Rars: a Rust RAR implementation, mostly written by LLMs
bitplane.net
May 13, 2026
9 min read
🔥🔥🔥🔥🔥
49/100
Summary
A RAR compressor for every version of RAR was developed using OpenAI Codex 5.5 and Claude Opus 4.7 in approximately five weeks, costing around £40 in tokens. The resulting code consists of 55,000 lines and is not optimized for speed.
Key Takeaways
- A RAR compressor was developed in 5 weeks using OpenAI Codex 5.5 and Claude Opus 4.7, costing approximately £40 in tokens.
- The project involved extensive reverse engineering, utilizing various free decompressor sources and tools like Ghidra and DOSBox-x to document the RAR file format.
- The author faced challenges with AI models generating incorrect code and had to implement manual interventions to maintain code quality and prevent issues.
- The development process highlighted the importance of monitoring AI outputs to avoid propagation of errors and the need for frequent refactoring.
Community Sentiment
Positives
- The project represents an ambitious use of AI in software development, showcasing the potential for LLMs to assist in creating functional tools.
- The open-source nature of the implementation allows for community-driven improvements and transparency in the development process.
Concerns
- Concerns about copyright and the legality of using machine-generated code highlight significant ethical implications in AI-generated software.
- The implementation is criticized for being slow and of lower quality compared to established tools like WinRAR, raising doubts about its practical utility.
- There is skepticism about the accuracy and correctness of the generated code, which could undermine trust in AI-assisted development.
