FLEETSENTINEL
HALL-A · liquid-cooled AI fleet · streaming telemetry
t = —
PLANT NOMINAL
Scenario
Window
480 steps · 4.0 h
Racks
8
Inject faults
Claude writes the briefing
RUN SCENARIO
PUE
—
total / IT power
Cooling capacity
—
of loop capacity
Power headroom
—
vs 1,250 kW envelope
Open incidents
—
at or above WATCH
IT load
—
fleet power draw
Hall map — rack severity
click a rack to inspect
Run a scenario to populate the hall.
▶
Rack detail
GPU temp °C
Coolant flow L/min
ECC errs/min
Severity 0–100
Predictions — time to breach
No forecasts yet.
Incident feed
No incidents.
Shift briefing
Run a scenario to generate a briefing.
Streaming 4 hours of fleet telemetry through the pipeline…