A star tracker is a camera that already flies on most spacecraft. It was built to recognise patterns of stars and report which way the bus is pointing [4]. FALCON asks the same sensor to pick out the points that move against that background, match them to catalogued satellites and debris, and then run the geometry backwards: if you know where those objects are, their apparent positions tell you where you are [5]. The hardware bill for that is close to zero, because the camera was already on board. What NASA and EraDrive added was software, an embedded pipeline the Stanford spinout calls Era-Core, plus an onboard catalog [3].
The arithmetic of the three-day run is worth sitting with. Controllers uploaded predicted orbits for roughly 20,000 objects [6]; the payload came back with improvements on more than 200 of them [7], which is about one percent of what it was carrying [16], and an average above 66 objects a day [17]. That is not a complaint. It is the shape of the measurement: FALCON refined the objects that happened to cross its field of view during 72 hours, not the catalog as a whole.
The number NASA's account does not supply is the one an operator would ask for first. The estimates are described as more precise than the existing catalog entries [8], and better than the position predictions the ground uplinked [10], but no magnitude and no units are attached [18]. A conjunction decision is made on covariance, not on the direction of an improvement. Until there is a metres figure, and one that separates a bright cooperative satellite from a tumbling fragment, this is a demonstrated capability rather than a usable input to collision avoidance.
Then there is the extrapolation to the Moon. GPS is weak or absent out there, which is the whole motivation [11], and FALCON is being read as the answer. But the landmarks in this experiment were other people's hardware and other people's junk, identified against a public catalog maintained by the U.S. Department of War [5]. Cislunar space has neither the population nor the survey. Read NASA's own framing closely and it points at future networks of satellites operating without GPS, and coordinated systems supporting activity on the Moon or Mars [13]: a swarm using its own members as references. That is a related problem, and a harder one, because the landmarks have to be launched before they can be sighted.
Where this lands soonest is the traffic problem it was tested inside. Roger Hunter, who manages NASA's Small Spacecraft and Distributed Systems program at Ames, put space-traffic monitoring and collision avoidance ahead of alternative navigation in his own list of implications [12]. The demonstration sits in low Earth orbit, and so does the crowd it needs.
