The Talmud is a continuously-edited knowledge base maintained, attacked, and extended by thousands of contributors for roughly 1,500 years. Its reasoning rules have been formalized as mathematical logic. Its argument structure anticipates modern argumentation theory. Its editorial discipline — preserve the disagreement, cite the source, never delete, only reinterpret — is a working answer to problems computer science is only now hitting: provenance, trustworthy revision, and how a system keeps its values while rewriting itself.
Jew.tech maps that convergence. Not by analogy, but with the actual mathematics — the formal logic literature that formalizes Talmudic inference rules as explicit algorithms — and with working systems: knowledge graphs built from Torah sources, AI models trained to cite or stay silent, and measurement regimes borrowed from a tradition that always demanded its claims carry their evidence.
A short history of the convergence
- c. 200 CE The Mishnah is redacted — oral law becomes a versioned, structured text. Dissenting opinions are recorded alongside the rulings that won. Nothing is deleted.
- c. 500 CE The Gemara layers argument onto the Mishnah, question by question. Objection, resolution, distinction, unresolved stand-off (teiku) — each a typed move in a reasoning protocol.
- 1st c. BCE – 2nd c. CE Hillel's seven middot, expanded to Rabbi Ishmael's thirteen — explicit inference rules for deriving law from text. A codified rule system for textual inference, two millennia before formal methods.
- 11th–15th c. Rashi, Tosafot, Rambam: commentary becomes a linked, cross-referenced network. Every page of Talmud is typeset as text surrounded by its citation graph.
- Since 2008 Dov Gabbay's Talmudic Logic Project formalizes the reasoning, volume by volume. Kal vachomer as matrix abduction; deontic, temporal, and non-monotonic logics of the Talmud — peer-reviewed mathematics.
- Now Knowledge graphs, retrieval-grounded language models, argumentation frameworks. Computer science needs what the Talmud engineered: provenance, structured disagreement, revision without amnesia.
Explore the intersections
Talmudic Logic
The thirteen middot and seven core operations of Talmudic argument, read as formal inference rules.
Dov Gabbay & the Mathematics
The logician whose labelled deduction, argumentation networks, and Talmudic Logic Project supply the proofs.
The Sugya Protocol
Structured disagreement as an engineering methodology for trustworthy, updateable knowledge bases.
The Torah Knowledge Graph
157,143 nodes of Jewish text and tradition — cited teachings, verses, and a machine-readable chain of transmission.
All Intersections
Gematria and number theory, halakha as algorithm, oral law as version control, machloket as argumentation theory, and more.
Grounded AI Sages
A case study in AI that cites or stays silent: eval gates, fabrication baits, and measured groundedness.
Research & Data
The corpus behind the site: paper graphs, text archives, and the databases that hold them.
Projects
Building at this intersection — sites, graphs, models, and pipelines.
Why this matters now
Modern AI systems summarize, answer, and forget. Their knowledge is hard to audit, harder to update, and brittle under adversarial pressure. The traditions of Jewish learning developed working answers to a version of each of these problems under the harshest possible test conditions — exile, dispersion, and the repeated destruction of every institution that carried the tradition. What survived was not a database. It was a discipline: keep the argument, label the source, preserve the minority view, and let reinterpretation — never deletion — carry the core forward.
That discipline can be made computable. The pages here show how — with published mathematics, working code, and measured results.