The intel stack
The map, the sensors, the layers, the call.
Hypha composes multi-modal intelligence inside the room with the data. Sensor tracks, doctrine, after-action reports, full-motion video, telemetry. The operator sees the layers. The commander makes the call.
01 / 02 · LAYER STACK
The layer stack
Five layers. One operator view.
Each semi-transparent plane in the map mesh is a Hypha intel layer. The operator sees them stacked, queries them in English, and exports the brief with provenance per layer. Read bottom to top: sensor ground truth resolves up into the composed view the commander reads.
-
LAYER 1 / GROUND TRUTH
Sensor footprints
Sparkle-blue cones for sensor coverage, gold annotations for points of interest. Updated as the data refreshes.
-
LAYER 2 / BASELINE
Pattern-of-life
The baseline. What "normal" looks like for the area, the unit, or the window the operator selected.
-
LAYER 3 / DELTA
Anomalies
Departures from the baseline that crossed the Anomaly Agent's threshold, with the model's confidence per anomaly.
-
LAYER 4 / REFERENCE
Doctrine and reference
RAG retrieval from operator-supplied doctrine, with citations. The "what does the manual say about this situation" layer.
-
LAYER 5 / OUTPUT
Operator synthesis
The composed view. The brief. The thing the commander reads.
02 / 02 · AUDIT TRAIL
The audit trail
Every layer has a source. Every source is citable.
The Visualization Agent does not paint a layer it cannot source. The intel architecture is built around provenance: every chip on the dashboard links back to the data that produced it, every citation links back to the document, every model output links back to the model card. The auditor and the commander are reading the same trail.
- Dashboard chip TRACE the data that produced it
- Citation TRACE the source document
- Model output TRACE the model card