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)/p^k, 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.