Building news that can explain what changed while you were away

Come back to a running story after a week away and every news product gives you the same thing: the latest article. What you actually needed was the difference between the last time you looked and now — what moved, what turned out to be wrong, and which parts are still only one source deep.

ITISYOU News is an attempt to build for that question instead. It is built and running but has not launched, and I will come back to what that means at the end.

Why the delta is hard

A news article is written as though you are arriving fresh. It re-establishes context, restates what is known, and folds any correction quietly into the current text. That is a reasonable convention for a reader who has just walked in. For a reader who was here on Tuesday, almost all of it is noise wrapped around the one paragraph that is new.

Worse, the thing that changed is often not in the text at all. A claim that was attributed to a single anonymous source on Tuesday might be confirmed on paper by Friday. Nothing in the prose necessarily changes. The confidence you should have in it changes completely.

So the underlying model has to record more than the words. It has to record the state of the reporting, separately, over time.

Signals that stay separate

The design decision I keep defending is that an event carries several independent labels instead of one score. In the reader you can see them: whether something is developing, whether it has been editor reviewed, whether it is emerging, whether it currently rests on a single source.

The temptation to collapse those into one number is strong, because one number sorts nicely and fits in a badge. It is also the mistake. Editorial review and independent corroboration are genuinely different signals:

• Editor reviewed means a person checked the piece against its sources. It says nothing about how many independent sources exist.
• Corroboration means several independent outlets or documents support the claim. It says nothing about whether anyone here checked the writing.

A story can be carefully reviewed and single-sourced. A story can be widely corroborated and sloppily written up. Averaging those into “78% confident” destroys exactly the information a careful reader wants, and does it while looking more rigorous than either label alone.

Revisions, and why corrections get their own log

Every event links to its evidence, its current status and its full revision history. That last one is doing quiet work. If a claim was strengthened, weakened or withdrawn, that movement is the story for a returning reader, and it is exactly what gets erased when an article is edited in place.

Corrections have their own log, linked from the front page — including when there is nothing in it. The log currently says that nothing published so far has required a correction, which is a less impressive statement than it sounds given how little has been published, and I would rather it said that than hid the log until there was something in it.

A corrections log that only appears once it has entries is a marketing surface. A corrections log that is always there, sometimes empty, is a commitment made in advance.

What this shares with the rest of the network

The resemblance to DEV.ITISYOU is not accidental, and I noticed it after building both rather than before. Dev attaches evidence to a specific revision so it cannot silently transfer to edited text. News attaches status to a specific version of an event so a reader can see how support changed. Both are the same instinct: the current state of a document is not the whole truth about it, and the history is not an audit trail hidden in a database — it is the part the reader needs.

That instinct is also the seed of an unnamed research question about whether posts generally could carry their own history. It is a concept, not a plan, and finding the same idea in three places makes me more suspicious of it rather than more confident.

What it is not

Being precise about the state of this one matters more than usual.

ITISYOU News has not launched. It is labelled Working everywhere it appears on this site, its public address at news.itisyou.app is still being brought online and may not answer yet, and there is no launch date — nothing in this article should be read as implying one.

Coverage is thin, deliberately. The build exists to test whether the structure holds up, not to be a news service. A model that works cleanly across a handful of events may well break under a busy week, and I have not seen it face one.

Advertising is planned for the eventual product and is not running. There is no advertising code in it. It earns nothing, costs something, and is being built in the evenings — which is the honest description of most of this network right now.