Text study that shows its work.

Three corpora, one discipline

Nothing in the method belongs to any one text. A body of writing qualifies when it is bounded, has a real order of authority, and gets argued about from memory more often than it gets read — which is scripture and law both. So every corpus is held to the same rules: retrieved text, sources ranked rather than averaged, and a citation you can follow to the passage itself.

The Bible

66 books, 1,189 chapters. The Hebrew and Greek rank above the 3 English translations held in full — BSB, KJV and WEB, each readable verse by verse in context — and a commentary sits below all of them, marked as commentary. A further 12 English translations are known by name but not licensed, so a reference in one can be found and its text cannot be shown: the tool says so instead of paraphrasing it.

Dansk lovgivning

29 acts and bekendtgørelser — among them barnets lov, adoptionsloven, forældreansvarsloven, forvaltningsloven, selskabsloven, ligningsloven, aktieavancebeskatningsloven, ligelønsloven, funktionærloven, aftaleloven, købeloven, markedsføringsloven, bogføringsloven, årsregnskabsloven, momsloven, ophavsretsloven and databeskyttelsesloven — cited the way they are cited: act, § and stykke.

EU-lovgivning

11 regulations and directives, in Danish and English — among them databeskyttelsesforordningen, AI-forordningen, NIS 2-direktivet, dataforordningen, forordningen om digitale tjenester and eIDAS-forordningen — with the articles, the betragtninger and the bilag each addressable on their own.

The law corpora are paired where the law itself is paired: a Danish act sits beside the EU instrument behind it, because “how does Denmark implement this?” is a question you can only ask when both sides are in the corpus. Not every theme has a second side — the rules on placing and adopting a child are Danish, with the European Convention on Human Rights above them rather than an EU instrument behind them, and the corpus holds the convention for that reason. Together they are a working slice — employment, data protection, consumer trade, bookkeeping, the duties of an online service, and the placement and adoption of a child — not a library of all Danish and EU law. Ask something further afield and the tool says the corpus does not hold it, rather than answering from somewhere else.

Three kinds of claim, and the difference matters

The text asserts it

It is on the page. Anyone reading the passage would see it.

It was inferred

It follows from the text, but the text does not say it. You can accept the premises and reject the conclusion.

It is open

The text underdetermines it. Reasonable readers land differently, and nothing in the passage settles it.

Most study tools blur these together, usually by accident — a smoothing phrase, a confident summary. That is how a reader ends up holding an interpretation and believing it is the text.

Sources ranked, not averaged

Every corpus carries its own order of authority, stored as data rather than built into the engine. In the Bible the Hebrew and Greek outrank the translations, and the translations outrank the lexicons and the commentaries. In Danish law the ladder runs grundloven, then the acts, then the bekendtgørelser, and every act is stamped with the rung it sits on. Where two sources disagree, the answer says which one governs and why — instead of quietly splitting the difference.

What it will not do

• Answer from memory when it cannot retrieve the text.
• Decide which tradition or reading is right. It reports what each one rests on.
• Close a question the text leaves open.
• Tell you what to do. On law it reports what a provision says; whether you have a case is a question for a lawyer.