Feed
themosaic.news/feedEvery ingested signal, scored 0–10 for credibility and significance and filterable by topic, country, and source trust before you read it.
Decides: Which of the raw intake deserves attention right now?
Baseline-aware signal intelligence
What changed—and the evidence behind it.
Mosaic monitors open information and physical-world signals, learns the normal rhythm of every country, source, and chokepoint it tracks, and connects statistical deviations into situations you can open, inspect, and verify.
Every surface is a public URL — explore the live system before you create an account.
500+ sources monitored · 300+ sources producing signals in a typical 24 hours · signals across 190+ countries · baselines recomputed every 15 minutes.
Source counts refreshed hourly from the live system.
Representative detectionsvs each subject’s baseline
Dark vesselssev 7.0
Strait of Malacca
19 underway vessels stopped transmitting AIS — receiver gaps cannot be excluded.
vs 7-day AIS baselineInspect →
Coverage spikesev 6.7
South Africa
Coverage 3.6× above its 30-day norm (27 mentions in 24h vs ~7.6/day).
vs 30-day baselineInspect →
Dark vesselssev 6.0
N. Red Sea / Suez
2 underway vessels stopped transmitting — receiver gaps cannot be excluded.
vs 7-day AIS baselineInspect →
Real detection classes in the live card format, shown when the trusted-baseline feed returns fewer than two mature results. The engine recomputes baselines every 15 minutes.
How it works
From a statistical deviation to an investigable situation
One real detection class, walked through the way the engine actually ran it — not a mock-up. Mosaic separates what it observed, what deviates, and what an analyst can conclude, and labels each along the way.
01
Establish the baseline
A trailing norm for one strait
Mosaic holds a continuous AIS feed over the Strait of Malacca — one of nine watched chokepoints — and learns its trailing 7-day norm: how many vessels are underway, and how consistently they transmit. A subject whose baseline has not matured is never scored as severe.
02
Detect the deviation
Transponders go quiet
When underway vessels stop transmitting inside the watch box, a dark-vessel anomaly fires — scored against the strait’s own norm, not a global average. The severity measures how far behaviour sits from baseline; it is not a claim about intent. Mosaic records that transponders went dark and that this is abnormal for this water — not why.
03
Correlate the evidence
Independent signals, checked
The corroboration engine scans the monitored sources for the same place and window. Confirmed has a definition — two or more independent sources; a single source is labelled single-source. When a physical anomaly and a narrative anomaly align in the same place and window, that becomes a fusion alert, the strongest class the system emits.
04
Build the situation
An investigable record
Related signals and anomalies cluster into a situation — with trajectory, key facts, and the full evidence trail. It is published only after passing a coherence and evidence gate; below the bar it is held for review rather than shown. Standing watches then evaluate in real time as new evidence lands.
No step involves prediction. Mosaic measures deviation from what it has observed, and links every figure back to the signals that produced it.
The product
Four surfaces onto the same evidence
Each surface answers a different question, and all of them are public URLs. This is the live system, not a rendering of one.
Situations
themosaic.news/situationsTracked events with trajectory and evidence depth, where confirmed and claimed are separated at the level of the individual claim.
Decides: What do I need to be across, and how solid is it?
Geo
themosaic.news/geoCountries coloured by deviation from their own trailing baseline. Coverage gaps are rendered as gaps — never coloured as an alarm.
Decides: Where is abnormal change concentrated?
Research
themosaic.news/researchRetrieval across signals, situations, and entities. Every result shows why it matched; answers cite their evidence rather than free-wheeling.
Decides: What is the recorded evidence on this specific question?
Methodology & trust
How the scores are made—and where they stop
Scores compress evidence; they never replace it. Everything below is inspectable down to the underlying items, and the limits are stated in the same type as the strengths.
What Mosaic computes
Source credibility is earned, not assigned. Every source is scored on its track record and capped below certainty — including wire services and official channels.
Corroboration is mechanical. “Confirmed” means two or more independent sources. One source, however credible, is labelled single-source.
Severity is distance from baseline. Each detector scores against its subject’s own trailing window, recomputed every 15 minutes.
Every number links back to its signals. Claim status, confidence, and severity are first-class fields, not presentation.
What Mosaic does not do
It does not predict events. A severity score measures distance from baseline, not the probability of what happens next.
It withholds severity on young baselines. Until a baseline matures, detections carry reduced weight or none — deliberately slower on new subjects.
It does not render absence as alarm. A quiet region may be a coverage gap, and the map says so rather than colouring it red.
It is machine-assisted and can be wrong. Every output links to primary sources — verify critical intelligence there.
Read the full methodology, including how every score is computed →
Coverage
What Mosaic watches
Monitored sources and recently producing sources are different figures, and Mosaic keeps them separate. Physical-world sensors are counted apart from sources — one number never pads the other.
Open information
500+ sources monitored, of which 300+ produced a scored signal in the last 24 hours — wire and national feeds plus 90+ curated Telegram channels, each credibility-scored.
25+ source languages, machine-translated to English; items that can’t be translated are labelled, never silently mislabelled.
16 analyst topics · 7 macro-regions, from Conflict and Defense to Energy, Cyber, and Climate.
Physical-world signals
Continuous AIS ship tracking across 9 strategic chokepoints — Hormuz, Bab el-Mandeb, Suez, Malacca, the Taiwan Strait, the South China Sea, the Black Sea, and others.
Military-aircraft monitoring (ADS-B) over six strategic zones, polled every two minutes.
Five detector classes — coverage spikes, source silence, dark vessels, flight surges, and physical-plus-narrative fusion.
Sources are monitored, not partners. No outlet named or counted here endorses Mosaic, and Mosaic implies no endorsement in return.
Use cases
Built for people who have to be right about what changed
Geopolitical risk
Track state and non-state behaviour against baseline rather than headline volume, and brief with an audit trail — confirmed versus claimed, per source, with trajectory.
Macro & market intelligence
Size a development against its own norm before the desk opens, with physical-world indicators most feeds never carry. An information service — not investment advice.
Enterprise & security risk
Watch the countries, chokepoints, and actors your operations depend on. Standing watches evaluate as each signal lands, with the evidence attached.
OSINT & investigative research
Curated Telegram channels beside wire feeds, auto-translated and credibility-scored. Source silence is itself a detector, and every claim traces to its primary.
See what changed today.
The feed, the map, and the methodology are public — no account needed. An account adds standing watches, alerts, and the morning brief. Free during early access.