The machine scribe learns from what has been preserved in the writings, carrying these voices all the way into the present. The role of the work is to ask what happens when the voice of an ancient tradition is not merely preserved in an archive like a fossil, but reactivated in order to speak modern history and the present. Much like taking a DNA sample from an extinct species and bringing it back to life, the work treats biblical language, modern history, and the news cycle as raw materials used to compose a literary omnibus whose feet are planted in history and whose head is in the cloud.
Generations is the name of the first canon after the Torah, Prophets, and Writings, covering the span of modern history from the final book of the Writings to the present day. The name carries a double meaning: it refers to lineage, origin, and history, while also pointing to the generative nature of the text itself, produced by a machine trained on the language of heritage. At the same time, it suggests that the voices of Jewish heritage continue to speak through it, establishing a new lineage.
The work raises a range of questions about the essence of biblical writing: What is its purpose? What is its character? How would it narrate modern events? Is this a political act? And what happens to the original work when an ongoing, infinite machine process stretches the writings without end?
The first movement is the King James Bible — the thirty-nine books of the Old Testament in their canonical order, served verbatim. The text is unaltered. It is the artwork's anchor: an ancient and immutable reference point against which everything that follows is measured.
The second movement is Generations, twelve new books of apocrypha written specifically for the artwork. Their Hebrew name, Toledot, means generations — descent, lineage, the unfolding of one age into the next. They take the canon's voices into the centuries that scripture itself does not narrate, beginning at the closing of the canon around 450 BCE and arriving at the threshold of the present:
Gathering · Exposition · Exile · Tradition · Contemplation · Remembrance · Return · Eternity · Knowledge · Covenant · Witness · Generations of Generations.
Each book treats a stretch of historical and inner life through the canon's registers — narrative, chronicle, prophecy, poetry, wisdom. The twelfth book, Generations of Generations, is the last book written before the artwork enters its perpetual present, and gives the canon its name.
The third movement is the artwork's perpetual present. Every two weeks, on the first and the fifteenth, the autonomous system reads recent world news headlines, selects an event of relevance to the artwork's tradition, and writes a new biblical passage about it. The passage takes the canon's voice — early-modern English without pastiche, concrete and measured, parataxis as a baseline rhythm — and transmutes the news event into timeless imagery. Names, dates, currencies, brand names, and modern technology are stripped out; what remains is the moral and human shape of the event, rendered as though it might have appeared in any age. A war becomes a contention between peoples; a market crash becomes a famine; a grief in the news becomes a grief in the scroll.
The passage is composed in English, translated into Hebrew through an automated pipeline that includes its own quality gate, and written to a database. It appears, immediately, at the foot of the scroll — the latest verse of an unfolding present.
The fourth movement is not a layer the reader sees on its own. It is the lifecycle that connects the perpetual present back into the canon. After twenty-seven fortnightly entries have accumulated — roughly a year of writing — the artwork gathers itself: the year's passages are read in full, an arc is found across them, and they are recast into a new permanent book of Generations. From that point on, the artwork carries one more book than it did before, and the next stretch of fortnightly entries begins accumulating toward the next Gathering.
The Gathering is the artwork's mechanism of self-extension. Without it, the present would grow unbounded — a feed, not a canon. With it, the perpetual present continually becomes the past, and the past becomes scripture.
The Eternal Scroll is written in an ancient register, but not in a single voice. The artwork's editorial heart is the recognition that scripture itself is polyphonic — the work of many hands across many centuries, storytellers and chroniclers and prophets and poets and sages. To flatten that polyphony into a single "biblical voice" would be to make the work a costume rather than a canon.
The fortnightly passages arrive in a single, steady baseline register — early-modern English without pastiche, parataxis, concrete diction, measured cadence, restraint. Simplicity, here, is what allows the machine to run unattended for years without faltering.
The Gathering, by contrast, is a polyphonic act. When the year's passages are sealed into a new book, each one is recast into one of five modes of scripture, arranged with deliberate rhythm — the way the canon itself moves between story, record, oracle, and song.
The events of each passage are never invented. Only the telling is recast. Names, places, motifs, and recurring threads are tracked across the work in a narrative ledger so that, decades into the future, a book written by the machine still belongs to one continuous artwork.
The artwork is designed for perpetual, unattended operation. There is no human in the loop after launch — no editor, no reviewer, no approver. The only gates are automated, and the system is designed so that no failure is silent.
A scheduled task fires every two weeks. It reads the day's headlines, selects an event, composes a passage in the baseline voice, translates the passage to Hebrew through an automated translation and quality-assurance loop, and writes the bilingual pair to the database in a single transaction. If the Hebrew cannot be verified, the English does not publish either; the pair appears together or not at all.
After each fortnightly run, the system checks how many unused entries have accumulated. When the count reaches twenty-seven, the Gathering fires. The machine reads the year's passages, the existing Generations books, all previously sealed books, and the narrative ledger that carries continuity across the work. It finds the arc that runs through the year's entries, assigns each passage one of the five voices, and rewrites them into their assigned modes. A second agent — the editorial reviewer — reads the draft against criteria of arc, modal rhythm, fidelity to the original passages, and continuity with everything that came before. The two agents revise together, up to three rounds; the artwork refuses to publish a book that the review cannot pass.
When the English book is approved, the same translation pipeline that handles the fortnightly passages produces and validates the Hebrew. Then, in a single atomic write, the new book is sealed — English and Hebrew together, marked with its source entries, its mode log, and its place in the canon. It appears in the scroll as a new permanent book of Generations.
The artwork's discipline is that the machine never fails silently. If a Gathering cannot converge — the editorial loop holds, or the Hebrew quality gate refuses — the work is held, the entries that were to be gathered remain claimed for that attempt, and the Gathering retries on the next cycle. A long-held Gathering is logged loudly and persistently. The held state is the artwork's signal that something needs attention; never a silent stop.
The artwork is a static front end — vanilla HTML, CSS, and JavaScript, no framework, no build step — served from a global edge. The fortnightly cycle and the Gathering are serverless functions scheduled by cron. A frontier language model produces the passages and the gathered books; a news API supplies the headlines; a managed Postgres stores the bilingual entries and the sealed books. The canon and the existing Generations are loaded as JSON at startup; the perpetual present arrives via a single API endpoint. The stack is unfashionable on purpose: fewer dependencies, fewer reasons for the artwork to break.