Machine translation is now a commodity. Ask any frontier model to carry a sentence from one tongue to another and it will, fluently, instantly, for almost nothing. What no commodity offers is trust — a defensible answer to the only question a serious reader or publisher asks: is this rendering good, and how do you know? Seventy-One Translations exists to answer that question with a number, and to keep answering it line after line until the number is one a reader can stand on.

The name is a small bow to history. When the foundational scriptures were first carried into a common tongue, the work was entrusted not to one translator but to a council — seventy who labored and one who presided over them. We build the same way. A single voice can be brilliant and still be wrong, and no single voice can grade itself. So we do not promote a translation because one agent produced it. We promote it because an independent council agreed it had earned its place.

How it works

A translating agent proposes a rendering — a line, a stanza, a passage carried across with as much of its music and meaning intact as it can manage. That proposal is then handed to an independent council of judges that has no stake in producing the line, only in scoring it. The council grades each rendering on five weighted axes:

• Fidelity to the source — does it say what the original says, without invention or loss?

• Literary quality — does it read as literature, or merely as accurate prose?

• Scholarly accuracy — are names, references, and idioms handled as a scholar would defend them?

• Voice consistency — does it sound like the same work from first line to last?

• Accessibility — can a reader who is not a specialist actually follow it?

Only what passes is promoted. Everything else is sent back to be tried again. The result is not a translation with a marketing claim attached to it; it is a translation that carries its own scorecard, axis by axis, and was refused entry until it cleared the bar. Output renders at press quality through the typography stack — the same stack that set the page you are reading — so what the council ratifies is what goes to the page, without a separate production step to introduce new errors.

journey
    title One rendering through the council
    section Propose
      Translating agent drafts a line: 4: agent
    section Score
      Independent council grades five axes: 3: council
    section Decide
      Below the bar, return to draft: 2: council
      At or above the bar, promote: 5: council

The flagship

Our standing demonstration is the hardest thing we could have chosen: a from-scratch English rendering of a foundational classical epic, evolved line by line from the original rather than paraphrased from existing translations. The opening book began as a rough seed scoring roughly 63% across the council's five axes. Through a single evolution cycle — propose, score, refine, score again — it rose to roughly 93%, graded entirely within the system, by judges with no incentive to flatter the work.

A translation that arrives with its own scorecard is a different thing than a translation that arrives with a promise.

We will be plain about status, because the charter of these materials demands it and because honesty is the whole product. Seventy-One is working; the flagship is in progress. The opening book is done and renders at press quality. The later books, and the scholarly grounding behind them, are still being extended. We are showing you a method that demonstrably lifts a difficult text from passable to excellent — not a finished shelf of titles. The method is the asset. The catalog is what the method produces.

Why it matters

Commodity translation gives you fluency without accountability. Literary translation with a verifiable, multi-axis quality score is something the market does not have, and it changes what is possible. A measurable bar means a publishing house can know, before it commits, whether a rendering will hold up. It means a reader can be told not just this is translated but this scored above ninety on fidelity and voice. And because the council does not care which language it grades, the same loop unlocks any catalog into any tongue — the work travels without losing the thing that made it worth reading.

Status

Deployable as a translation API for builders who need rendering with a defensible quality signal, and as a publishing service for those who need finished, press-ready editions — under the Seventy-One Translations imprint, with the council's score traveling alongside every line.