Every system ships with its world
Scenarios are generated in the thousands, run overnight and scored automatically — across terrain, weather and failure modes no test range can schedule. The same binaries that pass in simulation are the ones that fly.
Four scenarios, stepped one decision at a time
These are recorded runs from the scenario set, not live compute in your browser. The sandbox below runs the real thing.
Two aircraft hold overlapping lanes across the sector.
A crossing is detected on a linear front and has to be held in custody while a ground element closes.
A scenario set is a test rig, not a showreel
A scenario run is not a demonstration. It is one sample from a generated set, and what a run is worth is the distribution behind it — which is why the figures on this site are quoted as medians over runs rather than as a best case.
Terrain, weather, link quality, asset loss and target behaviour are sampled rather than scripted. Nobody writes the scenario the system is good at.
Thousands of variations run unattended. A change to the assignment layer is measured against the whole set before anyone flies it.
Each run is scored on custody continuity, operator inputs required, tasks completed and assets recovered. A regression shows up as a distribution moving, not as an opinion.
What passes goes to hardware in the loop and then to a range. The binaries do not change between the last simulated run and the first flight.
Build your own scenario and run it
Sandbox mode lets an evaluator place their own terrain, force mix and failure conditions, run the set, and read the scores against their own criteria rather than ours.
It is in build. Access opens to named programme offices and primes first, because a public sandbox that shows the assignment layer reasoning is a disclosure decision, not a marketing one.
Tell us your organisation and the scenario you actually want to run. We will say plainly whether the set covers it today.
Nothing on this page phones home. No analytics ship in anything delivered to a customer, and the site does not undercut that claim either.