The core argument: if the Socio-Cognitive Model of Trust applies to human-AI interactions (and by its own logic it does — it models the trustor's beliefs, not the trustee's inner life), and if micro-level trust aggregates into macro-level social capital (0.98 correlation with GDP in prior empirical work), then agent design properties are inputs to a social capital generating mechanism that nobody is currently measuring.