Structural-coherence pipeline and defect-index calculators v1.1

Creative insight shows four robust regularities: deliberate search on a resistant problem produces defective output; problems then resolve during low-control activity such as walking or showering; the triggering stimulus looks arbitrary yet is specific; and the solution arrives whole rather than assembled step by step. We model these regularities as properties of a single phase transition. The latent material of an unsolved problem is represented as a weighted semantic network, and the current cognitive state as an order-parameter field on that network. Two control parameters, saturation and constraint, locate the system on a phase diagram; insight is the nucleation of a coherent phase when saturation is high and constraint is released. The trigger is a structured endogenous perturbation whose efficacy equals the cross-correlation between the pattern the trigger induces on the network and the spectral structure of the latent material. A scaling law converts the poor quality of forced output into a quantitative prediction: the density of frozen-in defects in a produced text should decay as a power of the release time, with exponent d_s/4, where d_s is the measured spectral dimension of the substrate network. The prediction carries one measured input, three stated commitments, and no fitted parameters. The model also issues a set of pre-transition signatures: critical slowing, flickering, and involuntary return. An n-of-1 pilot on a timestamped corpus with contemporaneously recorded insight events, run under analyst masking, returned a mixed-to-null directional read, which we report under either-way commitments. All quantitative claims are deposited before data contact.