Morrison's Law of Identity-Preserving Governed Evolution

We introduce Morrison’s Law of Identity-Preserving Governed Evolution: a non- optimizing governing law for adaptive and self-modifying systems in which arbitrary proposed updates are admitted only through an invariant-gated projection onto an admissible manifold. The law separates proposal from permission, enforcing long- horizon stability via identity anchors, entropy/Lyapunov bounds, and idempotent halt semantics under empty admissibility. We formalize the underlying class of Invariant- Gated Evolutionary Dynamics (IGED), define convex and Lyapunov/entropy-gated projectors, prove non-expansive projection stability with respect to an anchor state, and provide failure semantics for hold/tighten/return governance actions. We further outline a distributed extension using sheaf-consistent verification and present an implementation- ready specification pattern (“ARK v71” mapping) illustrating operational validation. Implications include safe self-repairing memory, bounded drift for learning systems, and auditability for safety-critical deployments.