We present a unified variational principle governing the behaviour of complex systems across nine orders of magnitude in scale—from thermonuclear plasma toinstitutional economics. The central mathematical insight is that the classical principle of least action, which yields an extremum of the action functional, admits two fundamentally distinct classes of solutions: extractive trajectories that minimisecost and maximise yield (the saddle-point logic of parasites and nuclear fission),and creative trajectories that maximise long-term geometric proximity to a Diophantine reference lattice (the geodesic logic of creators and nuclear fusion).We prove that these two classes are distinguished by the geometric structure of the medium in which the system evolves. In an unstructured, three-dimensionalmedium, the extractive solution dominates: free energy flows from macro-scales to micro-scales via the direct Kolmogorov cascade. In a quasi-two-dimensional medium with Diophantine geometry, the creative solution dominates: free energy flows frommicro-scales to macro-scales via the inverse Kraichnan cascade. The transition between the two regimes is governed by the KAM winding index K, with the universal critical threshold K = 28 marking the Lorenz homoclinic explosion.We develop the theory in two parts. Part I formulates the geometric action principle: the Lagrangian of a complex system is not a function of energy alonebut of the Diophantine distance between the system’s trajectory and a reference lattice of creative configurations. The principle of least action selects the trajectory that maximises long-term geometric proximity to this lattice. We show that theclassical Lagrangian of mechanics, the Burgers Lagrangian of financial flows, andthe institutional Lagrangian of social systems are all special cases of this geometric action.Part II applies the geometric action principle to thermonuclear fusion. We prove the isomorphism between the Hasegawa–Mima equation for drift waves in magnetised plasma and the 2D Euler equation for incompressible fluids. The plasmapossesses two quadratic invariants—energy and generalised enstrophy—and therefore satisfies the Fjørtoft condition for an inverse energy cascade. The Coulomb barrier, which prevents spontaneous fusion, is identified as the effective viscosity of the plasma medium. A tokamak with a Diophantine magnetic field configuration reduces the effective dimensionality of the plasma from 3D to quasi-2D, activating the inverse cascade. Kinetic energy from thermal fluctuations is concentrated into coherent vortical structures that confine ions long enough for quantum tunnelling through the Coulomb barrier. Fusion becomes not a probabilistic event but a deterministic consequence of the inverse cascade.The theory makes a specific, testable prediction: the confinement time in a tokamak should exhibit a sharp improvement when the rotational transform (the safety factor q) is tuned to a Diophantine value—specifically, to powers of the golden ratio ϕ. Existing empirical data from JET, DIII-D, and ASDEX Upgrade are consistent with this prediction. The ITER baseline scenario, with q95 ≈ 3, lies near a rational resonance and may be suboptimal; a Diophantine scenario with q95 ≈ ϕ^2 ≈ 2.618 or ϕ^3 ≈ 4.236 is predicted to yield superior confinement.The theory contains no adjustable parameters. The golden ratio is optimal by virtue of its Diophantine properties. The critical threshold K = 28 is the Lorenzhomoclinic explosion. The fusion cross-section enhancement factor is determined by the Bruno distance between the plasma frequency and the Diophantine lattice. The geometric action principle unifies thermonuclear fusion, financial market stability,and institutional immunity as manifestations of a single mathematical structure: the inverse cascade in a Diophantine medium.