Directional Admissibility as a Structural Regulator in Long-Horizon Adaptive Systems: A Core Architectural Principle of PETRONUS

This work introduces Cybernetics of the 2.5-order as a distinct architectural class for long-horizon adaptive systems. Classical cybernetics of the first order treats regulation as external control over a system. Second-order cybernetics internalizes the observer, recognizing that observation itself perturbs system dynamics. Third-order cybernetics extends this reflexivity further, embedding meta-observation into causal decision processes and allowing systems to act upon their own self-descriptions. The PETRONUS / ΔE / ONTOΣ framework deliberately does not enter third-order cybernetics. Instead, it defines an intermediate regime — 2.5-order cybernetics — in which reflexivity is acknowledged, modeled, and structurally accounted for, yet explicitly prevented from becoming causally operative. In this architecture, the system does not optimize its own coherence, identity, or viability. These properties are observed non-causally as structural witnesses rather than control variables. Directionality, admissibility, and regime transitions are regulated through constraints on evolution, not through self-referential optimization. This is not a limitation of expressive power. It is a deliberate structural decision. Embedding long-horizon viability observables into the causal decision surface collapses observation into control, subjecting identity, coherence, and integrity metrics to optimization pressure, distortion, and silent degradation. The refusal to cross into third-order cybernetics is therefore a condition for structural honesty and long-horizon viability rather than an absence of reflexive sophistication. Cybernetics 2.5 thus occupies a precise and necessary position:- it models observation without weaponizing it,- it permits reflexivity without self-optimization,- and it preserves the possibility of witnessing long-horizon degradation without interfering with action.