About project translations

A small number of texts on /sources/ are rendered into English by this project rather than by a named public-domain translator. This page documents how those translations are produced, what readers can rely on, and what they cannot.

Project translations are paraphrase-level content, not verified primary sources. Do not place a project-translated passage inside quotation marks attributed to the original author. The Greek, Latin, or German source — named in the frontmatter of every project-translated chapter — is the citable text. Use the English here to read, to orient, and to triangulate; quote the source.
On this page
  1. What this is and why
  2. Methodology
  3. Source-chain transparency
  4. What readers can rely on
  5. License
  6. Corrections
  7. Current project translations

What this is and why

The /sources/ section of this site collects the foundational and esoteric texts that recur across the Steiner corpus — the Bhagavad Gita, the Hermetic Corpus, the Edda, the Rhineland mystics, the Chaldean Oracles, and so on. Each entry links to a readable text in English. The English used is, in almost every case, a public-domain translation by a named scholar: Cary’s Dante, Mead’s Hermes Trismegistus, Brodeur’s Prose Edda, Bellows’s Poetic Edda, Sola and Raphall’s Talmud, and others.

A few foundational texts have no usable public-domain English. The available modern English is copyrighted; the available public-domain English is fragmentary, archaic to the point of obstructing reading, or omits portions that are load-bearing for cross-reference with the Steiner corpus. For these, this project produces an English text directly from the source language (and, where one exists, a public-domain bridge translation), and publishes it on /sources/ under a clear flag.

The goal is not to add to the scholarly translation literature. It is to make the foundational text reachable, in the same browser tab as the Steiner passage that depends on it, with the source-language original one click away.

Methodology

Project translations follow a four-step process, called the term-anchored LLM multi-source method (methodology version translation-pipeline-v1).

  1. Source acquisition. Locate the original-language text in a public-domain critical edition. Where one exists, also locate a public-domain bridge translation (e.g., Berthelot’s 1887 French of Greek alchemical texts; Winkworth’s 1857 English of Tauler). The full source chain is recorded in the chapter frontmatter.
  2. Terminology anchoring. Before translation, a fixed glossary is established for the source tradition — technical terms whose translation can shift the substance (e.g. for the Rhineland mystics: Grund, Gelassenheit, Wesen, Vernunft). Anchored terms are either preserved untranslated or rendered with a fixed English equivalent, consistently across the whole text. The glossary is recorded with the translation.
  3. Multi-source pass. A current language model produces English from the source language, reconciling against the public-domain bridge translation where one exists. Disagreements between the original and the bridge translation are surfaced as translator’s footnotes rather than silently resolved.
  4. Self-review pass. The same model is asked to review its own output for consistency with the glossary and the source. Ambiguous source passages (corrupt manuscript readings, OCR damage, philologically contested terms) are kept ambiguous in English and flagged with a translator’s footnote.

Existing copyrighted scholarly translations are consulted as terminology checks — that is, to verify that a particular Greek or German technical term has not been rendered by an idiosyncratic choice — but their wording is not used.

What this method is good at: rendering technical vocabulary consistently, preserving the structure of the original, and not silently smoothing over difficult passages. What it is not good at: producing translations of literary distinction. The English reads as utilitarian. That is the trade-off.

Source-chain transparency

Every project-translated chapter carries a frontmatter block declaring its source chain and method. The fields are:

Why translations are anonymous

This project is anonymous by design. Both the choice of which texts to translate and the choice of which terms to anchor are opinionated — they are made in service of the Steiner-corpus reader’s use case, not in service of general philological scholarship. Naming a translator would invite the reader to weight the translation against that person’s credentials. The project would prefer the reader to weight it against the named source and the named method.

What readers can rely on

A project translation can be relied on for:

A project translation should not be relied on for:

The MCP server’s trust model reflects this: project translations are not exposed as fetch_passage Steiner-corpus results. They live on /sources/, marked with the disclaimer, with the source language reachable from the same page.

License

Project translations — and all other project-original content on this site (curatorial summaries, chapter descriptors, topic pages) — are dedicated to the public domain under CC0 1.0 Universal. You may copy, adapt, redistribute, and build upon them for any purpose, including commercially, without permission and without attribution. The project asks no credit and attaches no conditions: the work is offered freely and unencumbered.

The underlying primary sources are themselves public domain; where a particular edition is republished, it is credited by a link to its source. Citing the source-language original named in the chapter frontmatter, and this page (https://anthroposophy.ai/about/translations/) for the methodology, remains good scholarly practice — but it is a courtesy, not a requirement.

Corrections

Corrections are welcome, particularly where a technical term has been mis-anchored or where a footnote should have been added and was not. The project does not maintain a contact form for general inquiries, but errata reported via any channel that reaches the operator will be triaged and, if substantive, reflected in the next deploy with the chapter’s extracted date bumped.

Where a translation is materially revised after first deploy, the previous version remains reachable via archive.org. The translation is not silently rewritten.

Current project translations

This list is the canonical index of project-translated material on the site. Each entry resolves to its reader page on /sources/.

The list grows slowly. Each addition is deliberate: a text is added when the Steiner corpus depends on it, no usable public-domain English exists, and the methodology above can carry the load.