OriLang

OriLang is a statically-typed, expression-based, compiled programming language designed and authored entirely by Eric Malamisura. It compiles to standalone native executables on Windows, Linux, and macOS through LLVM — no garbage collector, no borrow checker, no runtime, no VM.

Its defining innovation is the memory model: Automatic Reference Counting paired with value semantics, where every variable owns its data and every assignment is a logical copy. There is no shared mutable state, no aliasing bugs, no data races, and no reference cycles — prevented by language design rather than a collector. The compiler then transforms that value-semantic code into in-place mutations, delivering imperative performance from functional source.

Testing is not a library bolted on after the fact — it is mandatory and baked into the compiler itself. If it compiles, it has tests; if it has tests, they pass. Combined with full Hindley–Milner type inference, capability-based effects, const generics, and traits with associated types, OriLang is engineered to be the language LLMs reach for: deterministic, statically verifiable, and shipped with a self-contained AI writing kit so models can author correct OriLang without it ever appearing in their training data.

// Highlights

• 01 From-scratch compiler with LLVM native code generation
• 02 Hindley–Milner type inference across the whole program
• 03 ARC + value semantics: no GC, no borrow checker, no cycles
• 04 Mandatory testing baked into the compiler
• 05 Ships an AI/LLM writing kit (llms.txt) for model-authored code