arXiv paper: Agentic AI rewrites C to idiomatic Rust via ship-of-Theseus method
A structured workflow turns unsafe C into safe, idiomatic Rust—one incremental step at a time.
Vasily A. Sartakov's new arXiv paper (2607.28835) tackles a core systems-engineering problem: how to migrate mature C codebases to Rust without breaking them. C underpins OS kernels, embedded platforms, and network infrastructure, but its manual memory management leads to undefined behavior, pointer misuse, and lifetime bugs. Rust's compile-time ownership model eliminates these failure modes, yet direct translation fails because implicit layout assumptions, aliasing patterns, and UB must be reconstructed first.
Sartakov's methodology treats migration as a ship-of-Theseus process: first, generate a semantics-preserving, non-idiomatic Rust baseline that compiles and behaves identically to the original C. Then, agentic AI incrementally rewrites the baseline into idiomatic Rust—swapping unsafe patterns for safe abstractions—while validating each step through compilation and behavioral testing. The paper applies the approach to iodine (12.5k SLOC), demonstrating that reliable C-to-Rust migration is a structured, verifiable workflow rather than a single blanket translation. For engineers managing legacy systems, this offers a practical roadmap to Rust's memory safety without the risk of big-bang rewrites.
- Two-phase approach: semantics-preserving Rust baseline first, then agentic AI-driven incremental rewrites to idiomatic Rust.
- Each transformation step is validated via compilation and behavioral testing, ensuring correctness throughout the migration.
- Demonstrated on the iodine codebase (12.5k SLOC), proving feasibility on real-world systems software.
Why It Matters
Gives teams a practical, verified path to replace unsafe C with memory-safe Rust—reducing vulnerabilities without risky rewrites.