The Deuterocanonical Books in the People's Open Bible

This document explains why the People's Open Bible includes the Deuterocanonical books (also called the Apocrypha, or Anagignoskomena), how we intend to translate them, and what commitments we make about their integrity and presentation. It complements DOCTRINE.md (theological stance), METHODOLOGY.md (pipeline), and REVISION_METHODOLOGY.md (revision process).

Status: working translations and ongoing source comparison. The former strategy-only status is obsolete. Existing drafts and bounded applications do not establish publication approval or a completed critical text for each book.

See also:

• APOCRYPHA_PROVENANCE.md — public-facing per-sub-pipeline account of how each of the 17 books was sourced and verified, with reader-auditable per-verse provenance. Written for anyone asking "where does this text actually come from?"
• REFERENCE_SOURCES.md — the operational policy for using copyrighted scholarly sources during translation (three-zone model: vendored / consulted / forbidden).
• EXTRA_CANONICAL.md — the scope document for texts beyond the 13 LXX deuterocanonical books (2 Esdras, 1 Enoch, Didache, etc.) that the People's Open Bible is committed to translating under CC-BY in phases after Phase 9.
• PRAYER_OF_MANASSEH.md — per-book account for MAN (not in Codex Vaticanus, so sourced from Charles 1913).
• 2ESDRAS.md — 2 Esdras (4 Ezra) sub-phase strategy.

This document is the "what and why" for the 13 LXX deuterocanonical books specifically. REFERENCE_SOURCES.md is the "how, per verse, mechanically" for all our translation work.

Position assessment — where we stand per book

Corrected 2026-10-04 to agree with the earlier reassessment in REFERENCE_SOURCES.md. The former April assessment's “optimal,” “definitive” and “sole surviving witness” claims were not demonstrated. Its prediction that a specialist would notice only 5–10 Tobit differences was untested. Available working inputs do not certify the best attainable source or a uniquely optimal translation.

The Sirach witness guide records survival and access limits. The IOSCS edition guide distinguishes Swete's diplomatic edition from wider critical work. Consultation does not recover absent material or settle redistribution rights. No book list, canonical status, verse wording or public release is changed by this correction.

Why include these books

The People's Open Bible exists to make the biblical textual tradition transparent and auditable. Excluding these books would itself be a theological editorial decision — one that curates which texts readers are permitted to evaluate. POB's mission is the opposite: present the tradition; let the reader and their faith community judge.

Concretely:

• Historical precedent. Luther's 1534 German Bible included the Apocrypha in a clearly labeled section between the Testaments, calling them "books not regarded as equal to the Holy Scriptures, yet useful and good to read." The 1611 King James Version likewise included them as a middle section. Their removal from most Protestant Bibles is a 19th-century development driven primarily by printing economics at the British and Foreign Bible Society, not theological consensus.
• Ecumenical reach. These traditions do not receive POB's entire collection identically. The Catholic canon list includes Tobit and Sirach, but not every work in the scope table below. The Ethiopian Orthodox list names a different collection, including Enoch and Jubilees. Identify reception by named community and exact work; do not generalize one list to every Eastern or Oriental Orthodox church. Clearly labeled inclusion serves readers across traditions without declaring all these works equally canonical.
• Intertestamental witness. These books describe the world Jesus was born into — the Hasmonean revolt (1-2 Maccabees), Second Temple wisdom traditions (Sirach, Wisdom of Solomon), and Jewish devotional life in the diaspora (Tobit, Judith). Reading the NT without them means reading a text whose cultural assumptions are partially invisible.
• Consistency with POB's stated philosophy. DOCTRINE.md commits to "preserve theological tension rather than resolve it" and "original-language primacy over translation tradition." Including texts that are canonical in some traditions and not in others preserves the tension; excluding them resolves it in one direction.

We do not pronounce on canonicity. We translate and publish.

Our theological stance on canonicity

POB takes no position on whether these books are canonical. This refusal is itself a stance, and we name it clearly:

• To Protestant readers: including these books does not assert they carry the same authority as the Protestant canon. Article 6 of the Anglican Articles of Religion names them as books the church "doth read for example of life and instruction of manners, but yet doth it not apply them to establish any doctrine." That is a reasonable stance for any reader to take while using this edition.
• To Catholic and Orthodox readers: including these books respects the canonical status your traditions have affirmed since antiquity. We translate them with the same rigor and per-verse provenance as the Protestant canon.
• To all readers: each book is labeled with its canonical status across the major traditions, so no reader encounters a book without knowing which Christian communities receive it and how.

This is the same posture DOCTRINE.md takes toward contested theological terms: name the controversy, present the alternatives, let the reader judge.

Terminology

Different Christian traditions use different names for this corpus. None of the terms is theologically neutral. We use them as follows:

In POB we use Deuterocanonical as our primary term in filenames and repository documentation because it is the most widely understood non-pejorative option in contemporary scholarship. In user-facing contexts (the mobile app, published editions) we will label the section as "Deuterocanonical Books (Apocrypha)" so readers from every tradition recognize what they are reading.

Scope — which books, from which sources

Books included

Our target scope is the complete corpus recognized by at least one major Christian tradition with a stable textual history:

Source editions and declared working forms

Declare the working form and edition for each book or passage, then compare the relevant witnesses before claiming historical priority. The former 12-of-14 “original or sole surviving witness” assertion is withdrawn. A Greek translation, Greek composition and Greek manuscript are different things; one available edition cannot certify all three or an optimal corpus. The adjudication method governs source selection, with parallel forms preserved where appropriate.

Field research (2026-04-18) confirmed that the commonly-referenced Rahlfs LXX 1935 has no CC-BY-compatible digital transcription we can vendor: every digital Rahlfs edition we evaluated is either CC-BY-NC-SA (eliranwong, CenterBLC), CC-BY-SA (Perseus, First1KGreek derivatives), or restrictive (CCAT). STEPBible's announced TAGOT has not shipped.

Instead, our LXX source is Henry Barclay Swete, The Old Testament in Greek According to the Septuagint, 3 vols. (Cambridge University Press, 1909–1930) — fully public domain (Swete died 1917) and hosted as scanned PDFs on Internet Archive. Swete's edition contains every deuterocanonical book we need. We transcribe the Greek ourselves from the archival page scans using AI vision, releasing the transcription under CC-BY 4.0. This is legally clean (PD source

• our own work) and produces a working source text whose provenance is auditable to specific public-domain page images.

For the two books where a non-Greek original survives (Sirach in Hebrew, Tobit in Aramaic), we reach for the original language wherever it is available under a clean license. The details differ by book:

• Sirach has recovered Hebrew for roughly two-thirds of the book. We vendor two Zone 1 Hebrew sources: the Sefaria/Kahana composite (CC0) of the Cairo Geniza manuscripts as a complete starting point, and Schechter & Taylor 1899 (PD) for MSS A–B directly from the first-publication facsimiles. Greek (Swete) covers the Hebrew gaps. Zone 2 consultation includes Beentjes 1997 and Skehan & Di Lella 1987 for cruxes.
• Tobit's ancient Aramaic witness (Qumran 4Q196–200, covering ~20% of the book) is held under IAA/DJD copyright and cannot be vendored. Our Zone 1 Hebrew reference is Neubauer 1878 (PD via Sefaria) — a 19th-century back-translation, not a Vorlage. Primary source text is the LXX Long Recension via Swete (Codex Sinaiticus, TOB_S); this tracks the same textual tradition as the Qumran fragments. Zone 2 consultation of Fitzmyer's DJD XIX reconstruction informs the ~20% Qumran-overlap verses at fact-level without reproducing his reconstructed Aramaic.

For 1 Esdras, which is a Greek composition reworking MT 2 Chr / Ezra / Neh, we vendor a verse-level alignment table mapping each 1ES verse range to its WLC Hebrew parallel (Zone 1), so the translator can lock proper-name spelling and idiom against our existing Hebrew corpus even though the Greek remains primary.

Per-book source path

Legend: ✓ = clean path, material vendored and Swete transcribed. ⚠ = partial block, documented workaround in place.

Masada Ben Sira scroll (Mas1h) preserves portions within Sirach 39:27–44:17 with gaps. It is not the only ancient Hebrew Sirach witness: Qumran fragments also survive. Photograph redistribution, published reading consultation and source selection are separate decisions. The witness guide and applied comparison supersede the former 43:30 endpoint and chapter-51 attribution. No automatic replacement of Greek by older Hebrew follows from acquisition alone.

Qumran Tobit fragments (4Q196–4Q200) preserve parts of the book in Aramaic/Hebrew, not a complete source corpus. Compare those actual units with the Greek recensions before choosing a reading. The former ~95% equivalence claim was untested and is withdrawn. An indirect Hebrew back-translation cannot supply missing direct Aramaic evidence. Access and redistribution terms must be recorded for each photograph, transcription and edition, not collapsed into one blanket block.

Scholarly editions (Beentjes 1997, Skehan & Di Lella 1987, Fitzmyer 1995 DJD XIX, etc.) are Zone 2 per REFERENCE_SOURCES.md — consulted as reference during translation, cited in footnotes for factual textual-critical claims, never reproduced.

Source acquisition status

The following table preserves an acquisition-planning snapshot, not a current complete source census or proof of every listed consultation. Consult the dated coverage audit, source-specific guides and research log for actual inspected evidence, access outcomes and completed applications.

Legend: ✓ = acquired and vendored · ◐ = request in flight / partial · (blank) = acquirable but not yet pursued.

Scholarly editions consulted during transcription (not reproduced): Beentjes 1997 (The Book of Ben Sira in Hebrew); Skehan & Di Lella 1987 (Anchor Bible 39); Ben-Ḥayyim 1973 (Academy of the Hebrew Language); the Göttingen critical LXX editions; Fitzmyer 1995 (Discoveries in the Judaean Desert Vol. XIX); Rahlfs-Hanhart 2006 (Stuttgart revised critical LXX); and further literature as relevant per book. These works inform our fresh transcription but are never copied into our published output.

Per-book drafting roadmap

The status of each book in scope, showing its most-original primary source, its Greek fallback (for verses or passages where the original is lost or damaged), and its overall readiness to enter Phase C (drafting). Status updates as sources are transcribed.

Translation pipeline

The translator-prompt builder (part of Phase 8-C — deuterocanon drafting) will assemble per-verse context using the three-zone model defined in REFERENCE_SOURCES.md:

Important contract note. In the source-prep/history layer we still refer to the aggregate Swete containers ADE (Greek Esther stream) and ADA (Greek Daniel-additions stream). In the actual translation layer, the draftable units are now:

• ESG — Additions to Esther (A–F)
• PAZ — Prayer of Azariah and Song of the Three
• SUS — Susanna
• BEL — Bel and the Dragon

ADE and ADA remain useful as audit/source buckets, but they are no longer the preferred Phase 9 drafting targets.

• Zone 1 (vendored): Swete Greek + Sefaria Hebrew parallel (for SIR, TOB) + WLC parallel (for 1ES) + any PD/CC-BY secondary readings. These are the sources the English output is allowed to derive from.
• Zone 2 (consulted): Fitzmyer DJD XIX (TOB), Beentjes (SIR), Skehan & Di Lella (SIR), Göttingen critical editions, etc. The translator-prompt builder injects a registry entry per Zone 2 source (name + usage guidance) into the verse context — no copyrighted text — so the model knows they exist and what role to use them in. tools/hebrew_parallels.lookup_with_consult(book, ch, vs) returns both layers in a single structured call.
• Zone 3 (forbidden): modern commercial English translations. Not consulted during drafting.

This three-zone discipline is what keeps POB CC-BY-redistributable while still benefiting from the full weight of modern critical scholarship. It is the same distinction every serious modern translation pipeline already enforces (NRSV, NABRE, ESV) — we just make it explicit and auditable per verse.

The same four-stage pipeline defined in METHODOLOGY.md applies:

1. Source preparation — vendored Swete LXX under sources/lxx/swete/ (our own OCR, CC-BY 4.0), with Zone 1 Hebrew parallels under sources/lxx/hebrew_parallels/ for SIR, TOB, and 1ES. Textual variants noted in the per-verse YAML source.text and in footnotes where material. Rahlfs-Hanhart and other Zone 2 references are consulted via tools/hebrew_parallels.lookup_with_consult(); see REFERENCE_SOURCES.md.
2. AI draft — GPT-5.4 with the same prompt family used for the Protestant canon, adapted only in the source-edition field. Per-verse YAML emitted with full provenance (ai_draft block).
3. Revision pass — Claude Opus 4.7 reviews drafts per the criteria in REVISION_METHODOLOGY.md. Revisions are discrete git commits; the drafter's ai_draft metadata is preserved.
4. Publication — tagged release on GitHub, mobile bundle regenerated, footnote apparatus cited transparently.

These books receive the same level of rigor as the Protestant canon. No corner-cutting because they are "deutero."

Transcription accuracy — honest expectations

AI-vision transcription accuracy varies sharply by source material. We state our expected accuracy by text type rather than claim uniform rigor, so readers can calibrate their trust in our deuterocanonical sources:

What we are not claiming. We are not claiming machine parity with Beentjes on manuscript plate transcription on a first pass, nor parity with Fitzmyer on Qumran reconstruction. These are specialist works produced over decades of career-deep engagement with specific material.

What we are claiming. The pipeline — fresh vision transcription informed by consultation of all leading scholarship, multi-model disagreement flagging, cross-source verification, and transparent public iteration — produces source-texts adequate for translation work, with every character auditable back to a public-domain photograph and every divergence from published scholarship documented. That is substantively equivalent to specialist-grade source fidelity for translation purposes, which is what the deuterocanonical section of the People's Open Bible is delivering.

Current measured baseline. First-pass Swete LXX transcription quality across all 572 deuterocanonical pages: 100% format-valid, 0.97% first-pass word-error rate on typeset Greek (Opus 4.7 vs GPT-5.4 cross-read on a 24-page stratified sample, ~4,944 words). See sources/lxx/swete/transcribed/TRANSCRIPTION_QUALITY.md for the per-page tally, error-pattern breakdown, and the planned improvement pathway.

Corpus build history — the passes we ran

The deuterocanon corpus was built in a sequence of passes, each one improving on the last. This table is the honest record of what each pass did and the measurable outcome. Latest snapshot numbers live in sources/lxx/swete/CORPUS_HEALTH.md and sources/lxx/swete/QUALITY_BENCHMARK.md; historical snapshots are archived under sources/lxx/swete/benchmarks/ so the trajectory of improvement is visible, not just the latest number.

The residual medium-confidence verses are dominated by obscure proper names in genealogy/place-name lists where Swete himself printed a ligatured or damaged reading from Codex Vaticanus. They are at the natural limit of what OCR + multi-source adjudication can resolve without a specialist paleographer consulting the Vaticanus manuscript itself — so we disclose them rather than fake certainty.

What is and is not "agreement"

When these benchmarks say 72.3% agree / 86.2% functional agreement with First1KGreek, they measure our corpus against Harvard/Leipzig's independent TEI-XML encoding of the same Swete 1909 edition (CC-BY-SA 4.0). First1KGreek is not ground truth. Both encodings are attempts to transcribe the same 1909 pages. Disagreements mean one of:

• Our OCR error (their reading is correct).
• Their encoding error (our reading is correct).
• Legitimate textual-tradition difference (Swete's diplomatic Vaticanus vs. First1KGreek's eclectic choice at the same lemma).

The scan adjudicator (Pass 4) resolves these by looking at what the actual printed page says. The 805 "major" mismatches that remain were independently verified against the scan — which means on those specific verses our corpus is more faithful to Swete's diplomatic edition than First1KGreek is. The benchmark undersells our fidelity where Swete and the eclectic tradition diverge.

Verifying this yourself

# Regenerate First1KGreek benchmark
python3 tools/generate_quality_benchmark.py

# Regenerate per-book health report
python3 tools/generate_master_benchmark.py

# Spot-check a specific verse's adjudication reasoning
jq '.verdicts[] | select(.verse == 15)' \
  sources/lxx/swete/adjudications/SIR_042.json

The adjudications/ directory is itself the audit trail. Every verdict in it records which reading won, which candidates lost, and the model's one-to-two-sentence reasoning citing specific features of the scan. This is the layer where trust is earned — not in a final benchmark number, but in ~3,500 individual verse-level decisions that a reader can audit at will.

Consulting scholarship, reproducing nothing

The scholarly apparatus around Hebrew Sirach and Qumran Tobit is the product of careers of specialist labor (Beentjes 1997 on Hebrew Sirach; Fitzmyer 1995 and the DJD volumes on Qumran Tobit; Skehan & Di Lella 1987; the Göttingen LXX critical editions for the Greek; among many others). These works are indispensable to accurate translation, and we consult them actively during our work.

What copyright restricts is reproduction of their specific text, not consultation. This distinction is the same one that governs every serious modern translation: a translator reads every major scholarly edition, weighs every variant, is informed by every argued reading — and then produces their own fresh work. The published output is the translator's own creation, informed by the literature but not copied from it. This is standard scholarly practice; it is uncontroversial under copyright law; it is what the NRSV, NABRE, ESV, and every other modern translation have always done.

Our commitment:

• We consult the leading scholarly editions actively during transcription and translation. Beentjes 1997 for Hebrew Sirach, Fitzmyer 1995 for Qumran Tobit, Skehan & Di Lella 1987 for English Sirach, the Göttingen critical editions for LXX variants, and further literature as needed. Where we notice a divergence between our fresh transcription and established scholarship, we weigh the evidence, document our reasoning, and either revise our transcription or note the disagreement with explanation.
• Our published transcription is our own creative work — produced freshly from public-domain photographs, informed by consultation of all relevant scholarship, but not a reproduction of any scholarly edition's specific transcribed text. This is a fresh and independent transcription, not a compilation.
• We cite scholarly conclusions in footnotes where they inform translation decisions — which is fact-level citation, not reproduction of creative expression, and is uncontroversial under US copyright law (Feist v. Rural Telephone Service, 1991). Example footnote shape: "Hebrew MS B at this verse reads differently from the Greek; the Greek appears to smooth an idiomatic Hebrew expression. See Beentjes (1997) and Skehan & Di Lella, Anchor Bible 39 (1987)."

The practical outcome: our transcription pipeline is informed by every scholar who has worked this material, while our output is cleanly our own. That is the path to specialist-grade accuracy without reproducing specialist-owned text.

Labeling in the final product

In both the repository and the mobile app:

• A clear section header: "Deuterocanonical Books (Apocrypha)"
• An introductory note to the section explaining the canonical-status landscape across traditions, and naming that POB takes no position.
• Each book's header displays its canonical status at a glance:
  • Canonical in: Roman Catholic, Eastern Orthodox
  • Considered useful but not canonical in: most Protestant traditions
• Internal references (cross-references in commentary, reading plans, search results) treat these books as first-class citizens of the library while preserving the label.

The reader should never be confused about what they are reading and which traditions receive it as Scripture.

Phasing

Phase A — Source acquisition (in flight, 2026-04-18).

The phases and estimates below are historical planning, not current readiness certificates or delivery promises. Completed Sirach applications are linked above; fresh manuscript readings still require the method's calibration gates.

• ✓ Swete LXX vendored: DjVu OCR text in repo, full PDFs manifested at sources/lxx/swete/MANIFEST.md with SHA-256 hashes.
• ✓ Schechter 1899 Wisdom of Ben Sira vendored in repo (PDF + OCR, sources/hebrew_sirach/schechter_1899/).
• ◐ Masada Ben Sira scroll and Qumran Tobit fragment photographs: acquisition in progress.
• ◐ Fresh transcription pipeline from public-domain photographs scoped for Cairo Genizah MSS C–F (sources/hebrew_sirach/ genizah_photos/README.md).

Phase B — Vision-transcribe LXX and Hebrew Sirach working text.

• Per-page vision-based transcription of Swete Greek into clean UTF-8 Greek with breathings and accents, committed to sources/lxx/swete/transcribed/.
• Per-folio vision-based transcription of Schechter 1899 Hebrew into clean UTF-8 Hebrew, committed to sources/hebrew_sirach/schechter_1899/transcribed/.
• Machine-readable verse indexes connect book/chapter/verse to source-text and source-image provenance.

Phase C — Draft and revise translations.

• Extend tools/draft.py to handle LXX and Hebrew Sirach source identifiers.
• Draft each deuterocanonical book from its most-original extant source per the per-book matrix above.
• Revision pass (Claude Opus) per REVISION_METHODOLOGY.md.
• Mobile bundle integration, tagged release.

Phase D — Pre-medieval Hebrew & Aramaic enrichment (when available).

• Compare published readings or calibrated fresh transcriptions of Qumran Tobit fragments; promote only adjudicated, actually surviving units.
• Compare Masada Ben Sira within 39:27–44:17, retaining gaps; a transcription or older copy does not automatically become the primary source.
• Republish tagged release including Phase D sources.

2 Esdras / 4 Ezra and the Ethiopian wider canon (1 Enoch, Jubilees, etc.) are not committed to any phase. If they are ever added, it will be under a separate dedicated strategy document specific to their distinct textual traditions.

Phase summary table

Commitments

We commit, as a project, to:

1. Equal rigor. These books receive the same drafter/reviser pipeline, the same per-verse provenance, the same footnote standards as the Protestant canon. No shortcuts because they are deuterocanonical.
2. Transparent canonicity labeling. Every book's header shows its canonical status across major traditions. Readers are never misled about what they are reading.
3. No doctrinal importation. Translation decisions follow the same DOCTRINE.md principles. We do not soften readings to align with Protestant doubts about these books, nor expand readings to strengthen Catholic/Orthodox claims about them.
4. Public scrutiny before publication. This strategy document is published before any verse is drafted so the approach itself can be challenged.
5. Honest limitations. We document what our LXX-only approach does and does not capture. A reader who wants Hebrew Sirach variants at every verse will need to consult scholarly editions alongside POB. That is stated openly, not obscured.

Why this honors the texts

These books are read as Scripture by hundreds of millions of Christians and have been for two millennia. They shaped the vocabulary of the New Testament (James echoes Sirach; Hebrews 11 lists Maccabean martyrs alongside the patriarchs). They were the devotional library of Jesus's own world.

Translating them carefully, with provenance and honest footnotes, under a free license, with the best tools we have — this honors them as Scripture for those who receive them as Scripture, and as early Jewish religious literature for those who do not. It refuses the shortcut of deciding for the reader what counts.

That refusal is the whole point of the People's Open Bible.

The Pseudepigrapha and expanded canon

Beyond the 13-book LXX deuterocanonical core covered above, several works are received as Scripture by specific Christian traditions (or are cited directly by New Testament writers) and warrant the same scan-grounded, multi-witness, CC-BY-4.0 treatment we applied to the LXX books. These are not in our deuterocanonical phase — each has its own source-acquisition pipeline because the textual situation differs substantially between books and between the LXX tradition.

Why include them:

• 2 Esdras / 4 Ezra is in the KJV 1611 Apocrypha, appears in the Vulgate appendix, is accepted by some Slavonic Orthodox traditions, and is a major text of early Jewish apocalyptic.
• 1 Enoch is canonical in the Ethiopian Orthodox Tewahedo and Eritrean Orthodox churches (part of their 81-book canon) and is quoted by name in Jude 14-15 of the New Testament. Readers of Jude benefit from direct access to the cited source; Ethiopian Orthodox readers have had no permissively-licensed modern English translation.

Why we approach them separately:

• Different primary languages. LXX work is Greek-based and uses our Swete + First1KGreek + Rahlfs + Amicarelli adjudication pipeline. 2 Esdras is Latin-primary with 6 daughter translations as witnesses. 1 Enoch is Ge'ez-primary with Greek fragments and (Zone 2) Qumran Aramaic fragments as witnesses.
• Different OCR backends. Azure GPT-5 handles Greek, Hebrew, and Latin at ≥98% high-confidence. It fails on Ge'ez (validated 2026-04-21), and Gemini 2.5 Flash hallucinates Ge'ez content. The Enoch pipeline therefore uses Gemini 2.5 Pro in plaintext mode for OCR, while LXX and 2 Esdras continue to use Azure GPT-5. We choose the tool that actually works for each script.
• Different textual-critical logic. LXX is "which of these Greek transcriptions matches the scan?" 2 Esdras is "do the daughter translations agree or disagree with the Latin?" Enoch is "does our Ge'ez OCR agree with the Beta maṣāḥǝft oracle, and where Greek fragments exist, do they corroborate?"

Works currently pursued

Shared Ethiopic pipeline

Enoch and Jubilees share a single Ge'ez OCR pipeline at tools/ethiopic/. Both use Gemini 2.5 Pro in plaintext mode with low thinking budget — the only combination we've validated for Ethiopic script. Azure GPT-5 fails entirely on Ge'ez despite handling Greek / Hebrew / Latin at 98%+ confidence; Gemini 2.5 Flash hallucinates Ge'ez content. Pro in plaintext mode produces character-accurate Ge'ez validated against the Beta maṣāḥǝft ground-truth digital text.

Building Phase 11 (Enoch) establishes the pipeline; Phase 12 (Jubilees) reuses it by feeding different source scans. This is why the Pseudepigrapha track sequences Enoch before Jubilees — not because Jubilees is less important, but because the shared infrastructure is built during Enoch and reused for Jubilees.

Works under consideration for later phases

If and when the above complete successfully, natural extensions in the Ethiopian broader canon include:

• Meqabyan 1–3 (Ge'ez-only, Ethiopian Orthodox canonical)
• 4 Baruch / Paraleipomena Jeremiou (Greek + Ge'ez witnesses)
• Testaments of the Twelve Patriarchs (Greek primary)

None of the works listed here are committed to in this document. (2 Baruch is now tracked separately in 2BARUCH.md as part of the broader extra-canonical roadmap.) They are named here only to keep the long-term trajectory visible.

What "pursue" means here

For each of these works our standards remain identical to the LXX deuterocanon:

• Primary source under a clean, permissive license (PD or compatible), OCR'd by us or vendored directly where a clean edition exists.
• Secondary witnesses where the textual tradition provides them, integrated as per-verse apparatus in the YAML.
• Zone 2 scholarship (Milik, Nickelsburg, etc.) consulted actively but never reproduced — standard scholarly practice.
• Per-verse provenance, auditable on the public reader, with every translation decision documented at the same granularity as the LXX books.
• CC-BY 4.0 output, free to redistribute.

We do not pronounce on whether any of these works are Scripture for any reader. We make them available carefully, transparently, and under a free license, and let each tradition and each reader judge for themselves.