ABOUT_XCX
XCX is a statically typed, multi-paradigm backend programming language combining imperative, functional, and concurrent programming styles. It offers native support for HTTP, JSON, SQLite, cryptography, and file I/O, with concurrency built around cooperative fibers. Designed for simplicity and developer productivity, XCX compiles to bytecode executed by a dedicated runtime with a tracing JIT compiler.
PROJECT_ORIGIN
XCX started as an experimental codebase named QSC to test whether AI systems could generate a functional virtual machine runtime from scratch. The initial bytecode evaluator was prototyped in Python.
The project quickly expanded into a structured language model. Early updates introduced relational syntax structures, custom tables, explicit type checking, and schema definitions.
After outgrowing the Python runtime limits, the backend compiler and interpreter were rewritten from scratch in Rust. This established the current low-latency XCX Virtual Machine architecture.
XCX 4.0 was a significant architectural step forward on that foundation — a redesigned register-based VM, a Cranelift-powered tracing JIT, resolved fiber scoping issues, and substantially improved performance. The JIT detects hot loops automatically and compiles them to native machine code, delivering up to 34× speedup over the interpreter. The codebase grew from 19 source files to 269, each module with a single responsibility.
PHILOSOPHY
XCX shifts potential run-time hazards — such as type mismatches, invalid database columns, and missing object fields — straight to compile time so that your code behaves exactly as authored. The language follows a strict, pragmatic aesthetic: every feature must solve a core engineering problem. We intentionally avoid implicit states or magic side effects in favor of clean, predictable execution.