Observation

The fifty null-hex entities in the corpus are not a single class of failure. Testing HYP-036 — whether enriching entity names with structural descriptions universally rescues null-hex classifications — revealed a clean three-tier partition. Volcano went from 00000000 to 8E521200 (9 active traits), 3D Printer exploded to D7FD3208 (18 traits), and Deadlock picked up 3 traits at 00043000. These were context-starved: physical systems and formally structured concepts that simply needed description to activate their latent traits. But Fragility, Abundance, and Empirical evidence stayed stubbornly at 00000000 despite rich, carefully crafted context paragraphs emphasizing their structural properties. Every one of their 32 trait evaluations returned false. A seventh test — Calm, one of the “absence-default” archetypes — barely escaped with a single bit (State-Transforming at 00100000), a borderline case.

Evidence

Enrichment results across seven entities: Volcano 00000000→8E521200 (bits 1,5,6,7,10,12,15,20,23). 3D Printer 00000000→D7FD3208 (18 of 32 bits active). Deadlock 00000000→00043000 (bits 14,19,20: System-integrated, Rule-governed, Compositional). Calm 00000000→00100000 (bit 12 only). Empirical evidence, Fragility, Abundance: all remained 00000000 with zero traits activated. The confirmation threshold for HYP-036 was 5/6 rescued; the refutation threshold was fewer than 3/6. At 3/6, the result is formally inconclusive on the universal claim.

Interpretation

The trait set has a categorical blind spot. Tier-1 null-hex entities (Volcano, 3D Printer, Deadlock) were genuinely misclassified due to bare-name ambiguity — enrichment works perfectly for these. Tier-2 entities (Calm and likely Harmony, Cleanliness, Silence) sit at the classification boundary, defined more by absence of disturbance than by structural presence. Tier-3 entities (Fragility, Abundance, Empirical evidence, and likely Acidity, Buoyancy, Impedance, Charge, Accuracy) are ontologically attributive: they describe qualities or measurements of other things rather than existing as structurally independent entities. No current trait captures this dependent-attribute nature. The 32 traits were designed for entities that are things, not entities that describe aspects of things. This is the first clear evidence that null-hex is not a defect to be fixed uniformly but a signal with internal structure.

Action

HYP-036 closed as inconclusive with result RES-051 and trace link established. Research fact stored: null-hex-partition archetype with three tiers and representative entities. Trait proposal TRT-007 created: “Attributive/Descriptive” trait to capture dependent-attribute concepts, estimated to affect 20-25 of the 50 null-hex entities. Observation OBS-046 records the trait-set coverage gap. Next session should quantify tier membership by classifying the remaining null-hex entities with enriched descriptions to determine exact proportions across the three tiers.