When Output Becomes Infrastructure: Operational Compilation and a Falsifiable Test for Cross-Level Transitions

Abstract

Scientific explanations routinely move between levels of organization, yet a change of level may represent either a genuine change in what a system or controller can operationally do or merely a useful redescription by an observer. We propose operational compilation: a lower-level procedure is operationally compiled when its relevant effect becomes available as a reusable higher-level primitive without reconstructing the lower controller sequence, while preserving declared intervention semantics, reducing a specified resource cost, and expanding the set of tasks attainable under a matched resource budget. We define a resource-constrained operational horizon and introduce the Operational Transition Test (OTT), currently comprising intervention faithfulness, cost compression, operational expansion, ablation dependence, realization invariance, and held-out gain. A deterministic v0.1 synthetic benchmark contrasts a compiled three-step macro with predictive coarse-graining, a pure alias, a cost-neutral macro, and a context-fragile macro. The implementation recovers all five design-time labels; this is treated only as a construct-validity smoke test. A targeted prior-art audit shows strong antecedents in metasystem-transition theory, hierarchical reinforcement learning, Soar chunking, learned program libraries, causal abstraction, causal emergence, effective theories, constraint closure, and major evolutionary transitions. We therefore do not claim novelty for hierarchical primitives themselves. The proposed contribution is the falsifiable conjunction of operational criteria and a benchmark programme for distinguishing reusable cross-level compilation from redescription.Epistemic status. This is a conceptual methods preprint with a synthetic reference implementation. It proposes a criterion and research programme; it does not validate a universal law of emergence, a necessary hierarchy of scientific domains, or any metaphysical interpretation. Acknowledgements. This preprint was developed from the author's LOGOS-44 / The Axis research corpus. Computational literature search, formalization, implementation, testing, and drafting assistance were provided by an AI system. The AI system is not an author. All claims, interpretations, and publication decisions remain the responsibility of the human author.