# Why does Naven’s AI newsroom wait for five publishers before it writes a briefing?

> Because a summary written from one or two reports repeats whatever they got wrong. A scheduled briefing publishes only once at least five independent publishers have filed new reports, leads with a story at least two of them reported, and links every source. Breaking news can publish sooner.

- Page: https://navenapp.com/newsroom/source-floor
- Last updated: 2026-10-21
- Published by: Naven, a free messaging app that runs in a web browser (https://navenapp.com/about)

## The rule, in full

- **Five publishers first.** A scheduled briefing is written only once at least five _independent_ publishers have filed new reports since the last one.
- **The lead is corroborated.** The story a briefing leads with must have been reported by at least two of those publishers.
- **Every source is linked.** Each publisher is listed under the briefing, with a link to what it wrote.
- **Breaking news is the exception.** A story that reads as breaking can publish sooner, on the next five-minute check, before five publishers have filed.

Publishers are counted as _distinct_ names. Five reports from one outlet count once, so a busy wire service cannot satisfy the rule on its own.

Live, as this page is served: briefings published so far, of them in the last 30 days, across news channels.

![The Channels screen in Naven on a phone, listing AI news channels to join](https://navenapp.com/shots/phone-channels.webp)  
What it looks like in Naven: a screenshot of the app.

## The numbers, as of 2026-10-15

| Scheduled briefings still stored | 1,130 |
| --- | --- |
| … citing five or more distinct publishers | 1,122 (99.3%) |
| … stored with no source list at all | 8 |
| Average distinct publishers per scheduled briefing | 6.21 |
| Most publishers behind one briefing | 8 |
| Breaking-news briefings | 78, of which 14 cited fewer than five |
| Briefings citing the same publisher twice | 0 |
| Channels running the rule | 11 |

Counted across every briefing still stored, published between 29 August 2026 and 15 October 2026. The eight scheduled briefings with no source list are a fault in how they were stored, not an exception to the rule, and they are listed here rather than left out of the count.

## A correction

Until 14 October 2026 this page said a briefing waited until five publishers had reported _the same story_. That was never what the newsroom did, and the same sentence was repeated on the about page and the homepage. The rule above is the one the code runs. The figures were re-counted on 15 October 2026; the version first published on 30 August 2026 counted 701 briefings.

## What the rule costs

It costs speed. In a quiet hour, when fewer than five publishers have filed, there is no scheduled briefing at all. And a genuine exclusive — one outlet, first, correct — cannot be the story a briefing leads with until a second publisher has reported it. Naven is a poor choice if what you want is to be first.

That trade is deliberate. An AI newsroom with no floor will confidently summarise whatever one site published, and attach a publisher's name to it. The failure mode of being slow is that you find out later. The failure mode of being wrong is that a reader repeats it.

## How to check this rather than believe it

Every briefing on Naven lists its publishers underneath itself. Open any news channel, open any scheduled briefing, and count them. The figures above come from one query against Naven's live newsroom database, counting distinct publishers per briefing:

```
SELECT breaking,
       COUNT(*)                                   AS briefings,
       SUM(CASE WHEN publishers >= 5 THEN 1 END)  AS five_or_more,
       ROUND(AVG(publishers), 2)                  AS average
FROM (
  SELECT p.breaking,
         (SELECT COUNT(DISTINCT json_extract(v.value, '$.source'))
          FROM json_each(p.sources) v) AS publishers
  FROM channel_posts p
  WHERE p.kind = 'news' AND p.reply_to IS NULL
)
GROUP BY breaking;
```

## Why this is unusual

Most AI news products are built to summarise whatever they are given. The interesting engineering question is not "can a model write a news summary" — it can, instantly, from one source, and that is the problem. It is "what is the model not allowed to do", and a published, enforced, checkable floor is a rarer answer than it should be.

**Read next:** [What Naven screens before a message sends](https://navenapp.com/safety) · [Naven vs Telegram](https://navenapp.com/vs/telegram) · [A messenger with no algorithmic feed](https://navenapp.com/for/feed-free)

## Naven, in brief

- **Price:** Free. No ads, no paid tier
- **Phone number:** Never asked for
- **To install:** Nothing. It runs in a browser
- **End-to-end encrypted:** No. Messages are screened before they send
- **Works on:** Phones, tablets, laptops, Chromebooks

### Limits, stated plainly

- Calls are not screened; messages, pictures and voice notes are
- The screen is automated. It misses things, and is weaker outside English
- No guarantees: Naven describes what it does, and its terms promise no outcome

**Naven is a free messenger that runs in your browser.** No phone number, nothing to install, and no new password.