, retained source uncertainties and the application/verification commands.

Why Enoch

Four reasons this book warrants a dedicated phase:

1. Canonical in the Ethiopian Orthodox Tewahedo Church — part of their 81-book canon. Eritrean Orthodox and Beta Israel also receive it. There is no permissively-licensed modern English Enoch translation serving these communities. R. H. Charles 1912 is PD but Victorian in register; Charlesworth OTP (1983), Nickelsburg Hermeneia (2001/2012), Black 1985, and Knibb 1978 are all copyrighted.
2. Quoted in the New Testament. Jude 14-15 explicitly cites 1 Enoch 1:9, and 2 Peter 2:4 alludes to the Watchers tradition developed in Enoch. Readers of Jude and 2 Peter benefit from direct access to the cited source.
3. Early Christian reception is deep. Tertullian, Irenaeus, Clement of Alexandria, and Origen all cite 1 Enoch. The book was lost to Western Christianity by the 5th century and rediscovered by James Bruce in 1773.
4. The textual situation is rich enough to do well. We have PD Ge'ez critical editions (Charles 1906, Dillmann 1851), PD Greek fragments (Codex Panopolitanus for chs 1-32, Syncellus extracts, Chester Beatty for chs 97-107), AND a validated digital Ge'ez text we can use as a cross-check oracle. This is actually a stronger source position than 2 Esdras.

Compositional structure

1 Enoch is a composite work of five sections, compiled by the 1st century AD:

The "Book of Parables" (chs 37-71), a.k.a. Similitudes, is distinctive — it contains the "Son of Man" material that influenced NT christology, and it survives only in Ge'ez. This is not textually problematic (modern scholarship converged on early 1st c. AD Jewish composition), but it means we have only one language witness there versus up to three elsewhere.

Canonicity

Per POB policy (see DEUTEROCANONICAL.md): we take no position on canonicity. We label the book clearly in its canonical context when published.

Source editions — vendored

Source-control clarification, 2026-10-05: critical editions are edited texts, not independent manuscript witnesses. The Enoch 1:9 and Jude comparison checks actual survival, editorial supplies and edition dependencies. Charles 1906 explicitly uses Greek and Jude in its apparatus at this unit; its agreement with Jude cannot automatically count as independent Ge'ez attestation. Local Greek OCR also clips the end of this complete printed verse. The older pipeline plans below are not evidence that these controls have passed for every verse.

Ethiopic (primary)

Greek (secondary witnesses)

The coverage descriptions above follow Flemming–Radermacher's actual introduction, printed pages 13–14. Do not route chapters 97–107 to that 1901 volume as Chester Beatty evidence. A verified edition of the later Chester Beatty–Michigan material is a separate acquisition and comparison task; its chapter labels do not establish every surviving word.

English reference (not reproduced; verse numbering + cross-check only)

See sources/enoch/MANIFEST.md for SHA-256 hashes and rehydration commands.

Zone 1 validation oracle (not vendored)

• Beta maṣāḥǝft digital Ge'ez Enoch (LIT1340EnochE.xml, 712 KB TEI XML, Hiob Ludolf Centre, University of Hamburg). Based on Michal Jerabek's 1995 Library of Ethiopian Texts digitization. 216 chapter divisions, ~179,000 Ethiopic characters, complete coverage of 1 Enoch 1–108.
  • License: outer wrapper declared CC-BY-SA 4.0; inner text attribution says "noncommercial purpose" and "cannot be sold." These are in tension; conservatively treated as non-redistributable.
  • Role in POB: validation oracle only. We do NOT vendor this XML; we do NOT derive our published Ge'ez from it. Instead, we cross-check our own fresh OCR of Charles 1906 and Dillmann 1851 against this digital text. Where they agree, our OCR is validated; where they disagree, we investigate the specific character and publish our reading with explicit provenance.
  • This pattern is identical to how First1KGreek TEI (CC-BY-SA) served as validation oracle for our LXX Swete OCR.

Zone 2 consult (scholarly, not reproduced)

Per REFERENCE_SOURCES.md:

• Milik, The Books of Enoch: Aramaic Fragments of Qumrân Cave 4 (Oxford, 1976) — Qumran Aramaic fragment reconstructions; coverage must be established passage by passage
• Nickelsburg, 1 Enoch 1: A Commentary (Hermeneia, 2001) — chs 1–36, 81–108
• Nickelsburg & VanderKam, 1 Enoch 2 (Hermeneia, 2012) — chs 37–71
• Knibb, The Ethiopic Book of Enoch (Oxford, 1978, 2 vols) — modern critical Ethiopic edition
• Black, The Book of Enoch or 1 Enoch: A New English Edition (Brill, 1985)
• Charlesworth (ed.), The Old Testament Pseudepigrapha vol. 1* (1983) — Isaac's translation

These are consulted during translation and cited fact-level in footnotes where their conclusions inform a reading. Their text is never reproduced in POB output.

Translation strategy

Same three-zone discipline as the LXX deuterocanon, adapted for Enoch's situation:

1. Primary source: our own fresh OCR of Charles 1906 Ge'ez, cross-validated against Beta maṣāḥǝft digital text (oracle) and Dillmann 1851 as a second edition control, not another manuscript vote. Preserve Charles's apparatus and supplied/emended wording separately from manuscript attestations. OCR agreement checks transcription of editions, not automatically their earliest-source priority. This OCR, once validated, becomes the Zone 1 Ge'ez Vorlage for English translation.
2. Secondary textual apparatus for chapters 1–32: Bouriant 1892 Greek (Panopolitanus) provides an additional witness to the lost Aramaic. For chapters 97–107, a separately verified Chester Beatty–Michigan edition is required; Flemming 1901 cannot supply that later acquisition.
3. Parables (chs 37–71): Ethiopic-only at Zone 1. No Greek, no Qumran. Our output disclose this in the per-verse YAML.
4. Qumran Aramaic parallels where they exist: Zone 2 consult via Milik 1976. We footnote specific textual agreements/disagreements between Qumran and Ethiopic without reproducing Milik's reconstructions.
5. Per-verse YAML follows the standard POB schema with witnesses added: zone1_sources_at_draft, zone2_consults_known, plus an enoch_witnesses block listing which languages attest that verse.

OCR validation (2026-04-21)

Critical finding during setup:

• GPT-5 (Azure deployment we've used for Greek and Hebrew) fails on Ge'ez. Multiple attempts returned "cannot read" errors even at 4000px resolution on clean Dillmann 1851 pages.
• Gemini 2.5 Flash fails. It can emit real Ethiopic Unicode, but it hallucinates the content and cannot be trusted for scholarly OCR.
• Gemini 2.5 Pro in plaintext mode succeeds. Validated against the Beta maṣāḥǝft chapter-1 oracle on Dillmann 1851: strong line-level agreement, real word separators (፡), sentence terminators (።), and Ge'ez numerals preserved, with the remaining disagreements largely traceable to witness variation rather than obvious OCR garbage.

Operational implication: our Enoch OCR pipeline uses Gemini 2.5 Pro in plaintext mode rather than Azure GPT-5 or Gemini Flash. Azure GPT-5 remains primary for Greek + Latin books. For Ge'ez, Pro is the model that actually passes validation.

Pipeline components (planned)

Phased timeline

Total ≈ 5-7 weeks of focused effort, scheduled after Phase 10 completes.

Unique strengths of this translation

When complete, our Enoch will be:

1. First CC-BY 4.0 modern English Enoch. All existing modern translations (Nickelsburg, Black, Isaac in OTP, Knibb) are copyrighted. Charles 1912 is PD but 112 years old and Victorian-register. Our output fills a real gap.
2. Ethiopian Orthodox / Eritrean Orthodox readers gain a permissive modern English. Existing English Enochs are produced for Protestant or Catholic audiences; an open translation serves Ethiopian canon readers directly.
3. Transparent textual provenance per verse. Which Ge'ez manuscript reading, which Greek fragment (if any), which Qumran Aramaic fragment was consulted (Zone 2) — every verse documents it.
4. Multi-witness Ge'ez grounding. Cross-validated OCR against the best scholarly digital text (Beta maṣāḥǝft) means our Ge'ez Vorlage is traceable, verifiable, and competitive with modern critical editions for textual accuracy.

Future related works (under consideration)

Enoch sits in the Ethiopian broader canon alongside works that would be natural extensions of this phase if we ever pursue them:

• Jubilees — Ethiopic Book, also at Qumran in Hebrew fragments
• Meqabyan 1–3 — Ethiopian Orthodox canonical, Ge'ez-only
• 4 Baruch / Paraleipomena Jeremiou — Greek + Ethiopic witnesses

None are committed to. The Enoch phase itself is scheduled; these remain speculative and would each need their own scope document when/if we pursue them.