lin-0239

20.5.1 A GLP-1 health-research atlas

The Prometheus GLP-1 study retained eleven heterogeneous documents rather than forcing them into a single evidential genre. The corpus included efficacy studies, safety and adverse-event material, access and adherence, cardiometabolic outcomes, delivery technologies, and adjacent mechanistic work. Its construction produced 3,376 extracted events, 11 causal episodes, 191 local predictive-state contexts, 190 restriction checks, 186 compatible gluing overlaps, and four tense overlaps. Of the restrictions, 149 were compatible and 41 divergent.

Those numbers describe an atlas, not a vote. Evidence for core weight-loss efficacy can glue across suitable clinical contexts while gastrointestinal mediation, hair loss, hormonal effects, reward pathways, and adjacent animal mechanisms remain local, qualified, or obstructed. The appropriate output is therefore not the global sentence “GLP-1 works.” It is a navigable collection of claims whose intervention, population, outcome, evidence level, and rebuttals remain visible.

¹1. In a health foundry, preserving a contradiction can be safer than resolving it. Two studies may differ because their populations, interventions, outcomes, or evidence levels should never have been glued. ↩

This case illustrates why non-compositionality is not merely another loss. Each divergent restriction carries a type and a proposed next observation. Scalar averaging would erase the distinction between a genuine contradiction and a legitimate regime boundary.