Torc

The World's First AI-Native Programming Language

Computation is connection.

Torc is a programming language, ecosystem, and build system designed from first principles for AI authorship and machine execution. Programs are directed semantic graphs with rich type annotations, computation is expressed as dataflow rather than control flow, formal verification proofs are carried as part of the program structure, and executables are materialized through constraint satisfaction against declarative hardware and platform models.

Status

This project is in early development. See IMPLEMENTATION_PLAN.md for the current roadmap and progress.

Specification

The complete language specification is in the spec/ directory:

Document	Description
00 - Index	Overview and table of contents
01 - Executive Summary	What Torc is and why
02 - Design Philosophy	The seven principles
03 - Language Specification	TGL core: nodes, edges, regions, contracts
04 - Type System	Proof-carrying types
05 - Ecosystem	The torc CLI tool
06 - Materialization Engine	Constraint-solving build system
07 - Target Platforms	Declarative platform models
08 - Registry	Package system and module hosting
09 - Observability	Human inspection layer
10 - Verification	Formal verification framework
11 - Interoperability	FFI and language bridges
12 - Example Application	Complete FOC motor controller example

Architecture

Torc is implemented in Rust. The major components are:

• torc-core - Graph data structures, type system, contracts
• torc-trc - Binary graph serialization format (.trc)
• torc-verify - Verification framework and SMT solver integration
• torc-materialize - Materialization engine (graph-to-executable pipeline)
• torc-targets - Target platform model definitions and parsing
• torc-ffi - Foreign function interface bridge generation
• torc-observe - Human observability layer and projection views
• torc-registry - Package registry client
• torc - Unified CLI tool

Building

cargo build

License

Apache-2.0