A seeded raid simulation across a real 150 km section of the southern Moldova–Ukraine border, built on the open-source Panopticon/BLADE engine. Ten raids, eighty-four intruders, three Falx roles — run end to end through detection, classification, human confirmation, and intercept, at two deployment densities.
Traced from the actual Moldova–Ukraine boundary (Natural Earth 10m), not a sketch — 15 Murus bases and 115 sensor towers placed to true scale along the line, Giurgiulești to Palanca.
Same seeded raid schedule (rng seed 42), same 150 km section, run twice: once at MVP density (5 bases, 20 towers), once at full density (15 bases, 115 towers). Every intruder passes the same five gates before a Falx Lupus is scored as a kill.
Going from 20 to 115 towers lifted detection from 60% to 83% of all intruders — the single largest swing in the whole chain, and the cheapest one to buy.
MVP launches exactly 20 Falx Lupus — its entire fleet — well short of its 41 confirmed tracks: it's interceptor-limited. Full launches 53, exactly matching its confirmed count: it's limited further upstream, at detection.
Every launch in both runs passes classification and a confirmation gate first — no Falx Lupus fires on detection alone. The final decision stays held by a human, by design, not by exception.
Built on the open-source Panopticon/BLADE simulation engine, driven through a custom detect → classify → confirm → launch → intercept layer written for this trial. Drone performance, tower detection radius, classification accuracy, and intercept probability are placeholder assumptions chosen to exercise the pipeline end to end — not vendor-confirmed specifications. Treat the percentages above as a validated methodology, not a performance claim, until those figures are replaced with real data.