What this edition is, how it was made, and what it found.
This is an edition of five verses and of what was said about them. Genesis 1:1–5 is read as ten cruxes — a crux being a difficulty that more than one tradition felt it had to answer — and on each crux every witness the edition could find is printed in its own language and in English, dated, placed, sourced to the column or the leaf it came from, and threaded to the other witnesses it cites, echoes, contests, transmits, or merely runs parallel to. The unit is not the commentary and not the commentator. It is the argument: one continuous piece of reasoning on one verse from one work. Augustine has five of them on the second verse alone.
Three benches sit on each page. The Latin runs from Origen in Rufinus's translation to Nicholas of Lyra and Paul of Burgos, and is drawn almost entirely from the Patrologia — the exceptions, a fourth-century Platonist and a fifteenth-century printed Bible, are transcribed from page images and say so. The rabbinic runs from the targums through Bereshit Rabbah and the Bavli to Rashi, Ibn Ezra and Ramban, and is drawn from Sefaria. The third bench is the smallest and the oldest: the Septuagint, and Philo of Alexandria, who read these verses in Greek before either of the other benches existed and who turns out to be the head of arguments both of them make. A witness's English is either a published translation, named and licensed, or a fresh draft made for this edition against a fixed list of renderings and marked as a draft until it was read. The colophon accounts for every text by licence.
The threads are the edition's claims, and each carries its evidence. A thread sayscites only when a witness names its source; echoes when the reuse is verbatim and unattributed; contests when one reading argues with another it has actually met;transmits when a text is the channel by which a reading reached a later bench. A thread that says parallel is a confession: the same image, the same argument, the same proof text, on two benches, with no channel between them that the edition can show. There are more of those than of any other kind, and the edition has not tried to explain them away.
What the ten findings add up to is not one thing but four. The first is that the two large benches, reading the same five verses in different languages for a thousand years, keep arriving at the same place by different roads. Both gloss the first word of Scripture asin wisdom and prove it from a wisdom text — a different wisdom text on each bench, and a different Wisdom. Both decide what the dividing of light from darkness means by reading a silence, and they find the same silence at consecutive clauses of the same verse. Both refuse the question of whether heaven or earth came first, and refuse it with the same psalm. Three benches choose the same two verses out of the Pentateuch to test the divine plural, by the same criterion, and none of them has read another. The pairing is not borrowed. It is what the Hebrew does to anyone who reads it carefully.
The second is that where contact existed, it was lost, and it was lost in both directions. Chalcidius in the fourth century reports Origen persuaded of the non-temporal beginningab Hebraeis, quoting the rabbinic bench's own verse, and nobody read Chalcidius. Basil's Latin Hexaemeron carries an accurate report of the rabbinic reading ofFaciamus hominem, and Bede copies it at Jarrow keeping every sentence of the argument and removing every Jew in it. And the rabbinic bench, which cites Philo not once, lost the one Jewish reader of Genesis 1 whose reading of its first verse its own commentary most resembles; he survives because the church copied him, and was recovered for Judaism only in the sixteenth century. Jerome, the one Latin who read the Hebrew, refuted the reading ofin principio as in the Son, and the Carolingians copied his page and dropped the refutation. Transmission on these benches is not a channel. It is a filter, and what it filters out is the knowledge that anyone else was in the room.
The third is one page. At Genesis 1:4, in a printed Latin Bible of 1487, Nicholas of Lyra calls the answer of Rabanus and Peter Lombard a piece of trifling and takes the reading he wants from Rashi; Paul of Burgos, who had been a rabbi in that city until he was forty and was afterwards its bishop, replies that Lyra credits Rabbi Solomon the Hebrew with what our own doctors found first; and Matthias Doering, defending Lyra, restates the charge and changes one word — he attributes the true answer to an unbeliever. Nobody in that exchange denies that the answer is right. An early reader of this copy underlinedRa. Sa. in ink on both leaves. Every other page of the edition shows two traditions reading the same verses without knowing the other is there; this one shows what it sounds like when they do.
The fourth is that the modern translations settle the cruxes by removing them. The ancient versions all kept day one; the English on both benches prints the first day, and the note that exists to ask why it is not first glosses the lemma as first. The JPS Tanakh prints When God began to create, which is Rashi's plain sense and Ibn Ezra's construct adopted as the text — the one reading no Latin witness could have raised, and the one Ramban had already refuted. The question each crux is made of survives only in the languages nobody now reads it in. That is the reason for printing the languages.
A page is laid out as a daf. The verse sits in the centre in Hebrew, Latin and English, with the Septuagint and the Vulgate beneath it and the Glossa's marginal glosses beneath those. The Fathers are in the left column and Philo above them; the rabbinic bench is on the right; the Carolingian and twelfth-century schools run across the foot. Every note opens on its lemma and an excerpt; hovering opens the whole note and draws its threads across the page; clicking pins it. The dialogue pages show the same witnesses by verse rather than by crux, and the finding on each crux page is the editor's own reading of what the page shows.
The edition was made in the late summer of 2026. Its data, its build, and every decision taken in making it are in the open repository, including the decisions that were taken and then reversed.