Situation Picture
Regional situation pictures, threat levels, dossiers on actors and assets, flights and ships.
- Good for
- One picture per conflict region: news, infrastructure, fires, flights and ships on one globe, with a daily assessment that a second model has to sign off.
- Measured
- 61,452 infrastructure objects in 12 classes, mapped inside 16 of 17 conflict zones and refreshed weekly, as of 8 October 2026.
- A daily assessment in which a second model checks every judgment against its evidence; 4 of 5 passed in the first published run.
- Open questions
- Do the threat levels move before the news does, or with it?
- How many published judgments still hold a week later?
Situation Picture
One dot, one day's value in one series. Each panel has its own scale.
Only 4 series are measurements (solid). 23 are claims by the two sides (grey), 2 are assessments (blue). The page keeps them apart.
29 series on the war in Ukraine, 2026-08-07 to 2026-11-05.
Source: Ukrainian General Staff, Russian Ministry of Defence, DeepStateMap, Institute for the Study of War
Measured 5 Nov 2026, 06:25 UTC
Three sources joined
- Geographic data. Flight and ship movements, infrastructure.
- News. From Retina.
- Actors. From Chimera.
- Situation picture. Regional threat levels, dossiers on actors and assets.
The situation picture joins geographic data with the news from Retina and the actors from Chimera: regional threat levels, dossiers on actors and assets, flight and ship movements.
What is mapped, and where
- 61,452
- infrastructure objects
- 12
- classes
As of 8 October 2026: 61,452 infrastructure objects in twelve classes, mapped inside conflict zones rather than worldwide and refreshed weekly from OpenStreetMap. Of the 17 zones, 16 currently hold objects; Gaza came back empty after a failed refresh on 4 October and waits for the next one. Satellite fire detections in those zones are ranked by closeness to that infrastructure, then by fire power. Aircraft and ships are shown as a selection, not as traffic: military aircraft worldwide plus civil traffic inside the zones, and state vessels worldwide plus all traffic inside the zones.
Two faces on the public site
- Front page. A globe, layers switchable on the left, an intel stream on the right, the highest-scoring region in the header.
- Analysis deck. Nine views: Situation, Zones, Ties, Movement, Signals, Narratives, Infrastructure, Ledger, Status.
- Each view. The question it answers, how to read it, what it is not.
On the public site, infinimind.info, the situation picture has two faces. The front page is a globe with its layers switchable on the left and an intel stream on the right; the header names the highest-scoring region rather than one number for the whole world. Behind it sits the analysis deck with nine views: Situation, Zones, Ties, Movement, Signals, Narratives, Infrastructure, Ledger and Status. Each view opens with the question it answers, a line on how to read it and a line on what it is not. The screenshots below, captured on 15 October 2026, go through them one by one.
The assessment of the day
- One model. Writes the judgments, each with a likelihood, what to watch for, the evidence.
- A second model. Checks every judgment against the evidence it cites.
- Published. Only judgments that pass. Too few: “No assessment today”, with the reason.
Above every view stands the assessment of the day: a short list of judgments on what matters most today, each with a likelihood, a line on what to watch for, and the evidence it cites. It is an AI draft. One model writes it, a second model checks every judgment against the evidence it cites, and only judgments that pass are published; on a day when too few pass, the panel reads “No assessment today” and gives the reason. The first published run passed 4 of 5 judgments, and the footer of the panel says so.
Why this exists at all, and the limits the lab holds itself to, are set out once for all perception instruments on the instruments page.
What comes out
Snapshot 5 November 2026Eight watched regions, each scored from what the layers show there: fires near infrastructure, military news, censorship, anomalies. The score is a reading of the lab's own layers, not an assessment of the world.
Region scores at 2026-11-05, 13:03 UTC
Scores run from 0 to 1. The live site shows the highest region and its direction over the last day, not one number for the world; the combined figure (the worst region plus 0.04 for every region at high or above) stays on its analysis page. The current picture is at infinimind.info.
- Ukraine0.60 high
- Iran0.60 high
- Gaza0.60 high
- Sudan0.45 high
- East Med0.45 high
- Russia0.45 high
- Gulf0.45 high
- Black Sea0.45 high
What drives Ukraine: med mil news 17, censorship high 10, tda mild 0.48, cyber apts 1.
Live counters, and what they count
| counter | shown | means |
|---|---|---|
| Fires near infra | no data | The layer delivered nothing when this snapshot was taken, so there is nothing to count. That is a gap in the reading, not a zero in the world. |
| Fires in conflict zones | no data | The layer delivered nothing when this snapshot was taken, so there is nothing to count. That is a gap in the reading, not a zero in the world. |
| Anomalies crit/high | 1 of 33 | Anomaly cells at critical or high severity, of all cells delivered. Volatile by design. |
| Regions high or above | 8 of 8 | Regions scored high or above, of all regions scored. |
Infrastructure catalogue, by class
- oil storage15,403
- HV substations15,209
- power plants11,285
- refineries9,351
- military sites5,814
- airports2,851
- dams1,955
- rail yards292
- depots285
- ports208
- data centres163
- nuclear sites110
On screen: the analysis deck, nine views
Captured 15 October 2026View Situation, 15:50 UTC. On top the assessment of the day, which heads every view. Below it the one global reading (high, 0.66 of 1.00, four of eight watched regions at high or above), the morning briefing, the region scoreboard with a 24-hour curve per region, and the conflict feed. The first stop: how bad the picture reads right now, and where it reads worst. View Zones. All 17 conflict zones in one ranked list, the biggest swing of the last 24 hours above it; here 8 of 17 carry an hourly situation report and Gulf scores highest at 0.50. The panel's own warning: a zone with no report is not a quiet zone, it is an unwatched one. Dossier and Compare open a single zone or set two side by side. View Ties. Who is pulling against whom: 32 actors tied into 27 pairs, 12 in conflict and 15 in alliance. Every pair carries weight, confidence and intensity and a sentence the extraction wrote from the headlines, so a reader can check why the tie is there. Sub-views for blocs, single actors and cyber. View Movement, sub-view Air. 494 of 507 aircraft in the feed carry a military marker, and the panel says what that marker is: a lookup in ICAO address ranges and callsign prefixes, not a judgement. Left the operators behind the marked aircraft, right every aircraft with altitude, speed, heading and when it was last seen. Sea and Space are the other two sub-views. View Signals. Everything that came in, ranked: 2,121 readings in this capture, of which the top 20 are listed with the sentence that put them there. The rank rewards a second, independent source at the same spot over volume. Every row opens down to the raw signals behind it; Anomalies and Sensors hold the raw lists. View Narratives. Who is claiming what: 50 claims in circulation, carried by 118 reports from 29 outlets, 9 corroborated, 40 open, 1 one-sided, and 41 of the 118 carriers state-controlled or state-aligned. Nothing here says something happened, only that these outlets said it did. Sub-views for entities, first reports and censorship. View Infrastructure, sub-view Burning. 1,287 fires within 2.5 km of a tracked site in this capture; the table lists the 500 the layer returns, with class, zone, distance and radiative power. Not a damage report, the panel says: a satellite cannot tell shelling from an industrial flare or a burning field. View Ledger. What Kyiv and Moscow each claim the other side lost: 29 series from 4 feeds, both figures side by side with their median ratio, and a further tile for Russian claims a Washington institute called unsupported. Nobody verified a single figure, and the view says so before it shows one. View Status. The deck checking itself: of the 8 layers this page checked, 2 report stale data and 6 carry no age at all, and a layer is judged against its own expected cadence, not the clock. Ages are taken when the public bundle is baked, every fifteen minutes, and counted forward. A statement about the lab's own installation, not about the world.