How do you PROVE a system will settle down, without solving its equations? Find an ‘energy’ — a quantity that’s zero only at rest and always DECREASES as the system moves. If such a function exists, the system must roll downhill to rest, guaranteed. Lyapunov’s trick (1892) is the bedrock of stability proofs, from robots to rockets. Slide the state and watch the energy fall.
Lyapunov’s direct method proves stability WITHOUT solving the dynamics: find a scalar function V(x) that is positive-definite (V > 0 away from the equilibrium, V = 0 at it) and whose time-derivative along the system’s trajectories is negative (V̇ < 0). Then V acts like an energy that can only decrease, so the state must converge to the equilibrium — it is asymptotically stable. For ẋ = −x with V = x², V̇ = 2x·ẋ = −2x² < 0, so the origin is stable. This one idea underlies stability certificates for control systems, neural nets and optimisation. A fail-loud self-check throws unless V decreases monotonically to zero along the trajectory. ◆ real control theory, node-verified.
A textbook V=x² on ẋ=−x (the exact positive-definite, decreasing-energy certificate); FINDING a Lyapunov function for a general nonlinear system is hard and an art — the if-you-find-one-it-proves-stability logic is exact.