Autonomous, self-optimising AI is being wired directly into physical infrastructure. In orbit there is no engineer to intervene — adaptive behaviour that destabilises the platform destroys the asset. Today, adaptation is permitted by default and safety reacts after the fact. [Background ¶0010–0014]
Authority is earned, not assumed. Every adaptive command is gated by measured physical stability. Watch each eclipse transit: as solar input collapses, the supervisor tightens authority before any limit is violated, and restores it only after stability persists. [Claims 1, 4, 7, 11]
Five nodes, five independent local supervisors — at any instant some are sunlit at full authority while others self-restrict in eclipse. Coordination emerges through the shared physical substrate, not a central controller. [Claims 14–15]