RBA (Refraction-Based Architecture) is a way of using AI to develop complex work through structured refraction rather than one-shot generation, so that meaning, architecture, and artifact coherence are preserved as the work evolves.
It is a governed human-AI collaborative process in which intent is progressively refracted across multiple artifact layers while preserving coherence, allowing lower-level work to inform and refactor higher-level understanding when reality reveals hidden constraints, opportunities, or structural mismatches.
(Collaboration was intentionally deferred because it semantically suggests two sentient beings working together, which risks anthropomorphizing AI in ways that are conceptually misleading. Refraction is more appropriate because, like light passing through a medium and changing direction or separating into visible components, it captures how human intent is perceived, transformed, decomposed, and delegated across different layers of artifacts, constraints, and AI-mediated processes.) |