How every number here was counted
This is the boring page, and it is here on purpose. A site whose subject is b'shem omro - report a teaching in the name of the one who said it - does not get to publish a figure it cannot show the working for. So every number that appears anywhere on jew.tech has a row below: what it counts, how it was counted, and the day the count was run.
Two standing rules govern the list. A figure is a snapshot of a live system on a particular day, so it carries that day. And a figure that has not been re-run does not get rounded up, moved forward, or quietly restated - it either gets measured again or it comes off the page.
The knowledge graph
One Neo4j database holding the classical Jewish canon. Nodes carry several labels at once, so every count below is label-scoped rather than derived from a first-label lookup, which would undercount badly.
| What | How | Count | Run |
|---|---|---|---|
| Total nodes | MATCH (n) RETURN count(n) | 157,143 | 2026-08-02 |
| Rabbinic teachings | MATCH (n:VaultTeaching) RETURN count(n) | 17,147 | 2026-08-02 |
| …carrying a citable reference | count(n.ref) over the same set | 17,147 (all) | 2026-08-02 |
| …carrying Hebrew source text | count(n.he) over the same set | 17,147 (all) | 2026-08-02 |
| …carrying a vector embedding | 768-dimension embedding present | 17,147 (all) | 2026-08-02 |
| Verses | MATCH (n:Verse) RETURN count(n) | 33,323 | 2026-08-02 |
| Named rabbis | MATCH (n:Rabbi) RETURN count(n) | 13,935 | 2026-08-02 |
| Teaching-to-verse citations | MATCH ()-[r:CITES_VERSE]->() RETURN count(r) | 15,910 | 2026-08-02 |
The transmission chain
The chain panel publishes three figures. All three are scoped to rabbi-to-rabbi edges only. The larger raw edge count that runs across every node type is not the same measurement and is not the one quoted.
| What | How | Count | Run |
|---|---|---|---|
| Teacher-to-student links | Rabbi-to-Rabbi transmission edges | 2,504 | 2026-08-02 |
| Sages in the chain | Distinct rabbis participating in those edges | 2,185 | 2026-08-02 |
| Largest connected component | Computed over all 2,504 edges | 1,163 sages (53.2%) | 2026-08-02 |
One honest limitation, since it matters more than the counts do: this is an extracted transmission network, not a historically validated one. The extraction contains chronological errors - it will happily assert that Moses taught Abraham - and the correct description of it is "what the sources say about who learned from whom," never "the verified lineage of the Jewish people." The site describes it that way and so should anyone quoting it.
The grounding rule
The rule the Torah & AI panel is built around, and how it is enforced.
| What | How | Result | Run |
|---|---|---|---|
| Only cited material is retrievable | Answers are drawn from teaching records that carry a citable reference; uncited material is not indexed for retrieval | Cited-anchors-only policy | standing design decision |
| Teachings carrying a citable reference | MATCH (n:VaultTeaching) WHERE n.ref IS NOT NULL RETURN count(n) | 17,147 (100%) | 2026-08-02 |
| Teachings carrying Hebrew source text | MATCH (n:VaultTeaching) WHERE n.he IS NOT NULL RETURN count(n) | 17,147 (100%) | 2026-08-02 |
| Every voice checked before serving | Citation checking offline, plus a live conversational check | Required to pass before release | logged per run |
The point that generalizes: a reference is not the same thing as the text it points at. A collection can name a source many times and hold none of its actual words, and nothing in the collection's own statistics will tell you the difference - you have to go and ask for the text. That is why every teaching here is required to carry both a reference and its source words, and why the coverage figures above are reported as percentages rather than totals.
The library and the machines
These are the inventory figures. They belong here rather than in the story, because the size of a collection is not the interesting thing about it.
| What | How | Count | Run |
|---|---|---|---|
| Files catalogued | SELECT count(*) FROM catalog.files | 34,457,680 | 2026-08-02 |
| Distinct works | SELECT count(*) FROM provenance.works | 43,174 | 2026-08-02 |
| Largest category: Torah & seforim | Category count over the same table | 12,203 works | 2026-08-03 |
| Corpus licensing | License field across all teachings | CC BY-NC 70.4% · CC BY 26.7% · unlabelled 2.9% | 2026-08-03 |
The site directory
The directory on the sites panel lists 47 properties: 46 destinations plus this hub. The count is the sum of the four directory groups: 19 Torah and Jewish learning, 16 knowledge and reasoning, 2 personal and hub, 10 engineering and business. It was set by a sweep on 2026-08-06. Every listed domain was rendered in a real browser with JavaScript executed and its text extracted, and an entry was kept only where substantive content appeared. Nine entries came off in that sweep. They are not listed as live anywhere.
What is deliberately not on this page
Anything that would need a client's permission, a contract term, or someone else's name. Anything private. And any figure I have not personally re-run, however good it would look.