How activ works, and how you can trust it
Trust you can check — not trust you're asked to give.

In a world of deepfakes, fabricated stories, and machine-made slop, the honest question about anything you read is: is this made up? Here is our answer, in a form you can verify rather than take on faith.

activ does not create facts. It curates and explains verifiable facts reported by named, established news organisations. Every claim you read traces to a named source you can open and check for yourself. We are not the authority — the source is, and we always show you which one.

We never fabricate. Everything traces to a named source.

Every article names its source and carries a numbered reference you can follow to the original reporting. We don't originate claims; we summarise what named sources report and point you straight back to them. If you ever want to check a story, the source is right there on the card — that is the whole point.

The sources we curate from

We don't ask you to take our roster on faith — you see the exact source on every single story. As a roster, the daily news is drawn from established news organisations, and ACTIV NEXT from primary science and research publishers. Among them:

The daily news
ReutersAssociated PressBBCThe GuardianAl JazeeraThe HinduDeutsche WelleFrance 24The Wall Street JournalThe Telegraph
ACTIV NEXT (science & discovery)
NASAESAMIT Technology ReviewQuanta MagazineIEEE SpectrumPhys.orgScienceDailyThe ConversationMongabay

A representative list, not the whole of it — the roster changes as we add and retire feeds. What never changes: whatever ran a story, its name is on the card, and you can go read the original.

On the technology we use

We use technology to help sort and summarise a large volume of news. Humans set the editorial standard. It never invents facts — it summarises what named sources report, and for our GLORY column it narrates a verified source record, checked and approved by an editor before it is ever published. We built activ specifically to prevent the fabrication that makes so much online content untrustworthy: the tool is a servant of the sources, never a substitute for them. It is a safeguarded tool — not the engine, and never a licence to make things up.

We flag uncertainty honestly

Preliminary is never presented as settled. Where a source flags a finding as early or preliminary, we carry that flag through to you — for example, an “early research” marker on a discovery that a single study suggests but that isn't yet confirmed. Showing what is not yet certain is part of telling the truth, not a footnote to it.

An editor stands behind it

Our GLORY pieces are held back as drafts and publish only after an editor has read and approved each one — nothing in GLORY reaches you unreviewed. Across the daily edition, editors oversee what ships and can correct or override any card at any time. And when we get something wrong, we don't quietly delete it — see below.

When we're wrong, we say so — in public

You can report an error on any story. An editor reviews it, and genuine corrections are published, dated, in our public corrections ledger — with the story card itself marked “corrected” and linked to the entry. Correcting yourself in public is something no fabrication operation will ever do; it is the plainest proof that a real newsroom stands behind the work.

The rest of this page is the detail: the specific rules our compiler must pass before an edition ships. If a rule cannot be satisfied, cards get dropped or the lead gets promoted — an edition is never shipped in violation of them. Watchdog observation is not enough; these are hard assertions.

Source discipline

Per-source cap: at most 3 stories per edition
No single publisher can dominate a 20-story edition. If Reuters produces four wire pieces you'd otherwise see, we ship the top three and drop the fourth. Applied at selection time, not as a watchdog warning after the fact.
Region quota (World): at most 3 stories per country
The World section has no more than three items from any single country per edition. Keeps a busy news day from any one place from monopolising a global read.
Sports cap in World: at most 2 stories
A World Cup weekend doesn't take over an edition otherwise dedicated to what's shaping the day. Editorial can flag one moment historic to override.

The lead slot

Crime never leads any section
Local doesn't open on the day's worst crime. Country doesn't either. Nor World. If the top-ranked story is a crime, the compiler promotes the first eligible follower to lead.
Sports never leads World
A cup final can be a story on the day it happens, but it doesn't become the day's most consequential development.

Meaning preservation

Source headlines are never softened
If a source uses the word "devastating," our AI is not permitted to rewrite it to "significant." A maintained list of load-bearing words (killed, banned, resigned, record, invaded, evacuated…) is guarded on every card. On a violation, we revert to the source's exact headline.
Card atomicity
Every card's headline, brief, and source footnote come from exactly one article. The compiler refuses to ship a card whose fields don't all trace back to the same story ID, and the editor is alerted when a merge bug produces one.
Summaries must inform
A summary that reads "My favorite movie is X" when the story is a policy speech isn't a summary. Decorative openings are detected and swapped for the first informative sentence of the body.

Category sanity

Wildlife stories are not crime
A rogue elephant story tagged CRIME is demoted to General. So is a business deal tagged GEOPOLITICS, and a film re-release tagged SCIENCE. Low-confidence tags fall back to a neutral label rather than a wrong specific one.

Fact-checks as first-class stories

When a story turns out to have been misleading — ours or a source's — the fact-check that clarifies it belongs on the card, not buried. Cards with published corrections carry a small "corrected" mark linking directly to the entry in the public ledger.