Valea is an AI-native systems programming language.

Programming languages were built for humans. Valea is built for humans and autonomous AI agents.

AI agents are increasingly capable of writing software.

But today's programming languages were not designed for autonomous code generation, repair, and verification.

Valea explores a simple idea:

What would a programming language look like if it were designed for both humans and AI agents from the start?

Valea focuses on five properties:

• Deterministic syntax: one obvious way to express most ideas.
• Explicit semantics: no hidden allocations, no exceptions, no magic behavior.
• Machine-readable diagnostics: compiler output should support both humans and tools.
• Canonical formatting: one stable style.
• Small language surface: fewer edge cases and easier automated repair loops.

Autonomous software development requires languages that are:

• easy to generate
• easy to analyze
• easy to repair
• easy to verify

Typical Valea workflow:

1. Agent receives a goal.
2. Agent writes Valea code.
3. Compiler returns structured diagnostics.
4. Agent applies fixes.
5. Program compiles and runs.

fn add(a: i32, b: i32) -> i32 {
    a + b
}

fn main() -> i32 {
    add(2, 3)
}

valea check examples/type_error.va --json

Example output:

[
  {
    "code": "E001",
    "message": "type mismatch",
    "line": 2,
    "col": 5
  }
]

valea ast examples/ok.va --json

valea fmt examples/ok.va

valea emit-c examples/ok.va

Valea is an early-stage experimental language.

Current milestone highlights:

• Rust MVP compiler
• deterministic lexer/parser/formatter
• JSON diagnostics
• JSON AST export
• simple C backend

See ROADMAP.md for planned milestones.

Build the compiler:

cargo build

Run checks:

cargo test
cargo run -- check examples/ok.va

Valea is designed as a community language experiment.

Good first contribution areas:

• parser improvements
• formatter rules
• JSON diagnostics
• examples
• documentation

Language and toolchain behavior are specified in SPEC.md.

https://asciinema.org/a/834560

MIT