A Cartha Project

the People's Open Bible

This is a Bible translation that builds on prior translations and the original Hebrew, Greek, and Aramaic source texts, then uses AI reasoning to render each verse. Anyone can inspect the sources, revision history, and publicly propose revisions. The goal is to make a translation that is the most faithful-to-the-originals and most transparent yet. Our data, methods, and code are published at the highest level of free and open possible.

Read Bible

How we work toward a faithful translation

We compare old copies of the Bible, study the differences, and work toward clear, faithful English.

  1. Compare old copies

    Leningrad Codex

    A Hebrew Bible copied around AD 1008. It is the basis of our main Hebrew source.

    Dead Sea Scrolls

    Much older copies, including the Isaiah Scroll shown here. Many words agree; some differ.

    Old Testament archive sources

    We compare Old Testament manuscripts and archive records to better understand the Hebrew text.

    Close-up examples, not the same verse. Our work brings together manuscript photos, source texts, and earlier translations.

  2. Check the differences

    When copies differ, we weigh the evidence and explain our choice. Older does not always mean better.

  3. Write clear English

    We use earlier translations and AI-assisted review, then check the English against the Hebrew and its context.

What makes this different

A source text built from many manuscripts

  • Bringing the evidence together. We compare surviving manuscripts and choose the best-supported readings to build the source text we translate—not simply follow one copy.
    See the source corpora →
  • Every translation decision is publicly documented. Every verse is a file in the repository with the source text, the English rendering, every lexical decision (source word, chosen gloss, alternatives considered, lexicon entry, rationale), and the contested theological readings with alternatives preserved. Every change is a signed git commit. Every disagreement has a permanent URL.
    See John 1:1’s source record →
  • Every verse is reproducible. Given the source text, the prompt hash, and the model identifier, any third party can re-run the pipeline and verify our documented draft. No other English Bible in history has offered this.
    Read the methodology →

Why we built this

A method, not a claim of authority

We’re not claiming authority. We’re presenting a method: use state-of-the-art AI alongside the original Hebrew, Greek, and Latin manuscripts to render a translation as close to what the original scribes wrote as we can get. Everything that goes into it is open-sourced — the source texts (much of which we had to restore from public-domain manuscripts), every pass from multiple frontier models, every lexical decision, every revision, every disagreement.

Use it freely

Free to use in any way

Released under CC-BY 4.0: read it, fork it, use publicly in any way, at the highest level of free and open possible. All is packaged and organized for anyone to use.