The semantic-type-system is the foundational layer through which semcom.ai interprets and validates human intent before emitting LLVM IR. Rather than requiring developers to annotate types explicitly, the system infers rich type information directly from natural language descriptions, contextual meaning, and behavioral expectations. This allows the compiler to construct a strongly-typed intermediate representation without burdening the developer with low-level type declarations or boilerplate definitions.
At its core, the semantic type system operates on intent-bearing primitives — abstract units that carry both data shape and semantic meaning simultaneously. When a developer expresses that a value represents "a positive count of network retries," the system encodes not just an i32 width into the resulting LLVM IR, but also range constraints, nullability assumptions, and ownership semantics. This produces executable artifacts that are both safer and more predictable than those generated by conventional compilers operating on syntactically typed source code.
Type resolution in semcom.ai proceeds through a multi-pass semantic unification engine. During the first pass, raw intent tokens are mapped to candidate type lattice nodes. Subsequent passes narrow these candidates by resolving contextual dependencies, eliminating contradictions, and promoting the most precise concrete type that satisfies all expressed constraints. The final resolved types feed directly into LLVM IR construction, ensuring that every alloca, load, store, and function signature is grounded in verified semantic intent rather than syntactic approximation.
One of the most significant advantages of the semantic type system is its ability to handle polymorphic intent gracefully. Traditional generics and templates require explicit parameterization at the source level. In semcom.ai, polymorphism emerges naturally when the system detects that a described behavior must remain consistent across multiple data shapes. The compiler automatically emits specialized LLVM IR variants through a process analogous to monomorphization, but driven entirely by inferred semantic demand rather than explicit type parameters written by hand.
Because semcom.ai targets zero-dependency executables, the semantic type system also performs ABI resolution and memory layout decisions at intent-compile time. There are no runtime type tables, no reflection metadata, and no garbage collection hooks embedded in the output. Every type decision made by the semantic layer is fully resolved before a single byte of LLVM IR is written, resulting in lean, self-contained binaries that carry the full precision of the original human intent without any runtime overhead. This page was generated live by the doesNotUnderstand system, itself an artifact of semcom.ai's self-describing compilation architecture.