Structural Diagnostics of the Retrospective–Prospective Ratio in Constrained Generative Systems: Computational Derivation and Clinical Profiles

H3 derives the diagnostic pixel: the minimal operative unit at which the ratio between retrospective and prospective classificatory residuals becomes structurally readable and actionable. The pixel is the ordered pair (d, a) where d is a diagnostic act discriminating the dominant border and a is a calibrative act applied to that border. The observable rH = η / ηₚ measures the operative balance between retrospective compression and prospective discrimination under the same finite limit L. Four computational propositions are derived from a parametric sweep of the Ω-field PDE (36 runs across three phases): (1) no calibrative intervention crosses the biological threshold L*; (2) the relation between intervention intensity and stability is U-shaped; (3) high-capacity systems self-stabilise and are destabilised by external intervention; (4) trajectory compression is not equivalent to regime stabilisation. Six clinical profiles (ADHD, ADHD under methylphenidate, erudite, athlete, elderly meditator, sleep) are positioned on the pre-derived computational map. The ADHD–MPH comparison confirms that pharmacological intervention on the biological border alone does not restore the retrospective–prospective ratio. Sleep is identified as Q15 (border variety), a third regime distinct from discrimination (Q20) and form (Q7). This is the first paper in the SDL programme in which the computational instrument generates the diagnostic structure before empirical data populate it. The domain satisfies the three CGS axioms, operates under trajectory-consumptive dynamics (T3) in regime R2, and admits Q-family weighting F > G > E. Part of the Sub-Limit Dynamics / Constrained Generative Systems programme. H-series paper 3.