You're on version 2026-08-01. Change notes are in English.
Cleaner, smarter entity tracking
Aug 1, 2026 · 2026-08-01
The names we track behind each story are now deduplicated. The platform records the key people, places, and things mentioned in every story. That list had grown noisy — website boilerplate like "Continue reading" was being logged as an "entity," generic words like "Scientists" or "Police" slipped in, and "New York" was being mangled into just "York." Those are cleaned up.
The same real thing now has one name, not five. "Trump" and "Donald Trump" used to be counted as two different entities; "America" and "United States" as two more; a Shahed drone and its Russian name "Geran-2" as two separate weapons. They're now recognised as one each — so searching or grouping by an entity finds everything about it, not just the stories that happened to use one particular spelling.
It learns from our investigations. The correct names come partly from a curated core list and partly, automatically, from the Special Assessments we run — so entity tracking gets a little smarter with every investigation. Nothing you do changes; the story lists and Situation Room just get more accurate underneath.
Retired the Angle Desk
Jul 26, 2026 · 2026-07-26
The Angle Desk has been removed. The header no longer carries an "Angle Desk" — the surface that turned a beat's coverage into audience-tagged, ready-to-post angles is gone. As ДОZОР settled into its identity as a neutral intelligence-analyst platform (decision-support, not "what to publish"), an audience-targeting angle generator no longer fit what the platform is for, so we've taken it out rather than leave it sitting unused.
Nothing you rely on changed. The per-story "Suggested angles" you see when you expand a story are unaffected — those come from the story analysis and stay exactly as they were. The cross-outlet blind-spots / omissions read still lives in the Situation Room, where it belongs. Navigation is now Signals Desk · Situation Room · Special Assessments.
Faster pages, and they stay fast as the archive grows
Jul 25, 2026 · 2026-07-25
The dashboard, browser, and Situation Room now load quickly. The dashboard used to take around 11 seconds to show its headline brief; it now appears in well under a second. The "Top Stories" list is about 18× faster (roughly 1.7 seconds down to a tenth of one), the All Stories browser and search open in a fraction of the time they used to, and the Situation Room's board is now cached so it opens near-instantly on a repeat visit (and quicker on the first one too). The server also no longer maxes out its processor drawing these pages.
Why it was slow — and why it won't creep back. Opening any of these pages used to re-read and re-rank the *entire* story history from scratch. That was fine at a few hundred stories, but the archive has grown past 8,000 and keeps growing, so the work piled up on every click. The platform now does that ranking once, in the background, and saves it — so opening a page is a quick lookup, not a from-scratch recount. And that background work is now limited to the active window (the last several days of live stories) instead of the whole archive, so pages stay fast however large the archive gets. The background ranking pass itself is about 3× faster as a result (roughly 12 seconds down to 4).
Nothing changed about what you see. Same stories, same ranking, same filters and search — just delivered far faster. Closed, older stories still appear in the All Stories browser; when you sort by priority, live stories now sit above closed ones.
Headlines in the desk's own neutral voice
Jul 25, 2026 · 2026-07-25
Story headlines are now neutral and attributed, not the source's loaded original. A wire headline often carries the outlet's framing — "invaders wiped out," a contested casualty figure stated as fact, or one side's claim presented as settled truth. An intelligence desk shouldn't repeat that framing in its own voice. So when a story is analyzed, the platform now also writes a clean, neutral headline for it: loaded or emotive wording is stripped, and any contested claim is attributed to whoever is making it ("Russian Foreign Ministry says…", "Ukrainian officials report…") rather than asserted. That neutral headline is what you now see across the dashboard, All Stories, the Situation Room, and the daily brief — in Russian too.
The original headline is never lost. The source's real headline is kept underneath and still does all the behind-the-scenes work — grouping related coverage, search, and beat tagging all read the original, so nothing about how stories cluster or how you find them changes. Only what's *displayed* becomes neutral.
It rolls in as stories are analyzed. New and freshly-updated stories get a neutral headline on the next analysis pass; older stories keep their original headline until they're next analyzed. No extra cost — the neutral headline is produced by the same analysis step that already writes each story's summary.
A new source: Ukrinform (Ukraine's national news agency)
Jul 24, 2026 · 2026-07-24
Added Ukrinform's English service to the wire. We spotted a real gap: a story worth seeing — Poland's offer to jointly produce Patriot air-defense missiles with the US and Ukraine — slipped past the platform because it ran mostly in regional Central/East-European outlets we didn't yet cover, even though our wires carried the wider Patriot story. Ukrinform closes that regional angle, especially on the war and the defense industry.
Tested before it went on, and we read the full article. Like every source we add, it was checked against the live pipeline first — its coverage came through fresh and on-mission — and we pull the complete article body (not just the one-line feed teaser) for the stories we analyze, so the cross-outlet read stays well-grounded.
Situation Room reads now track today's coverage
Jul 24, 2026 · 2026-07-24
The Situation Room's per-beat "read" is no longer anchored to old news. For each beat (Ukraine–Russia, US–Iran, Energy/Markets, …) the longitudinal read and its event timeline are built from the beat's coverage. On a busy beat the system used to fill its reading budget with the *oldest* coverage first and run out of room before reaching today's — so the read (and its timeline) could sit a week or two behind the actual news. Now it always includes the *freshest* coverage and drops closed, week-old stories, so the read reflects where the beat is *now*.
Each beat shows how current its read is. A small "Read as of <date>" line sits under the beat name; it turns amber ("· new coverage since") when fresh coverage has landed that the current read doesn't yet reflect — so you can always tell at a glance how up-to-date a read is.
Quiet beats don't churn. A beat with no fresh coverage keeps its last read as-is instead of being needlessly re-generated.
A fresher board: today's news leads, old news steps aside
Jul 23, 2026 · 2026-07-23
Stories are now bounded events, not endless topics. Before, a story could keep pulling in new coverage for weeks — so a write-up first seeded two weeks ago would quietly swallow today's related articles, and the analysis attached to it stayed anchored to that older moment (you might open a surging Energy/Markets story and find its read reflected a picture from days earlier). Now a story spans a bounded window of a few days; when fresh coverage of the same subject shows up later, it starts a new story with its own up-to-date analysis. What you see reflects *today's* developments, not last week's snapshot.
Finished stories move off the active board. Once a story is more than about a week old it's treated as closed — it stops taking new articles and drops off the daily brief, the Situation Room, and the dashboard's Top Stories. Nothing is deleted: it stays fully searchable, remains in the All Stories browser, and any assessment that cites it is unaffected. It just no longer crowds out live coverage.
No more "magnet" stories re-floating to the top. A big, sprawling story that used to climb back to the top of the brief on a trickle of follow-up coverage now settles and steps aside as it ages. In practice the brief, Situation Room, and Top Stories now lead with stories from the last day or two.
**Situation Room beats are ordered by what's hot *now*.** The row of themes (Ukraine–Russia, US–Iran, Energy/Markets, …) sorts by *recent* coverage instead of all-time volume, so a beat surging today rises to the front even if another has more coverage banked over the whole fortnight. Each beat's story count now matches what you get when you click in.
Why it matters. This was the fix for the most-reported problem: boards showing slightly stale reads — an energy summary anchored to a two-week-old picture, or a months-old story sitting near the top because coverage kept dribbling in. The daily board should read like *today's* intelligence — now it does.
Less noise: daily "live blog" roundups filtered out
Jul 23, 2026 · 2026-07-23
Rolling "[region] news live" roundups no longer clutter the feed. Some outlets publish one ever-updating daily page (e.g. a Guardian "Australia news live" roundup) that was slipping in and clumping into odd catch-all stories. These are now turned away at collection — while genuine event trackers on our beats (an "Iran war news live" live-blog, say) are kept.
Special Assessments, now in Russian
Jul 22, 2026 · 2026-07-22
Every long-form assessment now reads in Russian too. Flip the site's EN/RU toggle in the top navigation and the whole report switches — text, charts, and all — hand-translated, not machine-translated. A second assessment has also been added to the desk (team-only and read-only, like the rest).
Special Assessments: long-form intelligence reports
Jul 19, 2026 · 2026-07-19
A new Special Assessments desk. Open it from the header (next to the Angle Desk). It holds full, long-form intelligence reports built on top of the ДОZОР corpus — the kind of deep, cited assessment that doesn't fit in the daily brief. Each report appears as a card; click through to read the whole thing, with its interactive charts, maps, and source list intact.
First report: Ukraine's strikes on Russian oil refining. A structured assessment of Ukraine's long-range drone-and-missile campaign against Russia's refineries and petroleum exports (2023–July 2026) — the target-selection pattern, the industrial and economic damage, the weapons and command behind it, and a confidence-graded read on what it has and hasn't achieved.
Team-only and read-only. Reports sit behind the platform login like everything else — there's no public link — and they're for reading, not editing. Every factual claim is either drawn from the corpus or externally cited, with one-sided (belligerent) claims clearly flagged.
Cleaner, consistent navigation
Jul 19, 2026 · 2026-07-19
One header, the same on every page. All surfaces now share a single top navigation bar, so moving around the app is predictable. The three main areas — Signals Desk, Situation Room, and Angle Desk — always sit in the same place, with the one you're on highlighted.
Signals Desk is now a menu. Click Signals Desk to jump to any of its pages — Dashboard, All Stories, All Articles, and the Briefs Archive — from anywhere in the app. Before this, some of those pages (and the Angle Desk) couldn't be reached without first going back to the dashboard.
Log out and switch language from any page. Both used to live only on the dashboard.
Less chrome noise. Removed the "LIVE" badge and the "auto-refreshes every 30s" line from the header — neither reported anything you needed to act on. The dashboard still refreshes on its own; it just no longer announces it.
The Angle Desk: turning coverage into audience-ready angles
Jul 18, 2026 · 2026-07-18
A new Angle Desk. Open it from the header (next to the Situation Room). For any beat, it turns the day's analyzed coverage into a handful of sharp, audience-tagged editorial angles — concrete takes the team can actually post — instead of leaving you to work them out from the raw stories. It's decision support for "what does the English-speaking audience want to hear," built on top of the Situation Room's read of each beat.
Pick a beat, generate angles. Choose a beat from the row at the top and hit Generate — one pass produces a few angles, each labelled with the audience lane(s) it fits and a suggested format (an X post, a clip, a chopped podcast segment, a long-form sit-down). Regenerate any time the coverage has moved. It's on-demand by design: nothing is auto-generated, so you only spend when you ask for it.
"What mainstream is skipping." Each beat leads with the cross-outlet blind spots — what the coverage is *not* saying — because that gap is where the strongest, most differentiated angles come from. Every angle sits on that context.
Grounded, and honest about it. Each angle shows what it's built on under "Grounded in" — the stories it draws from (click through to read them and their sources) and the omissions it addresses — and carries an honesty flag whenever it rests on an interpretation or a contested claim rather than an established fact. It stays an internal, human-curated drafting aid — nothing is ever auto-published.
A sharper Signals Desk, and an intelligence-style brief
Jul 18, 2026 · 2026-07-18
The main dashboard has a new look. The Signals Desk — the daily working surface — now shares the visual language of the Situation Room: a cold, console-style board with a live status bar, an at-a-glance stat strip, and cleaner panelled sections. Nothing moved — the same brief, top stories, most-recent, emerging, and filtered lists are in the same order and navigated the same way. It just reads better on a big screen, and still works on a phone. The All Stories, All Articles, and Briefs pages — and the pop-ups behind them (create / add / merge / split) — carry the same look.
The at-a-glance counts are now accurate. "Stories tracked", and the "to cover / to monitor / emerging" figures, now count the *whole corpus* instead of just the first page of it — so the numbers match what you get by filtering All Stories.
The morning brief now reads like a real intelligence brief. Its opening synthesis is no longer one long paragraph: it's a BLUF (bottom line up front), a Key developments list, and a Priority · cover now list — the same substance the brief always gave you, just structured so you can scan it in seconds. Each story in the brief is now a clean bullet with a priority-coloured marker, and full summaries are still shown so it stays a no-click read. Older archived briefs read exactly as before.
Putting the full article text to work
Jul 15, 2026 · 2026-07-15
The AI analysis now reads the full article, not just the teaser. Both the daily run and the per-story "Analyze" button now work from the complete article body — for outlets that publish only a short feed summary — building on the full-text fetching added earlier the same day. Stories we've already covered still send only the genuinely new coverage each day, so cost stays proportionate.
Why it matters. A side-by-side test showed the fuller text surfaces concrete specifics the teaser leaves out — names, dates, figures, and how outlets differ — for a better-grounded read. Where an outlet's full text isn't available, the analysis falls back to the feed summary exactly as before, so a story is never worse off.
Reading the whole article, not just the headline
Jul 15, 2026 · 2026-07-15
Fetching full article text, behind the scenes. For the stories we analyze, the platform now pulls the *full article body* from outlets that publish only a one-line teaser in their feed — BBC World & Africa, the Guardian, Fox News, Foreign Affairs, Defense One, the Jerusalem Post, Al Jazeera, and The Africa Report — instead of leaning on the short summary alone. It's stored next to each article, banking a far richer body of source text. Nothing changes in how you use the platform yet.
Why it matters. The most valuable editorial signal — how different outlets frame the same event, and what each one leaves out — is impossible to see in a 100-character teaser. Gathering the full text now (at no AI cost — it's just fetching public web pages) sets up the deeper cross-outlet analysis coming next, and lets us prove the material's quality before we spend anything to process it.
Careful by design. Every fetch is quality-checked to confirm it's real article text — not a paywall notice, cookie banner, or navigation clutter — and if anything looks off it simply keeps the original feed summary, so a story is never worse off. Sources were switched on one at a time, each only after we eyeballed real extractions from it; a couple already carry their full text in the feed and were left as-is.
Keeping an eye on what the AI costs
Jul 15, 2026 · 2026-07-15
Day-by-day cost tracking, behind the scenes. Every paid AI step — the once-daily intelligence analysis and the translations that render the dashboard in Russian — now records what it cost the moment it runs, saved so the figures survive app updates and restarts. Nothing changes in how you use the platform.
Why it matters. It gives us an honest, at-a-glance picture of daily spend, so as we add more AI-powered features we can grow them deliberately — and catch any surprise the same day instead of finding out weeks later.
A wider, sharper source list
Jul 15, 2026 · 2026-07-15
Nine new outlets on the wire. We widened coverage where the mission actually lives: the US State Department's official releases (back online and detailed), AP, BBC World, the Guardian (incl. its Russia desk), the Washington Post, Politico, Defense One, and Responsible Statecraft — plus, for context, the English edition of Russian state media (TASS) so we can see how Moscow frames the same events for English readers. These were chosen for coverage and framing you can't get elsewhere, and every one was tested against the live pipeline before being added — not assumed.
Dropped a dead source. CNN's public RSS feeds have been defunct since 2023; they were quietly collecting nothing, so we removed them. (No free CNN feed exists today; we'll find another route if we want CNN back.) The All Articles source filter now lists only sources we still actively collect, so retired feeds like CNN no longer clutter the dropdown.
Russian state media, clearly labeled. We've added TASS (the English-language edition of Russian state media), mainly for Russian military/weapons and Ukraine-front coverage and to see how Moscow frames events for English readers. Every item shows its source, so TASS is never mistaken for independent Western reporting. We treat it as context, not corroboration — and formally separating it from any "multiple outlets agree" signal is part of the framing work still ahead.
Africa on the board, and a cleaner corpus
Jul 15, 2026 · 2026-07-15
New AFRICA beat in the Situation Room. A ninth theme joins the eight standing beats, covering the Sahel, the Horn, Nigeria, and Sudan/DR Congo — security, coups, terrorism, and Russian/Chinese engagement across the continent. It's fed by four new Africa-focused sources chosen for coverage you can't get elsewhere: BBC Africa (broad, reliable), The Africa Report (Sahel / France–Russia realignment), African Arguments (deep analysis), and the Long War Journal (militant-group tracking).
Less noise on the board. We now filter out recurring filler that was slipping in as if it were news — the Times of India "Quote of the day" / "Word of the day" columns, auto-generated stock-price pages, and sports results — and we cleared the back-catalogue of it out of the corpus entirely. The filtering is deliberately precise (it keys on the exact filler phrasing), so real stories that merely mention a word like "quote" or "puzzle" are untouched.
Even less noise — now learning from your "mark irrelevant" calls. We reviewed every story the team has flagged as irrelevant and taught the filter to catch the recurring off-mission patterns on its own: celebrity and royal-family tabloid coverage (the Prince Harry saga), sports (football, cricket, rugby), and "wow science" curiosities (fossils, dinosaurs). It stays topical, not geographic — UK and European politics remain in (they bear directly on Russia); only genuinely off-mission material is dropped. Carefully tuned so real stories survive — an energy piece on "fossil fuels" or a report that happens to mention a football-fan activist is kept.
We stopped paying to translate junk. Items the filter marks irrelevant are no longer sent for machine translation, trimming running cost with no effect on the stories that matter.
Sharper topic detection under the hood. The behind-the-scenes list of people, places, and organizations each story is about is now cleaned up — common words and duplicates removed, related names merged (e.g. "Russian" folded into "Russia"), and multi-word names kept whole ("Strait of Hormuz"). It makes the beat filing and coverage read more accurately, and lays groundwork for the blind-spot and impact analysis coming next.
The Situation Room: the day's coverage, by beat
Jul 14, 2026 · 2026-07-14
A new top-level view groups everything into the beats you actually follow. Open the Situation Room from the header, and the hundreds of individual stories roll up into eight standing themes — US–Iran, Ukraine–Russia, US domestic, Israel–Palestine, NATO/Europe, China/Indo-Pacific, the wider Middle East, and Energy/Markets — each with a live coverage trend so you can see at a glance where the volume is moving.
Click any theme to drill in. You get that theme's on-beat stories, a coverage-over-time chart, a narrative read and its threads, an event timeline, a cross-outlet frames panel (how different outlets frame the same story — stance, emphasis, and what each one leaves out), a "negative space" blind-spots note, and a second-order impact model.
The theme lists are tightly on-beat. We reworked how a story gets filed under a theme so it keys on the *headline*, not a stray keyword buried deep in the coverage. Before, a big story — a senator's death, a Hormuz explainer — could leak into five or six themes at once; now each lands under the beat(s) it's genuinely about (and a truly dual-beat story, like a Hormuz item that's both US–Iran and energy, still shows under both). Open CHINA and you get China stories, not Russia and Iran ones.
It works on your phone. The Situation Room now lays out cleanly on a small screen — no more content running off the edge.
About the deep panels: the narrative reads, frames, and impact models are a hand-authored snapshot for now — real and rich, but a single point in time.
Safer, clearer story curation
Jul 13, 2026 · 2026-07-13
Working with a story is now split into two clearly separated zones. Actions that affect *one or more selected articles* (Split, Move) and actions that affect *the whole story* (Confirm good, Mark bad, Merge) no longer sit jumbled in one row. And while you have articles ticked, the whole-story Merge and Mark bad buttons lock — so you can't accidentally merge an entire story when you only meant to move one article out of it.
Merge, Move, and Split now confirm before they act. Each pops a short dialog naming exactly what it's about to do — how many articles, and into which story — before anything happens. Merge especially, since it folds a whole story into another and deletes the original.
Move articles straight from a story. You can now select articles on a story's page and move them into another story in one step (with that confirmation), instead of detaching and re-adding them.
Splitting a cluster now asks why — and confirms scope. Peeling articles into their own story shows how many you're peeling and asks for a reason (the same short list as marking bad), which quietly sharpens the tuning that improves grouping over time.
Cleaner pop-ups. The old plain browser prompts (for merge confirmation and naming a split) are gone, replaced by tidy in-app dialogs that match the rest of the tool.
See a story's review status, age, and how coverage built up
Jul 13, 2026 · 2026-07-13
Each story now shows whether you've marked it good or bad. A small badge — e.g. *"✓ Reviewed good"* or *"✗ Reviewed bad · Mixed events"* — appears on the story wherever it's listed, with the date on hover, so you can tell at a glance which clusters you've already judged (and re-judge freely; the badge just tracks your latest verdict). It's purely informational and never changes a story's scoring or ranking.
Stories show when they were first detected, right on the card.
A story's sources are grouped by the day they were attached, newest first — so a long-running story reads as a timeline of how its coverage built up.
Jump from an article to its story. On the All Articles page, an article that's part of a story now has a "View story" button (opens the story in a new tab); the headline still opens the original article.
The brief keeps a story only while it's actually developing
Jul 13, 2026 · 2026-07-13
A story now stays near the top of the brief only while it's genuinely developing — not just because articles keep trickling in. Before, *any* new article counted as "fresh," so an old, winding-down story could stay stuck near the top on a drip of follow-up coverage. Now the ranking watches for a *real development* — more than one outlet bringing genuinely new information (the same signal behind the "What's new" line) — so an actively-unfolding story stays prominent while one that's just getting repeat coverage of old news gently drops off.
Why it matters: your daily brief stays focused on what's actually moving, instead of a two-week-old story lingering at the top.
A note on timing: the system needs a few days to build up this "what's developing" history across all stories, so you'll see it sharpen over the coming week. Until a story has a recorded development, it's ranked by how recently it first appeared.
Long-running stories now show "what's new"
Jul 13, 2026 · 2026-07-13
A big, ongoing story now tells you what just developed — instead of re-showing the whole pile. For a story that keeps gaining coverage (a war, an election, a fast-moving crisis), the system now tracks each new turn and shows a "What's new" line — e.g. *"The US conducted a third round of strikes as Iran re-declared the Strait of Hormuz closed"* — on the story card and in the daily brief, in English and Russian. The full summary is still there for context; the new line just surfaces the latest development.
It's also cheaper and faster on exactly the stories that cost the most. Re-analysing a large story used to re-read *every* article in it each time; now it reads only the new coverage against what it already knew — about 5× less work per update on an ~80-article story. So the tool keeps big stories current without the cost climbing as they grow.
Under the hood it also starts a per-story timeline of developments — groundwork for a future "how this story evolved over time" view.
Less noise in the feed: sports and auto-generated stock pages
Jul 12, 2026 · 2026-07-12
Football and auto-generated stock-ticker pages no longer clutter the feed. The relevance filter — the doorman that turns away obvious non-news before it ever reaches you — learned two new kinds of noise. A burst of FIFA / World-Cup football coverage had been slipping in because those headlines say "FIFA" rather than "World Cup", and Reuters publishes auto-generated stock-quote pages ("AAPL.O — Stock Price & Latest News") that Google News lists as if they were articles. Both are now turned away at collection.
Why it mattered. The stock-quote pages all read nearly identically, so dozens of them were clumping into a single fake "story" and climbing the rankings on sheer item count — one such cluster had about 66 of them. Dropping them at the door removes both the clutter and the phantom story.
Scoped so real news stays. The rules match only unmistakable tells, so genuine coverage is untouched — "political football", "referee in the Gaza talks", and ordinary market news like "Tesla stock price soars" all stay in. The exact rules are listed on the How it works page.
Sharper grouping: fewer wrong merges within one outlet
Jul 12, 2026 · 2026-07-12
Same-outlet articles are now grouped more strictly. The system used to sometimes lump together two *different* articles from the same source just because they shared a lot of wording — most visibly Reuters' auto-generated stock-price pages, which all read nearly identically. It now needs a noticeably stronger match (or a shared distinctive name/place) before merging two articles from one outlet, so those junk merges stop. Grouping *across* different outlets is unchanged, and genuine same-outlet follow-ups (a report and its later update) still join as before.
This is the first change to come out of the new evidence-driven tuning: it was measured against your own Confirm / Detach feedback and spot-checked on real headlines before shipping. The exact thresholds shown on the How it works page are updated to match.
Confirm or flag a story's grouping
Jul 12, 2026 · 2026-07-12
You can now tell the system when an auto-grouped story is right — or wrong. On a story's page, use ✓ Confirm cluster if the articles genuinely belong together, or Mark bad (pick a reason) if the grouping is off. These are quiet signals: they record your judgement for later tuning and do not change the story's scores, ranking, or contents in any way — a marked-bad story keeps working normally.
Detaching an article now asks a quick reason. When you detach an article that doesn't belong, you pick one of four reasons — *coincidental overlap*, *mixed events*, *off mission*, or *duplicate story* — in the same click. Same short list is used for Mark bad, so the feedback stays consistent.
This completes the hands-on story-curation tools (Create · Add · Split · Merge · Confirm/flag). What you record here becomes the ground truth we'll use to keep improving the grouping and scoring.
Split a story, or merge two stories together
Jul 12, 2026 · 2026-07-12
You can now split articles out of a story into their own story. Open a story's page, tick the sources you want to peel off, and use Split selected into new story. Those articles move into a fresh, human-curated story (you can give it a title); the original keeps its remaining articles, and both are re-scored right away. This is the fix for a story that swept in an article that doesn't belong with the rest.
You can now merge one story into another. On a story's page, use Merge this story into… — a picker opens with the most similar stories plus a search box. Pick a target and this story's articles all move into it; the target is re-scored and marked curated, and the now-empty story is removed. The merge is fully recorded behind the scenes.
Together with Create and Add, this completes the hands-on tools for grouping coverage the way an editor sees it. Lightweight confirm/flag signals are the last piece, coming next.
Add articles to an existing story
Jul 12, 2026 · 2026-07-12
You can now move articles into a story you already have. On the All Articles page, tick the articles you want, then use the new Add to story… button on the toolbar. A picker opens with suggested stories (the ones most similar to what you selected) plus a search box to find any story by title — pick one and the articles move in.
This works whether an article was on its own or in another story. Moving an article re-homes it: it's added to the story you chose and removed from wherever it was, and both stories' scores update right away. An article you'd previously detached from a story is fully rescued this way — it becomes a normal member again.
Splitting a story, and merging two stories together, are coming next.
Build your own stories
Jul 11, 2026 · 2026-07-11
You can now group articles into a story yourself. On the All Articles page, tick the checkbox next to any articles that belong together, then use the Create story bar that appears at the top of the list and follows you as you scroll — turn them into one story, give it a title (or leave it blank to borrow the first selected article's headline), then click through to the new story.
Hand-built stories are marked "Curated." The automatic grouping will never pull apart what you put together — the articles you chose stay in the story and won't be moved out. (New, closely-matching coverage can still attach to the story on its own as it's discovered, the same way it does for any story, so a story you start by hand keeps growing.)
This is the first step of manual curation. Adding articles to an existing story, and splitting or merging stories, are coming next.
A sharper daily brief
Jul 11, 2026 · 2026-07-11
Sprawling "magnet" stories no longer dominate the brief. Some stories quietly grow by pulling in loosely-related articles that share a keyword (e.g. a "jet-powered drones" story that kept absorbing every drone/missile item), and each new addition used to shove it back to the top of the brief day after day. Now a story the AI judges loosely-grouped gradually settles as it ages, while a genuinely coherent multi-day event (a summit, an unfolding crisis) keeps its place as long as fresh coverage keeps arriving. The result: the top of the brief moves more, and well-sourced coverage that was getting buried now surfaces.
Less accidental clustering. A shared country name alone ("China", "Russia") no longer merges two unrelated stories — so a factory-fire story stops getting stapled onto a car-factory acquisition.
Fewer wasted AI runs. A story is only re-analysed when genuinely new coverage from new outlets arrives, not every time one more article trickles in.
A story's articles now list newest-first, so the latest coverage is at the top and the original seed article sits at the bottom.
Under-the-hood cleanup
Jul 11, 2026 · 2026-07-11
A housekeeping release — nothing about how the platform works has changed. This was a maintenance pass: we removed duplicated code behind the scenes and brought the internal documentation back in line with the app, so it stays easy to maintain and extend as it grows. Scoring, sources, analysis, and the brief all behave exactly as before.
More consistent navigation. The links at the top of each page (Dashboard · All Stories · All Articles · Briefs) are now the same everywhere, so you can jump straight to any section from any page.
The "How it works" page, brought up to date
Jul 9, 2026 · 2026-07-09
The public explainer at `/about` now matches how the system actually works today, and is restyled to the same dark theme as the dashboard. It had drifted out of date; it's now a faithful, top-to-bottom description of the live pipeline.
What's new on the page: ranking is now driven by the two scores that actually order the brief — *importance* and *editorial value* — with *momentum* retired and *audience interest* shown as dormant until social sources arrive. The mission-keyword method is shown in full (the CORE / CONTEXT / MARKET tiers and every term), alongside how a story's AI confidence can demote or drop it, and the current cover / monitor / wait cut-offs. The source list is the real thirteen, and the charts were redrawn to match. Still bilingual (EN / RU).
Clicking a story's title now opens that story's own page — the same view you get from the Share button — instead of jumping out to a single news site. From there you can read every source, see the analysis, and grab the share link. This is especially handy in Emerging Stories and Most Recently Detected, which only show a title. (Before this, the only way to open a single-story page was to copy the Share link and paste it into the address bar.)
A "← Back" button on a story takes you back where you came from — the story list, the dashboard, or the briefs archive. Returning to the list keeps your place (your scroll and any "Load more" pages you'd opened).
Detach a mis-grouped source from a story's own page — the "detach" control next to each source article now appears on a story's individual page too, matching the rest of the app.
Sharper story scoring
Jul 9, 2026 · 2026-07-09
Story scores tell stories apart again. "Editorial value" used to max out at 100 for anything that mentioned a handful of relevant words, so dozens of stories tied at the top and the number stopped meaning much. It now weighs *how much a story is really about our core beats* — Russia, the war, weapons, Russian politics — above passing mentions of other topics, and it rewards stories whose subject is right in the headline. A genuinely on-topic, well-sourced story now clearly outranks a keyword grab-bag. "Importance" got the same fix, so a story covered by eight outlets no longer ties one covered by two.
Broader vocabulary: added terms for weapons systems, the war map, Russian officials, and Africa / the Sahel, and cleaned up false matches (e.g. "war" was being triggered by the word "toward").
The "cover immediately / monitor" cut-offs were re-tuned to match the new scale. Expect these to be adjusted again as the team gives feedback — tell us if the brief feels too full or too thin.
Analyze any story on demand
Jul 8, 2026 · 2026-07-08
The "Analyze" button now appears on every un-analyzed story, not just the ones in Top Stories. Open the All Stories page, find any story that hasn't been analyzed yet, and run the AI analysis on it right there.
A sharper morning brief
Jul 7, 2026 · 2026-07-07
The brief now leads with its best-sourced stories. Every analyzed story carries an AI confidence rating; the brief and Top Stories now use it. Stories the AI flags as incoherent or thinly-sourced are held back, and well-corroborated stories — several outlets covering the same event — rise to the top. In practice: a strong multi-source story that used to sit buried can now lead the brief, and mis-grouped "junk" clusters no longer take up a slot.
Clearer story scores
Jul 7, 2026 · 2026-07-07
You now see only the scores that actually matter. Each story shows Importance and Editorial value — the two numbers that decide what rises to the top — plus Audience interest, which is greyed out for now with a note: it switches on once we add social-media signals (Reddit, YouTube, X). The old "Momentum" number wasn't changing anything you saw, so it's been removed to keep the display honest and uncluttered.
"Emerging Stories" now lists the most recently active stories first.
Share a story with a link
Jul 7, 2026 · 2026-07-07
New "Share" button on every story. It copies a direct link to that one story. Anyone on the team with the platform password can open the link and lands on that story by itself, expanded to the full analysis — handy for flagging a strong example to a colleague (e.g. over Telegram).
More sources & sharper market coverage
Jul 6, 2026 · 2026-07-06
Five new outlets now feed the platform: The New York Times (World & Politics), Foreign Affairs, Al Jazeera English, Times of India, and The Jerusalem Post — widening coverage across the US, the Middle East, and international affairs.
Better economic radar. Stories about oil, gas, grain, fertilizer, key metals, and major cryptocurrencies now score higher, so market and commodity news surfaces in the brief instead of getting lost.
Less junk. Personal-finance ad copy that used to slip in ("0% APR", "home equity loan") is now filtered out.
Version footer & What's New
Jul 6, 2026 · 2026-07-06
Every page now shows the current version in a small footer at the bottom.
Clicking it opens this What's New page, so the whole team can see at a glance what changed and when.
Added a plain-language guide to how the whole platform works — the pipeline, the scores, and how stories are grouped — as a shareable page.
A clearer picture and a better brief
Jul 5, 2026 · 2026-07-05
All Stories page. A new page lists every current story in one browsable, searchable, sortable place — so a genuinely new story can't hide between the dashboard's separate sections anymore.
Tidy up your feed. You can now remove an article that doesn't belong in a story, mark an article as noise so it disappears from every story and brief, or promote a wrongly-hidden article back into the feed. Every one of these is reversible.
Richer morning brief, once a day. The brief now opens with a short overview of what's in it and why, and each story expands to the full analysis — how it's being covered, questions worth exploring, suggested angles, and sources. It now arrives once each morning instead of several times a day.
Cleaner controls. Removed the buttons that let anyone trigger expensive batch work; the only manual analysis is a per-story "Analyze" on stories that haven't been analyzed yet.
Smarter story grouping. Improved how articles are grouped so unrelated news stops getting mixed into the same story.
Fresher rankings. Stories are now ranked with a freshness boost, so the brief and Top Stories reflect what's actually developing rather than staying frozen on older items.
Launch
Jul 4, 2026 · 2026-07-04
The platform went live at intel.vbn.news.
A live dashboard that refreshes on its own, with a "Most Recently Detected" section showing new stories as they arrive.
All Articles archive: browse every collected article and jump straight to the original reporting.
The whole dashboard can now be read in Russian, not just the menus.
A briefs archive that keeps every past morning brief, and clickable source links throughout.