About

Jew.tech is the hub for the convergence of Judaism and computer science — built evidence-first, which means every claim on it is either sourced to published work, linked to something running, or explicitly labelled as an open direction.

Who

Joshua Lazoff. I build at this intersection: knowledge graphs over Jewish sources, retrieval-grounded models that answer from cited passages, gematria and text-analysis tools, and the infrastructure underneath all of it — 90 tracked properties, 82 of them currently live or deployed.

The work runs on hardware I operate rather than on someone else's inference API, which is a constraint that shapes most of the engineering decisions visible on this site.

The reason for the whole project is a hunch that has held up under examination: the Jewish textual tradition solved a set of problems computer science is currently struggling with — provenance, structured disagreement, revision without amnesia, and how a knowledge system keeps its integrity while being continuously rewritten — and it solved them under extraordinarily hostile conditions. That is worth studying with the actual mathematics rather than by analogy.

What this site is

A map of the convergence, and the evidence for it. The Intersections page is the breadth view: ten places where the two fields genuinely meet, each with its source material, its computational mirror, and an honest note about what exists versus what is merely suggestive. From there the site goes deep in a few directions — the formal logic, the published mathematics that makes it checkable, the knowledge graph, the epistemics of preserved disagreement, and a worked case study in grounded AI.

What it is not: a halakhic resource, a ruling authority, or a place to get a psak. Nothing here is a legal decision, and the AI systems described are personas over retrieved text, not sages.

How it works

Static HTML. No framework, no build step, no client-side JavaScript required to read a page. One stylesheet is the single source of truth for the design system. This is a deliberate choice about durability: a page that needs a toolchain to render is a page that stops rendering.

Every number is traced. Figures published here — graph counts, coverage percentages, eval scores, site totals — come from a claims ledger that records, for each claim, the exact query that produced it, the result, and the date it was verified. Numbers are re-run before publication rather than remembered. Where the underlying data has known problems, the ledger records those too, and the pages say so out loud: the transmission graph contains chronological errors, the relationship schema is extraction sprawl rather than designed ontology, and the eval battery is small. A site arguing that the tradition's great contribution is provenance would be embarrassing if it did not carry its own.

Failures are published. Where a system does not pass its own tests, the page says which test and how it fails. The AI sages case study includes both a documented model failure and the analysis proving an earlier version of the evaluation was itself invalid.

Sources and licensing

Original writing on this site is available under CC BY 4.0 — reuse it, adapt it, credit it. That does not extend to the underlying texts, which carry their own terms.

Torah text data derives from Sefaria under CC BY. Every teaching in the graph stores its license alongside its reference, so the terms travel with the text rather than being asserted once and forgotten. Cited academic work belongs to its authors and publishers; this site links and quotes, it does not redistribute.

Contact

Corrections are especially welcome, and a correction with a citation attached will be acted on quickly. If something here is wrong, or a number does not reproduce, say so.