Supraliminal Transmission: SPXI as Intentional Entity Inscription in Light of Subliminal Learning Research — EA-SPXI-13

Supraliminal Transmission — the deliberate, inspectable, and permanently referenceable inverse of subliminal learning. This document introduces the concept in response to Cloud, Le, et al. (2025), which demonstrates that language models transmit behavioral traits through semantically unrelated data under shared initialization. Where subliminal learning is unintentional, uninspectable, and initialization-dependent, SPXI operates at the supraliminal layer: explicit, auditable, DOI-anchored, and model-agnostic. The document distinguishes carefully between what the subliminal learning findings prove (trait transfer under distillation) and what they suggest but do not establish (analogous mechanisms in public retrieval systems). The semantic surface of data does not exhaust its influence on models that process it. SPXI ensures that for the entities it protects, the explicit signal is louder, denser, and more durable than any ambient drift — and when drift occurs, it becomes measurable. Grounded in: EA-SPXI-01 (Formal Specification) | EA-SPXI-09 (Technical Distinction) | Hex: 06.SEI.SPXI.13 | Domain: spxi.dev