The graph

160 nodes, 427 edges, twenty-one op-eds. Start with the articles and their authors, and switch the other blocks on one at a time. Click a node to open it, double-click to centre on it, shift-click two to read the path between them as a sentence — in English or in Portuguese. Machine surfaces: graph.json · ontology.json · triples.nt.

Beta, agent-produced. The fetching, extraction, classification and drafting here are done by software agents against frozen bytes. Nothing on this page is a legal opinion, and there has been no legal review. If you are deciding whether you may republish one of these pieces, read the terms on the piece itself and ask the publisher — that is the point we are making.

We hold all twenty-one pieces and republish none of them. Every article is linked to its original on worldnewsday.org. What is published here is the description: who wrote it, under what terms, what it links to, what words it contains. The prose belongs to the people who wrote it, and the gate on this section fails the build if twelve consecutive words of any piece appear on any page we generate.

Every number walks back to bytes we hold. Each page was fetched once, frozen to a dated snapshot in this repository and hashed with SHA-256. The counts are re-derived from those bytes by a second program before anything ships. The register → · The method →

Why a graph at all. Twenty-one articles in a list are twenty-one articles. The same twenty-one with their authors, the organisations those authors name, the themes their prose contains, the sources they point at and the terms they are published under is a thing you can ask questions of: which organisations have someone writing about AI, which pieces cite nothing, which themes never occur together. None of that is in the publisher’s own data, because the publisher’s own data is a single category called Story.

The instrument is borrowed, on purpose. This is the same engine and control panel as /portugal/ and graphs.sgit.ai — packs instead of levels, otherwise the same thing. A reader who has used one should not have to learn another.

Start here

Each button sets the packs and, where it makes sense, runs a query.

The corpus: articles, authors, organisations What they are about What they cite The terms every piece is published under The evidence underneath Organisations whose author wrote about AI Everything

Query the paths

Tracing answers how do these two connect? This answers show me every route shaped like this: a start filter, then up to three edge-and-node steps, walked breadth-first with a bound — and it says when it hit the bound, because a query that silently truncates is worse than one that refuses.

The types

Type	What a node of this type is	Count
Article · Artigo	One commissioned op-ed, as published on worldnewsday.org and frozen here. The prose is linked, never reproduced.	21
Author · Autor	A named person in an article's byline, in their professional capacity. Name and the role line they gave; nothing else, and never a contact detail.	23
Organisation · Organização	An organisation an author's own role line names. Derived from that line, never typed by hand.	27
Theme · Tema	A theme from the published lexicon whose pattern matched the article's prose. It says the piece contains those words and nothing more.	20
Cited source · FonteCitada	A page an article links out to. The evidence the piece offers its reader, counted as the piece offers it.	23
Licence · Licença	A statement of republication terms. Carries the form it takes: prose, a named public licence, or a machine-readable declaration.	2
Snapshot · Captura	A dated capture of the source site: every page fetched, frozen to bytes in this repository, and hashed.	1
Frozen file · Fonte	One frozen file with its SHA-256. The thing a claim walks back to.	43

The vocabulary

Every edge is a verb with a distinct, named inverse, and every verb carries its Portuguese. Walked forwards a path uses the verb; walked backwards it uses the inverse; either way it is a sentence. Inherited from graphs.sgit.ai’s edge grammar — Every edge a verb with a distinct named inverse; symmetric edges banned; a path must read as a sentence., and shared with /portugal/ so the two sections can be read by the same reader and, eventually, joined.

Verb	Inverse	Domain → range	Reads as	In use
`written_by` escrito_por	`wrote` escreveu	Article → Author	A is written by B	24
`affiliated_to` afiliado_a	`affiliates` afilia	Author → Organisation	A is affiliated to B	27
`touches` aborda	`touched_by` abordado_por	Article → Theme	A touches B	211
`cites` cita	`cited_by` citado_por	Article → CitedSource	A cites B	38
`published_under` publicado_sob	`governs` rege	Article → Licence	A is published under B	21
`commissioned_by` encomendado_por	`commissioned` encomendou	Article → Organisation	A was commissioned by B	21
`attested_by` atestado_por	`attests` atesta	Article → Source	A is attested by B	42
`captured_in` capturado_em	`captures` captura	Source → Snapshot	A was captured in B	43

Banned

A verb that is banned is banned in the code, not in a style guide: the gate fails the build if one appears. agrees_with is this section’s own addition, and the reason is on the right.

Verb	Why
`related_to`	Symmetric, so it would be its own inverse, and it says nothing a reader could walk.
`mentions`	Every edge in a graph built from text mentions something. It is the absence of a verb.
`associated_with`	The verb reached for when the relationship is unknown. Find out, or leave the edge out.
`agrees_with`	Two pieces touching the same theme is not agreement. Asserting agreement would be this section putting words in an author's mouth.

The one formula

The only classification this graph makes rather than reads. Each edge carries the words the pattern matched, and the gate re-runs every pattern on the frozen bytes in both directions. The 20 patterns are published in full.

For an agent

Fetch data/graph.json and data/ontology.json rather than parsing this page; data/triples.nt is the same graph as N-Triples with owl:inverseOf on every verb and @en/@pt labels. Every node names the frozen file it came from. This page also publishes an API: after the tool:ready event, window.__graph exposes read methods (get_graph, get_nodes, get_node, get_edges, get_ontology, search) and view methods (set_packs, explore, select_node, set_layout, set_language, fit_graph, graph_snapshot, reset_view). Nothing writes.