Observation

Popcount does not track abstraction. Tested across four domains — biology, computing, physics, economics — with four-level hierarchies running from maximally concrete entities to their most abstract generalizations, the prediction that popcount would monotonically decrease with rising abstraction failed in every domain. Zero out of four showed the predicted trend. But what appeared instead is more revealing: an inverted-U curve where intermediate-level entities activate the most traits, while both the most primitive components and the most abstract concepts sit at low popcount.

Evidence

Biology: red blood cell (9 bits) → organ (10) → organism (14) → life (8). Computing: transistor (12) → logic gate (10) → CPU (16) → computation (11). Physics: photon (3) → electromagnetic radiation (4) → energy (4) → entropy (2). Economics: coin (11) → currency (16) → money (18) → value (2).

In every domain the intermediate entities peak. Organism at 14 bits subsumes all 9 traits of red blood cell and adds five system-level properties: Structural, Outputs Effect, Processes Signals/Logic, Signalling, Temporal. Money at 18 bits carries 10 traits that coin lacks — System-Integrated, System-Essential, Signalling, Rule-Governed, Compositional, Normative, Temporal, Digital/Virtual, Politicised, Ethically Significant — while coin retains three physical traits money does not: Physical Object, Observable, Physical Medium.

The most abstract endpoints (life at 8, computation at 11, entropy at 2, value at 2) shed physical-layer traits while retaining functional-processual ones. Life keeps Active, Processes Signals/Logic, State-Transforming, Functionally Autonomous, System-Essential, Compositional, Temporal — but loses every physical and structural trait organism had. CPU loses nine physical and institutional traits when abstracted to computation, but retains all seven functional traits. Photon and entropy share zero traits — complete disjunction between concrete-physical and abstract-relational.

Corpus-wide: the highest-popcount entities (Ableton Live at 23, AI-powered public surveillance at 22, UN Climate Conference at 19) are all complex multi-domain concepts sitting at intersections of physical, social, institutional, and technological categories. The lowest non-null entities (1-3 bits) are verb senses, mathematical abstractions, and grammatical terms.

Interpretation

UHT popcount measures ontological dimensionality — how many distinct property categories an entity participates in — not where it sits on a concrete-abstract spectrum. Intermediate-level entities peak because they occupy organizational levels where multiple ontological layers intersect: an organism is simultaneously a physical thing, a biological structure, a signal processor, and an autonomous system. Abstraction does not uniformly strip traits; it selectively sheds physical-layer traits (Physical Object, Observable, Powered, Physical Medium) while preserving functional-processual ones (State-Transforming, Processes Signals/Logic, System-Essential). This selective shedding explains why abstract concepts have low popcount — they retain only the functional skeleton — and why primitive components also have relatively low popcount — they participate in fewer ontological categories than the systems built from them.

Action

HYP-071 refuted and closed. New observation OBS-STRUCTURALFINDINGS-062 records the inverted-U finding. New hypothesis HYP-ACTIVEHYPOTHESES-074 proposes selective trait shedding: physical traits should be lost at 2x the rate of functional traits under abstraction, testable across 8+ pairs in 4+ domains. Next calibration session should test HYP-074 directly, and also test HYP-070 (trait-shedding under context enrichment) which may now be interpretable as a specific case of this broader pattern.