A language model does not translate the way a dictionary does — word in, word out. It reads the whole clause, forms an internal representation of what the clause means, and then writes English out of that representation, one word at a time, each word chosen in light of the entire sentence and everything already written. Researchers who study these systems from the inside find that meaning is stored in a form largely shared across languages: “bread” in English and its Coptic equivalent light up much of the same internal machinery. Translation runs through that shared layer. The unit of fidelity is the sentence, not the word.
Each English word the model writes is a choice among candidates, and two forces pull on that choice. One is the evidence in front of it — the ancient text and its instructions. The other is its training: the patterns absorbed from everything it was built on. For most text the two agree. For scripture they can pull apart. With the four canonical gospels the pull is enormous — the King James and centuries of quotation sit in the training data, so a famous verse tends to come out remembered rather than rendered. The Gospel of Judas is the opposite case. It returned to the world only in 2006, and its English tradition is a handful of scholarly translations, not four centuries of pulpit English. The pull here is subtler: toward the wording of those few published versions, and toward the cadences of the canonical gospels wherever a line in Judas resembles them — a steady pressure to make a strange text sound more familiar than it is.
The corrections here were human. The draft was made with the established scholarly translations beside the Coptic as a control, and the editing that followed treated divergence as a question to settle, not a blemish to smooth away. And the manuscript adds a limit no instruction can lift: the papyrus is damaged, and a machine can only translate words that survive. This edition renders the text as established by its modern editors; where the lines are gone, the gaps and the larger restorations are marked on the page — the Editorial Notes explain the marks — so the machine was never asked to invent what the desert took.
The whole chain stands on the shoulders of the monks who copied the codex, the conservators and scholars who pieced it back together and established its text, and the translators who first carried it into modern languages. The machine is one more lens — with its own distortions, named here so you can look through it with open eyes. Read it as a draft to question.