The feed
Reading the feed, what every signal on a story card means, and how to filter down to what matters.
The feed is the first item in the sidebar and where everyone lands after signing in, whatever their role. Editorial and Monitoring are role-gated; the feed is the one story surface the whole desk shares. It lists the events your selected sources are covering, one card per event, newest change first.
The header reads Feed, with the line "Scan source stories, see whether an article exists yet, and open the right one fast." That is the job of this screen. It is a triage list, not a publishing queue.
Two things about it are worth knowing before you read a single card:
- A card is a source story plus the state of your article for it. ErmisAI groups articles from several outlets into one event and writes a synthesised version behind the scenes, but nothing is written into your workspace as a finished article. Every story arrives with placeholder text until someone opens it and runs compose.
- Nothing here is published. The pipeline never publishes. Publishing is a human approval in editorial review.
What one card tells you
On a wide screen the list has two column headers: Story and article state on the left, Pipeline on the right. Each card follows the same layout.
| On the card | What it is |
|---|---|
| Signal pill | A four-band descriptor for the confidence score: Verified signal, Strong signal, Developing signal, or Low signal |
{n}/100 | The confidence score itself. It never goes below 40 or above 99 |
| Article status line | Whether an article exists yet, and what state it is in |
| Source icon and headline | The favicon and headline of the primary outlet in the cluster |
| Supporting line | One of three fixed lines about your saved article |
| Category badge | The event's category, in your interface language |
{n} sources | How many distinct outlets are in the cluster |
breaking pill | Present when the cluster is flagged breaking |
| Date and time | When your newsroom's record of this story last changed |
| Owner name | The person the story is assigned to, if anyone |
| Queue badge | In my queue or Pinned in my queue |
| Pipeline chip | Where the story sits in the production chain |
| Editorial badge | Draft, In review, Rejected, or Published |
| Priority badge | High or Urgent. Standard priority shows no badge |
| Buttons | Add to queue / Remove from queue, and Open |
Clicking the headline and clicking Open both take you to the story workspace.
The headline is the source's, not yours
The card always shows the headline the original outlet published, even after someone has written and saved an article with a completely different headline. Your headline appears on the line underneath, prefixed Saved article:.
The article status line
This is the fastest thing to read on the card, because it tells you whether there is any work in the story yet.
| Line | Meaning |
|---|---|
| Needs first article in English (or your article language) | No article exists. The story is placeholder text and nobody has composed yet |
| Article ready | An article is saved and sitting in Draft |
| Article in review | An article is saved and has been submitted for review |
| Published article | An article was approved in editorial review |
Directly below the headline, the supporting line repeats this in longer form: The source story is available, but no article in {language} has been saved yet. for an uncomposed story, Saved article: plus your headline once one exists, or "Saved article is ready for editing." when your headline and the source headline match.
Pipeline chip versus editorial badge
These are two independent status systems and they routinely disagree. The pipeline chip describes production; the editorial badge describes the editorial decision.
In practice the pipeline chip reads Ready to compose on anything nobody has written yet, then Review or Published. The chip also carries a small dot labelled live or stable — English in every interface language.
A card reading Ready to compose next to In review is normal and not a bug. It means the pipeline routed the story for review because of its confidence or subject matter, but there is still no article, so it will not appear on the editorial board. Compose it first.
Full detail on both systems is in Pipeline phases and editorial states.
Breaking and priority
The breaking pill fires on any source article published in the last 90 minutes, and on any headline containing the words "breaking" or "live". It is a recency signal much more often than an editorial judgement, so expect a lot of them on a busy morning.
Priority starts at High for anything flagged breaking or sensitive, and Standard for everything else. Nothing sets Urgent automatically — that only appears when someone sets it by hand in the story workspace.
The timestamp
The date and time on the card is when your newsroom's record of the story last changed — a compose, an edit, a review decision, or a fresh source joining the cluster. It is not the source's publication time, and it is written out in full rather than as "2 hours ago".
The signal descriptors, the breaking pill, and the live / stable dot are English in all ten interface languages. Everything else on the card is translated.
The three counters
Above the filters sit three numbers. They appear once the first load completes.
| Counter | What it counts |
|---|---|
| Stories | Cards matching your current filters — not the size of your newsroom |
| Delivered | Approved articles released into the feed today, counted against your plan's daily limit |
| Queued | Approved articles held back because you hit that limit. Turns amber above zero |
Why an approved story is not on screen
When Queued is above zero, an amber banner appears reading {n} queued for delivery (daily cap resets at midnight UTC), with a Manage subscription link into billing settings.
Those stories are not greyed out or moved to the bottom. They are removed from the list entirely. The counter and the banner are the only evidence they exist.
Only fully composed and approved articles consume the daily limit. Anything still in Draft, In review, or Rejected is always visible and never counted, so the limit never blocks you from working — only from releasing.
| Plan | Stories delivered per day |
|---|---|
| Free | 20 |
| Plus | 100 |
| Pro | 500 |
| Business Plus | 100 |
| Business Pro | 500 |
| Enterprise | No limit |
The count resets at midnight UTC, not in your local timezone. If a subscription goes past due or is cancelled, the whole newsroom reverts to Free limits until it is settled. See Plan and quota reference.
Filters and search
One search box, four dropdowns, and one checkbox.
| Control | Options |
|---|---|
| Search box | Placeholder Search headlines… |
| Category | All categories, then the standard category list plus any custom categories your newsroom has added |
| Pipeline | All pipeline, Collecting, Clustering, Ready to compose, Review, Published |
| Editorial | All editorial, Draft, In review, Rejected, Published |
| Priority | All priority, Standard, High, Urgent |
| Checkbox | Breaking only |
A small badge shows the count of active filters, and a Reset button clears them all. Your filters are written into the page address, so you can bookmark a filtered view or send it to a colleague.
Three things about filtering are easy to trip over.
Filtering by Collecting or Clustering always returns nothing. Both are listed in the dropdown, but no card ever carries either chip. Use Ready to compose to find work that has not been written yet.
Search looks at one field only. It is a plain case-insensitive substring match on a headline — no summaries, no article bodies, no phrase handling. And it changes target once an article is saved: for an uncomposed story it matches the headline printed on the card, but once someone composes, it matches your saved headline while the card carries on showing the source headline. If you are hunting for words you can see on screen and getting nothing, that is why. Search the words you wrote, or drop the search and use the dropdowns.
Breaking only is close to a recency filter, for the reason described above.
Sorting and scrolling
There is no sort control. The order is always most-recently-changed first, and confidence plays no part in it.
The list loads twenty cards at a time and pulls in more as you scroll. When you reach the end it says No more stories.
My queue
The personal queue is your own shortlist. It is visible only to you, and no one else in the newsroom can see or change it.
The personal queue is switched on by default, but it is a platform-level feature that can be turned off. When it is off, Select multiple, View queue, and the per-card queue buttons do not appear at all.
- Add to queue on any card adds it. The button flips to Remove from queue, and the card gains an In my queue badge.
- Select multiple turns on checkboxes, and
Queue selected ({count})adds them in one go. Exit select leaves selection mode. - View queue opens a drawer titled My queue, described as "Stories you've lined up in your personal editing queue." An empty one reads "Your queue is empty. Add stories from the feed to line them up here."
Each row in the drawer shows a position number, the headline, the pipeline label, and buttons Pin / Unpin and Remove. Pinned rows carry a Pinned badge and sort to the top; everything else stays in the order you added it. There is no drag-to-reorder — pinning is the only ordering control, and the position numbers are recalculated each time you open the drawer.
The queue drawer shows your saved article's headline once one exists, so a row here can read differently from the same story's card on the feed.
Keeping the feed fresh
ErmisAI re-reads your sources in the background on a ten-minute cycle, so the feed keeps filling whether or not anyone has the page open. Newsrooms are refreshed one after another within a time budget, and the starting point rotates each cycle so nobody is permanently at the back of the line. A busy cycle may not reach every newsroom, so treat the schedule as approximate.
The feed also updates itself while you are looking at it. When a colleague composes, submits, or approves something, your list refreshes on its own without a reload.
Refresh sources forces your newsroom's turn immediately. The button reads Refreshing… while it works.
Clean feed
Clean feed is destructive and only appears for owners and admins. In a team newsroom it is shown only to organisation admins; on a solo desk it is always shown.
It opens a confirmation dialog titled Clean feed asking "Clear the current local feed state and placeholder articles from this workspace?", with Cancel and Clean feed. On success you get a Feed state cleared toast.
What it actually does:
- It hides, it does not delete. Existing drafts are marked as hidden behind a timestamp. Nothing is removed from your newsroom.
- It clears everyone's personal queue, not just yours.
- It does not refund the day's Delivered count. Cleaning cannot buy you more releases.
The feed does not stay clean. Cleaning immediately triggers a forced refresh, and a forced refresh ignores the hide marker — so the drafts you just cleared reappear straight away. The next ordinary page load hides them again. Any use of Refresh sources later on will surface them once more.
Warming up, empty, and broken
A brand-new newsroom builds its first feed on demand, and that build is slow because it fetches, clusters, and writes in one pass. While it runs you get an animation and a rotating line — "Warming up the newsroom…", "Scanning the wires…", "Sifting through the sources…" and similar — plus the note "Fetching, clustering, and writing the newsroom articles can take up to a minute. Please be patient - and stay hydrated."
The empty state always reads "No stories match your current filters." even when you have no filters set at all. There is no separate copy for a newsroom that simply has nothing yet, so on a brand-new desk it usually means "nothing here", not "your filters are wrong". Check the filter badge before you go looking for a fault.
A red strip means the load failed. The most common cause is a first build that took too long. Press Refresh sources and give it another go; more failure messages and what they mean are in Troubleshooting.
Your first story, end to end
From a feed card to an approved article, one worked example.
Composing and refining a story
What happens after you press Open.
Pipeline phases and editorial states
The two status systems on every card.
Confidence, sources and conflicts
What the signal score is built from, and what it does not prove.
Getting approved stories out
The daily cap and where approved articles can go.
Choosing and adding sources
What ends up in the feed in the first place.
