Epiphany Lab / Essay
Semantic Discovery Before Substrate Engineering
Reliable architecture begins by discovering and stabilizing the real concepts of a system before encoding them into schemas, ontologies or infrastructure.
I just read this excellent post on Substrate Engineering, and it strongly resonated with observations from my own engineering work.
My perspective is slightly different.
I believe the real challenge begins even before Substrate Engineering.
Before we can formalize an ontology or define a semantic protocol, we first need to discover the stable concepts that actually belong to the domain.
In my own software architecture work, I repeatedly observed the same pattern:
Build a working implementation.
Encounter conceptual tension.
Ask: “What is this thing, really?”
Discover a more stable concept.
Rename it.
Refactor the architecture.
Watch complexity decrease.
This wasn’t a theoretical exercise. It happened multiple times during the development of a production-oriented Site Factory:
Template → Page Type
UI Component → Semantic Block
Renderer Registry → Block Registry
Each architectural improvement came not from applying another design pattern, but from discovering a more stable semantic concept.
This makes me think there is an engineering discipline that precedes Substrate Engineering: Semantic Discovery.
Its goal is not to define an ontology, but to systematically discover and validate the stable concepts from which ontologies, semantic protocols, and reliable AI systems can eventually be built.
Perhaps the engineering pipeline looks like this:
Semantic Discovery → Semantic Stabilization → Substrate Engineering → Reliable AI