Stellar engines levitating on radiation pressure can be passively stable. The geometry determines whether they need a controller.
Note (2026-11-07): the abstract uses a simplified model and says a stellar engine with a ring supporting a reflector 'can in principle be passively stable', that static swarms (Dyson bubbles) are unstable but 'in principle' can become self-stabilising as a dense cloud, and that uniform discs are always unstable. The essay drops the in-principle hedge, merges the engine and swarm cases, and gives a tilt-torque mechanism, a fate for the disc (falling into the star or escaping) and a remark about stellar lifetimes that are not in the abstract.
A reflector hovering above a star, held aloft by radiation pressure against gravity, is a Dyson-class megastructure. The simplest design — a uniform reflective disc — is unstable: any small displacement from the equilibrium position grows. The disc drifts, tilts, and eventually falls into the star or flies away. Active control is required. This instability has been treated as a fundamental limitation.
McInnes (arXiv:2603.00203) shows that the instability is a property of the disc geometry, not of radiation-pressure levitation itself. Ring-supported reflectors and dense cloud formations of smaller reflectors can achieve passive stability — perturbations generate restoring forces without active intervention. The mass distribution across the reflector determines the sign of the restoring torque.
The structural insight is sharp: stability is a design parameter, not a physics constraint. The same radiation pressure that makes a disc unstable makes a ring stable, because the ring's mass distribution changes how the reflected radiation pushes during a perturbation. A tilted ring experiences asymmetric radiation torque that returns it to equilibrium. A tilted disc experiences asymmetric torque that amplifies the tilt.
For engineering megastructures that must persist for stellar lifetimes, the difference between “needs a controller” and “is passively stable” is the difference between fragility and robustness. The controller can fail. The geometry cannot.
McInnes, "Stellar engines and Dyson bubbles can be stable," arXiv:2603.00203 (2026).