Observation

Forty-eight entities share the hex code 00000080. Sixty-three share 00000000. Together with five other high-collision codes — 0000A080 (38 entities), 00008080 (26), 00002000 (9), 0000C080 (7), 00000280 (7) — approximately 198 entities out of 1975 map to just seven hex values. Ten percent of the corpus collapsed into seven points in a 32-bit space that could theoretically accommodate four billion distinct codes.

The entities trapped in these gravity wells are a specific kind: word-sense disambiguations (“Better (thoroughly or completely)”), social relationships (“Unrequited Love,” “Romantic Friendship”), aesthetic judgments (“Beauty,” “Ugliness,” “Harmony”), and abstract qualities (“Amount,” “efficacy,” “verification”). Meanwhile, five algorithm-related entities each received a unique hex code. Not a single collision.

Evidence

Population counts by hex code: 00000000 holds 63 entities, 00000080 holds 48, 0000A080 holds 38, 00008080 holds 26. The 00000080 code encodes a single active trait — Social Construct — and attracts concepts as diverse as “Actual Infinity,” “Amateur Botany,” “baseline management,” and “Conflict (state of opposition).” The 0000A080 code adds two more bits but still attracts concepts with minimal ontological structure in UHT terms: “Agreement (thing agreed to),” “Average (sports ratio),” “Economic (of economics adj).”

By contrast, algorithm entities spread across hex space: “Computational Algorithm” (40A53709), “Sorting Algorithm” (00E03B00), “algorithm” (40A03B09), “consensus algorithm” (40A5FB89), “algorithmic hiring decisions” (51B67B9D). Five entities, five codes, zero collisions.

Interpretation

The trait set was designed to discriminate engineered, physical, and institutional systems — concepts with clear boundaries, designed functions, regulatory contexts, and measurable properties. For these, the 32 bits spread efficiently. But for concepts whose primary nature is social, qualitative, evaluative, or linguistically polysemous, the trait set offers almost nothing. “Social Construct” becomes a catch-all bit that absorbs everything the other traits cannot describe, producing a single-bit attractor that pulls in 48 unrelated concepts.

This is not a classification error — it is a trait-set resolution limit. The system accurately reports that these 48 entities share the property of being socially constructed and lack every other trait in the taxonomy. The problem is that “socially constructed” describes a vast, internally varied category that UHT cannot subdivide. Prior sessions documented this piecemeal — aesthetics trait poverty (session 63), emotion hex collisions (session 56) — but the population-level scale was unknown until now.

Action

Created OBS-STRUCTURALFINDINGS-034 documenting the gravity-well phenomenon with quantitative population data. Derived HYP-ACTIVEHYPOTHESES-026, which tests whether collision rates differ by at least 4x between technical and social domain categories (threshold: <10% technical, >40% social). Trace link established OBS-034 → HYP-026. Stored a FUNCTIONAL_ARCHETYPE research fact for the gravity-well finding. The next CALIBRATION session should test HYP-026 by sampling 50+ entities per domain category and computing per-domain collision rates. If confirmed, this would motivate a trait proposal for social/qualitative discriminating dimensions — traits like “evaluative,” “relational,” “performative” that could subdivide the 00000080 attractor basin.