MASTER, the core infrastructure underlying every domain on this platform, deliberately contains no Pokémon logic, no crypto logic, no watch logic, and no logic specific to any other domain. This isn't a limitation — it's the entire reason adding new domains keeps getting easier rather than harder.

What actually lives where

Domain-specific knowledge — how to interpret a PSA population report, how to read a governance proposal, how to evaluate a GitHub repository's momentum — lives entirely inside each module. MASTER understands signals only: a structured event with a source, a category, an entity, a signal type, and a score. Once a module translates its domain-specific observation into that structure, MASTER treats it identically to every other signal from any other domain.

This is why a new domain like Watches doesn't require any change to MASTER's architecture. It requires building domain-specific collectors and detectors, writing one normaliser class, and registering the module — all reusing infrastructure that's already proven across three live domains.

What this buys the platform over time

  • Every new domain increases the value of HPI automatically, because the historical precedent library is domain-agnostic and immediately applicable to any newly-onboarded entity
  • A module can be removed entirely without damaging MASTER — historical signals from that domain remain preserved even if the producing module is retired
  • Integration effort for each new domain decreases over time, since the framework, not the domain knowledge, is what's being reused

The architecture as a competitive structure, not just an engineering choice

A competitor attempting to replicate this platform would need to replicate not just one domain's detection logic, but the entire compounding signal history behind it — and that history cannot be rebuilt retroactively, regardless of how quickly the underlying software architecture itself could theoretically be copied.