The φ Theorem: Uniqueness of the Golden Ratio as the Self-Referential Fixed Point in Quantum and Cognitive Systems

We prove that the golden ratio φ = (1+√5) /2 is the unique stable fixed point of self-referential Bayesian dynamics with balanced-dynamics condition. The proof is algebraic: for a system with likelihood L (p) = (1−p) /pᵏ, the posterior map f (p) = 1/ (1+p^ (k−1) ) has |f' (p*) | = (k−1) (1−p*). The condition |f' (p*) | = 1−p* forces k = 2, yielding p² + p − 1 = 0 with solution p* = 1/φ. We show this same equation emerges from free-energy minimization at the golden temperature T* = 1/log₂ (φ), Born-rule self-similarity |β|² = (|α|²) ², phase-damping decoherence at t* = 3·ln (φ) ·T₂, and entropy partition h₀/h₁ = √5 − 1 (all exact). Combined with prior results — embedding distance bands in 37 models (Azpiroz, 2026a), cognitive insight prediction in 4, 482 CRA trials (Azpiroz, 2026b), genius/delusion separation with pre-discovery signal (Azpiroz, 2026c), and four independent derivations from information theory and neuroscience — this establishes 14 convergent routes to p² + p − 1 = 0 across quantum mechanics, information theory, neuroscience, and machine learning. An experimental protocol for trapped-ion verification (IonQ/Quantinuum) is provided.