Welcome · Link 1 of 10
Three thousand years of argument, and the losing side is still on the page.
Welcome! I'm Joshua, an engineer. A famous essay recently asked whether the world's oldest library might hold the secret to making AI honest - because for three thousand years it never deleted an argument and always named who said what. I had quietly spent ten years building the answer. This site is it - ten short steps, every claim checkable. The magic was not invented here. It was engineered into these books three thousand years ago; I only turned the lights on.
The chain in the menu is the whole site. Each link is one short step, and they build in order - but every one stands on its own, so jump anywhere you like. The receipts live on the evidence page, the counting on the methodology page, and the experiments still to run on the agenda. Nothing here asks anything of you until the last link, which asks one thing.
What is the Torah?
The five books, the oral tradition, and the rules that kept them alive.
The machineTorah & AI
A knowledge graph of the tradition - and what it taught a modern AI.
Your first stepStart learning
Clear first steps, whatever your background - including none at all.
Discover · Link 2 of 10
What is the Torah?
Say "the Torah" and most people picture a scroll. What is on that scroll is five books, Genesis through Deuteronomy, given at Sinai - and it is the smallest part of the answer. In the full sense the Torah is something much larger: the entire transmitted teaching of the Jewish people, written and oral, carried from teacher to student for over three thousand years and still growing today.
That teaching lives on a bookshelf. Here is what sits on it:
- Tanakh תנ״ך
- The written scripture: Torah (the Five Books), the Prophets, and the Writings.
- Mishnah משנה
- The oral law, organized and written down around 200 CE - the first great act of preservation after exile.
- Gemara גמרא
- Centuries of recorded argument about the Mishnah. Together they form the Talmud - question, objection, resolution, and sometimes an honest "unresolved."
- Commentaries
- Rashi, Tosafot, Rambam and hundreds more - layers of explanation, each citing the layers before it, printed together on a single page.
- Codes
- The law organized for living: what to do, derived from all of the above and traceable back to it.
- Responsa
- Questions and answers, from the age of the Geonim to your generation. New ones are being written right now.
The rules that kept it alive
What makes this bookshelf different from every other ancient library is a set of editorial rules the tradition enforces on itself:
- Cite the source. A teaching is reported in the name of the one who said it. Attribution is a religious obligation, not academic etiquette.
- Keep the losing argument. Rejected opinions stay in the canon, permanently - available to any generation that discovers the rejection was wrong.
- Never delete - reinterpret. The text is not rewritten. New understanding is layered on top, with the full record kept underneath.
- Record the unresolved. The Talmud has a formal verdict for an open question: teiku - it stands. An honest "we don't know yet" is a legitimate result.
This is not a museum. Every layer of that shelf is studied every single day, argued with, and added to. It is one of the longest-running collaborative projects in human history - and you are allowed to walk in.
Discover · Link 3 of 10
The unbroken chain
How does a teaching cross three thousand years with no empire to guard it? No army, no capital, exile after exile - and it arrives intact, with the names still attached? The tradition wrote its own answer at the front of Pirkei Avot:
The claim of this site is simple, and it is not modest. As a system for keeping knowledge honest - sourced, versioned, argued in the open, silent where it has no license to speak - the Torah’s architecture is still ahead of the AI industry. The pages below make that case piece by piece, and a live graph lets you check it.
"Moses received the Torah at Sinai, and handed it on to Joshua; Joshua to the elders; the elders to the prophets…"the opening of Pirkei Avot, the tradition's own record of its transmission
Every teacher in this tradition learned from a teacher, who learned from a teacher, in a chain that runs back further than almost any institution on earth. Exile scattered Jewish communities across every continent - and the chain held, because it was carried in books built by the rules you read, and in the teacher-student bond those books require.
On this site the chain is not a metaphor. Draw the classical sources as a map of who taught whom, and the chain that opens Pirkei Avot appears on its own. Thousands of teachers and students, most of them joined in one unbroken web from the era of the Mishnah onward. It is still there, and it still holds.
Here is the part that is about you: the chain has no closed end. It grows the same way it always has - someone decides to learn. That is the entire admission requirement.
Explore · Link 4 of 10
Torah & computer science
Here is the thing that changed how I think as an engineer. Every system I spent my career building - keeping every old version, tracking who said what, showing where a fact came from, keeping the argument on the record - was built here first. On paper. And it ran for two thousand years without losing anything.
- A versioned knowledge base. It was edited for about 1,500 years, and every new layer was laid on top of the ones before without rubbing them out. You do not delete. You answer.
- A citation network on paper. Open any page: the text sits in the middle, the people arguing with it run down the margins, and the pointers to other pages sit around the edge. Printers in Venice set it that way in 1523, and nobody has moved it since.
- Named inference rules - the thirteen principles of Rabbi Yishmael. Recited daily for two thousand years, they are a formal system. Kal vachomer - the a fortiori argument. Gezerah shavah - analogy, but only where the tradition grants the license; you may not invent one. Binyan av - generalizing from cases. Ten more besides. An inference stands only when a named rule licenses it. Today’s language models still reason with neither the names nor the licenses.
- A verdict for the unresolved. Teiku - the recorded open problem. It is written down and kept, exactly like any other answer. Machines are still learning to admit that much.
- Built-in adversarial review. Almost nothing stands without its strongest objection printed right beside it. The objection is kept even after it loses.
Not analogy - mathematics
This is not a nice comparison I made up. Since 2008 the logician Dov Gabbay - with Michael Abraham and Uri Schild - has been writing this reasoning out as real mathematics. The result is the Studies in Talmudic Logic series: more than a dozen volumes and counting, peer-reviewed, shelved with the rest of formal logic.
His results read like a translation table between the beit midrash and the machine - each one takes something the Talmud has always done and shows what it solves:
- Every statement travels with its name tag
- In his logic, a statement never travels alone - it carries a tag saying where it came from, exactly as the Talmud's page always has. For AI, that is how an answer shows its source instead of asking to be trusted.
- “How much more so” became a method
- The Talmud’s signature argument (kal vachomer) turns out to be a precise way of filling in a half-finished table of cases (Studia Logica, 2009). For AI, it is a checkable way to extend known rules to new situations - with the working shown.
- The objection is part of the answer
- In his formal treatment, a claim stands only with its strongest challenge recorded beside it - the Talmud’s question-and-answer rhythm, made into a soundness condition. For AI, that is review built into the knowledge itself, so it holds up the first time someone pushes.
- Minds may be changed in an orderly way
- He works in the logic of conclusions that new information can revise - without the rest of the system falling over. The tradition has done this since the Mishnah: layer, never erase. For AI, it means knowledge that can be updated instead of rebuilt.
- Different kinds of reasoning join without breaking
- He showed how to join different kinds of reasoning into one system without breaking either - which a page of Talmud has done all along, running law, math and story side by side. For AI, it is the recipe for a machine that can think in more than one way.
- “Unresolved” is a real verdict
- The Talmud records teiku - the question stands - as a permanent result. For AI, that is a system allowed to say “unknown” and have it count as an answer.
- The Ship of Theseus has an answer on the shelf
- Whether a thing rebuilt piece by piece is still itself has a formal treatment grounded in Talmudic sources (Logica Universalis, 2019). For AI, that is the deepest question there is: whether a system that rewrites itself is still the system you trusted. The tradition has been working that problem for two thousand years.
And this work has weight in its field: in Volume 18 of the Handbook of Philosophical Logic, the longest chapter of all is "Principles of Talmudic Logic."
If you want the full argument with every citation, it is written up in the research paper. If you want to feel it, keep going - the next link is where it meets AI.
Explore · Link 5 of 10
Torah & AI - now, and next
Modern AI has a truthfulness problem: machines invent sources that sound right. Here is the astonishing part - the Torah's editors solved exactly that problem, on paper, with one rule: report a teaching in the name of the one who said it, or don't report it. I spent years teaching a machine that discipline.
What I built
A knowledge graph of the classical Jewish canon. Every rabbinic teaching in it carries a reference you can look up, all but six carry the Hebrew source text, and each is linked to the verses it turns on. On top of it sits a council of AI voices - David, Abraham, Moses, Solomon, Hillel, Akiva, Rashbi and the Rambam - each one grounded in the graph. They answer only from the graph, and they cite everything they quote. You can talk to one right now at app.openrabbi.com.
Why this is the future
The hardest open problems in AI - knowing where an answer came from, revising a system without corrupting it, keeping your values while rewriting yourself - are not new problems. They are the problems the Torah's editorial discipline was engineered to solve, and it holds the longest continuous track record on earth of solving them. That is why this tradition belongs in the future of AI, not only its past.
The Hebrew text in the graph comes from the open Torah library; the corpus is predominantly CC BY-NC, with a CC BY minority.
And one more thing. If a rule this old can steady a machine this new - what else is sitting on that shelf, waiting? The next link is what surprised me most.
Explore · Link 6 of 10
Things that surprised me
Eighteen findings from the books, the history, the record and the graph!
Learn · Link 7 of 10
Learn - start here
What would it take for you to open one of these books tonight?
You don't need Hebrew. You don't need a religious background. You don't need anyone's permission. Every path below starts from zero and uses sites I built so the first step would be easy. Pick the one that sounds like you.
"I'm starting from nothing."
Three short visits, in order - how the books are organized, how the reasoning works, and how it reached you:
- torahverse.com - how the Tanakh is organized: books, chapters, verses, the yearly reading cycle.
- talmudiclogic.com - the thirteen rules of Talmudic reasoning, explained plainly.
- torahchain.com - the transmission, teacher to student, visualized as data.
"Let me talk to it."
The most direct door: a conversation where every answer carries its source.
- app.openrabbi.com - speak with an AI sage grounded in the knowledge graph, with a source under every answer.
- openrabbi.com - what the project is and how the grounding works.
"Give me the deep end."
The mathematics and the machinery:
- Torah & computer science on this page, then the full research paper.
- proofanalyzer.com and truthknowledgebot.com - argument structure as working software.
One honest note: websites are doors, not destinations. Torah has been learned the same way for three thousand years - with a teacher, and with other people. If something here moves you, the real next step is a class, a community, a rabbi. The sites will still be here.
Explore · Link 8 of 10
The sites
I build a site when I want to understand something well enough to explain it out loud. These are the ones that survived. If you only want three, start with talmudiclogic.com, torahverse.com, and app.openrabbi.com.
Torah & Jewish learning - 19 sites
Knowledge, reasoning & libraries - 16 sites
Personal & hub - 2 sites
Engineering & business sites - the work that funds this - 10 sites
The person · Link 9 of 10
My story
For years I lived two lives, and only one of them had witnesses. At 5:50 every morning I was in a study hall, bent over books that argue with themselves. By nine I was on the trading floor at Goldman Sachs, bent over systems that are not allowed to fail. Nobody at the bank knew about the mornings. I barely understood them myself.
Then there is the archive nobody asked for. In 2013 I started saving recorded Torah classes to disk - lectures, essays, short teachings, anything I could keep. I could not have told you why. I only knew that a voice explaining a hard text felt like something the world should not be allowed to lose. I kept collecting for thirteen years.
In June 2015 I gave up the “perfect job” - the one you are not supposed to leave - and started two companies in the same month. One took client work. The other was for this. For years my work email was joshua@torahtech.com, and I would watch people read it twice in meetings, deciding whether to ask.
Here is the part I did not see until much later: every client job was secretly a lesson.
A children's education company, then an Orthodox learning platform built with a team in Israel, then rebuilding a beloved Torah video library - each one taught me that organizing knowledge is harder than writing code. Six years running the websites behind MetaMask taught me to keep a front door open for millions of people entering a new world. Three and a half years at Chainlink - a company whose products are called oracles - taught me how machines report facts from outside themselves. Someone might as well have been writing me a curriculum.
One month before my first day in blockchain, I had already written down the sentence the whole thing turns on: the books of the sages are like the blockchain of history, copied exactly, generation to generation, with very specific rules.
The last piece was the machine. I built an NVIDIA supercomputer in my own house - a research computer whose whole job is to read the Torah's library, connect it, and question it. The archive became a graph. The graph learned to speak. And what makes it worth listening to is the oldest rule on the shelf, written straight into the code: say who said it, or say nothing. Eight voices answer from that library now, out loud, and anyone can talk to them.
So the secret turned out to be the opposite of what I thought. The Torah was never the second career. It is the only thread that runs unbroken from a youth group in San Juan to this page - through the trading floor, through MetaMask, through the oracles - and the engineering was what finally gave it the tooling. This site is where the two lives meet in public. The longer version, in order, is at joshualazoff.com.
The next link · 10 of 10
Write to me
If this page moved you to open a Jewish text, visit one of the sites, or ask a single honest question about the Torah - write and tell me. That is the entire reason this site exists.
I answer engineers who want to argue about the graph, students who don't know where to start, and people who read the Torah & AI panel and want a machine like it in their own house.
joshua@digitaltwinpro.comThe chain in this menu ends with a dotted ring labeled the next link is you. It has no closed end - it grows the way it always has, when someone decides to learn. Tell me what you decided to learn!