Ladybird Browser Goes Memory-Safe Porting LibJS to Rust Using Claude & OpenAI.

 

Ladybird Browser Embarks on Rust Migration: Leveraging AI to Port 25,000 Lines of Code

Ladybird, the ambitious new web browser project built entirely from scratch (independent of Blink or WebKit), has officially announced a strategic move to integrate Rust. This decision follows a long search for a "memory-safe" language to enhance the browser’s security and stability. While the team initially considered Swift, they found its support outside the Apple ecosystem insufficient. By choosing Rust the same language utilized by Firefox and Chromium Ladybird aligns itself with industry-leading safety standards.

The "LibJS" Milestone: AI-Powered Porting

The first major component selected for the migration was LibJS, the project’s JavaScript engine. To ensure absolute accuracy, the team relied on the test262 suite a rigorous set of 52,898 tests verifying compliance with ECMA-262, ECMA-402, and ECMA-404 standards.

Lead developer Andreas Kling achieved a remarkable feat by utilizing Claude Code and OpenAI Codex. Through a continuous stream of hundreds of prompts over two weeks, Kling successfully generated 25,000 lines of Rust code. This new code mirrors the original C++ implementation byte-for-byte in output, passing both the test262 suite and Ladybird’s internal tests. Performance remains on par with the original code, marking a significant milestone in AI-assisted software migration.

The Roadmap Ahead

While the current Rust code is largely a direct translation from C++, the team plans to "idiomatically" refine the codebase later. For now, Ladybird will continue to port smaller sub-modules to Rust incrementally, ensuring seamless interoperability with the existing C++ foundation.

Data from Google and Microsoft indicates that 70% of critical security vulnerabilities stem from memory safety issues in C++. Ladybird's early adoption of Rust effectively "closed the door" on these future vulnerabilities, a key selling point for modern browsers.

Andreas Kling's case clearly illustrates that AI doesn't replace programmers, but rather acts as a "supplement." Writing 25,000 lines of complex code like the JS Engine within two weeks by a single person was impossible in the past, but AI made this "tedious and complex task" (repetitive logic mapping) manageable.

The biggest challenge of this project is implementing a Foreign Function Interface (FFI) to allow Rust code to communicate with C++ without performance overhead bottlenecks. If Ladybird succeeds, it will serve as a blueprint for other large open-source projects to gradually migrate their code to more secure languages.

Further information indicates that Swift's problems on Linux stem from its ABI stability and tooling, which are not as stable as Rust's. Choosing Rust, therefore, is about securing the most readily available "community" and "tools" for system programming. 

 

The Director of AI Safety vs. OpenClaw Why Even Experts Can't Always Control AI.

 

Source: Ladybird