TheoFonsAt the sources of the faithTheoFonsAt the sources of the faithTheoFons brings the Catholic Magisterium, Scripture and the theology they cite together in a single tool — to question them, connect them, follow their cross-references both ways. Each of the pieces below exists elsewhere on its own. Their combination, where you pass from a verse to a magisterial paragraph and back without changing tool, is what did not exist online.
A magisterial document's apparatus is not decoration: it is the trail its author left. Here every note is a navigable edge, followed in both directions — what this paragraph cites, and who cites this paragraph. A backlink is treated as first-rank evidence, because it is a relevance judgment made by an authority rather than a resemblance between words.
Reading a verse shows which magisterial paragraphs invoke it, and reading a paragraph opens the verse it rests on. The bridge works by range overlap, not string equality — a paragraph citing Mt 5, 44 is found from the pericope that contains it.
Cross-reference traditions do not agree, and one of them ignores the deuterocanonical books entirely. They are merged rather than picked between, and each edge keeps the name of the dataset it came from — so a disagreement stays visible instead of being silently arbitrated.
Scripture is divided into pericopes — units of sense, not isolated verses: a narrative, a parable, a discourse. The division is nested, a long section holding shorter passages within it; we keep the level whose grain answers a question, descending while a unit is too long to carry a single sense. Each passage is then matched to a question by its sense and not by its words, so a question reaches the text that answers it without naming a single reference.
Scripture, the Magisterium and theology do not carry the same authority, and the site never pretends otherwise — the Summa, like the Fathers, is classed as theology, not as a magisterial act. The theology present is not added at will: it is what the Magisterium cites directly — level-1 adjacency in the graph — and only insofar as an accessible source exists. These texts are held in a single apparatus, with the Summa's own structure preserved (objections, sed contra, respondeo) and its internal cross-references navigable. Reading an encyclical and following it into the article it argues from is a move you cannot make anywhere else.
The assistant composes its answer from the texts found in the corpus, not from the model's general knowledge; it cites each claim to its address, § by §. In Scripture it grounds on the pericope and not on the isolated verse: that is the level at which search by meaning works, and the unit that carries a whole sense. Where the corpus does not answer, it says so rather than filling the gap. The retrieval trace is shown: you can see which passages were used, and why they were selected.
The rules this project holds itself to. They are testable: each one is either visible in the site or it is not being kept.
Every magisterial claim carries the address of the passage it rests on. An answer that cannot cite is an answer that must refuse.
The statistics page lists 1,695 “ghosts” — documents the corpus cites but that are not yet ingested — and 3,599 references that lead nowhere. A reader is entitled to know where the map stops.
No number on this site is written by hand. All are recomputed at build time from the artefacts the reader itself consumes, so a page cannot quietly drift away from the data.
Every document is compared against its own edition in the other language: same paragraphs, same notes, same call-outs? The alignment is produced by a data-mining algorithm and checked semi-automatically — this is what separates a failure of our extraction from a real divergence between the Holy See's own editions.
The Holy See publishes some French pages WITHOUT their apparatus of notes. For 15 documents — Centesimus annus, Veritatis splendor, Evangelium vitae, Ut unum sint among others — the notes you read in French are mounted from the English edition, call by call. This is not invention: every note exists in an official edition of the same text, and without the mounting the apparatus would simply be missing. But it is a reconstruction, and it should be stated rather than undergone. Where the two editions genuinely differ — Inter mirifica carries in French a note from the Council Fathers absent in English, and places in the body citations the English sends to footnotes — no alignment is forced.
Retrieval quality is measured, not assumed: the cross-references the authors themselves drew serve as the yardstick — does a search by meaning recover the passages they linked? Where a measurement contradicted an early idea, the idea gave way, not the measurement.
Every text and dataset is attributed, including those that impose no such obligation. Where a source declares no licence, none is claimed on its behalf.
The whole corpus is served as static pages at stable addresses: one paragraph, one address — no session, no database queried while you read, no page that changes according to who opens it. What you cite today in a paper, a class or a homily still resolves in ten years, and it is cited the way a book is cited: by paragraph, not by whatever screen you happened to have in front of you. The dead link is the disease of citation online, and the Holy See has moved its own pages more than once. Static files can also be archived elsewhere than here — and what is archived outlives the site that produced it.
Stated here rather than discovered by you.