TheoFonsAt the sources of the faithTheoFonsAt the sources of the faithThis site does one thing, and does it in a way that is worth explaining: it turns the footnotes of Church documents into paths you can walk. Nothing here is a summary of the texts — the texts are given whole, with the trail their own authors left between them made walkable in both directions.
Five tools over one corpus. They are not separate products: each of them ends up at the same passages, at the same addresses.
Every document is taken from its official publication — the Holy See's own pages, and the published cross-reference datasets for Scripture. Nothing is fetched while you read: pages are archived once, parsed, and served from that archive. So what you read cannot change under you because a remote site was reorganised, and an address you cite today still resolves tomorrow.
The corpus currently holds 96 magisterial and theological documents and 73 biblical books — 17,037 numbered passages in all. That is a fraction of the Magisterium, and the site says so rather than implying completeness.
This is the part nobody sees, and it is most of the work. The Holy See's pages were written by hand over some twenty-five years, and they were made for eyes, not for a program: what a reader takes in without thinking — a rule separating the text from its notes, a heading that is not a paragraph, a number badly printed — has to be reconstructed rule by rule, document by document.
On top of that come defects a reader really does suffer: the same document gives two differently shaped pages depending on the language, and some French pages appear without their apparatus of notes. So each document is confronted with its own edition in the other language — same paragraphs, same notes, same call-outs? Where the French carries no notes, the apparatus is mounted from the English call by call: 17 documents are in that case, and each says so.
Why the trouble: a paragraph number is an address. Off by one, and every citation pointing at it lands on the wrong text — without any visible sign.
This is the central move, and it is invisible when it works. A document's apparatus is full of lines like “Cf. Second Vatican Council, Pastoral Constitution Gaudium et spes, 22”. To a reader that is a pointer. To a website it is usually just text — you read it, and you go and look the passage up yourself, somewhere else.
Here each of those lines is read, identified, and turned into a link to the exact paragraph it names. 9,994 notes have been processed this way, yielding 45,629 working references — of which 29,135 point into Scripture, reaching 10,356 distinct verses.
The result is that an author's own citations become navigable. You are not following a resemblance a machine noticed between two texts; you are following a link a pope or a council deliberately made.
Every passage in the corpus is read in two directions, and the difference matters more than any other idea on this site.
The texts this passage draws on: what it quotes, what it rests upon, what came before it. Follow this direction and you go back towards the sources — often to Scripture, or to an earlier council.
The later texts that cite this passage. Follow this direction and you go forward through history: you see what a paragraph became once it had been written, who took it up, and what they made of it.
In the technical documentation these are called out and in. Those names describe which way the arrow points in the underlying graph, not which way the influence runs — and it trips people up, because a paragraph's outgoing references are exactly what comes into it as nourishment. If you ever meet the words, read out as “what feeds it” and in as “what it produced”. On the site itself we avoid them.
Not chosen by hand: at each build the site looks for the paragraph that best shows both directions at once, and shows that one. Here it is today.
The truth is that only in the mystery of the incarnate Word does the mystery of man take on light. For Adam, the first man, was a figure of Him Who was to come, namely Christ the Lord. Christ, the fin…
A paragraph of an encyclical is written ls/217 — document, then paragraph number. A verse is jn/3/16 — book, chapter, verse. Every page of this site is one of those addresses, and the address is in the URL.
That is not a technical detail: it is what makes a citation hold. You can send someone a link to the exact paragraph you mean, and it will still be that paragraph in a year. It is also why the corpus can be linked to itself at all — a reference has to point somewhere, and here everywhere is somewhere.
It matters to be precise about this, because a reader who is not told will assume it is everywhere.
Turning “Cf. Gaudium et spes, 22” into a link to that paragraph is done by explicit rules, not by a model. This is deliberate. A model asked to guess a reference would guess wrong some of the time, and a wrong link is worse than a missing one: it looks exactly like a right one. The graph, the addresses, the paragraph numbering — none of it is generated.
The rule underneath: AI selects and phrases. It never establishes a link, never a number, never a text.
Documents cite documents the corpus does not hold. Rather than drop those references, the site keeps them as named markers — you see that the passage cites a work, and what it is, even though there is nothing yet to open. There are 751 such names. The most cited are:
Along with 2,496 references that lead nowhere at all — most often to a text with no accessible edition. Both figures are published rather than hidden: a reader is entitled to know where the map stops.
The project, in figuresSources & creditsCorpus statisticsPublic API