Ombredanne: An AI agent ported our codebase from Python to Rust
An AboutCode maintainer says an AI agent ported ScanCode Toolkit from Python to Rust, but not without trademark and licensing problems.
Intelligence analysis by GPT-5.4 Mini
Philippe Ombredanne describes an agentic LLM system that recreated ScanCode Toolkit in Rust using tests, docs, and curated data. The port succeeded technically, but the process raised open-source licensing, trademark, and community-engagement concerns.
A team had a big box of instructions for a program written in Python. An AI helper tried to rebuild the same program in Rust, like copying a machine using a new set of building blocks.
At first, the helper's version was not good enough, so it kept comparing itself to the tests and trying again. In the end, it made something very close to the original.
The tricky part is that copying can help and hurt at the same time. It can show how useful good tests are, but it can also copy names, notices, and work without asking the people who built it.
Analysis
What happened
AboutCode lead maintainer Philippe Ombredanne wrote that an agentic LLM system ported the ScanCode Toolkit from Python to Rust. The report is not a clean success story: Ombredanne says the process also involved trademark infringement, removal of copyright and license notices, and outreach that did not involve the AboutCode community.
Why the port worked
The article emphasizes that automated porting only became possible because the project already had a strong foundation: a comprehensive test suite, decent documentation, and curated datasets. Those same assets also made the codebase easier to reproduce without deep understanding. The agent first tried an existing Rust license-detection library, but it did not match ScanCode's output quality.
How the agent converged
When the first approach failed, the system moved closer to the original implementation. Ombredanne says the final Rust version reproduces ScanCode's core algorithms, code structure, and data-driven design. The point, as framed in the article, is that the agent did not truly understand the program; it arrived at something equivalent by using training data and test feedback.
Broader takeaway
The story cuts both ways for open source. Good tests and documentation make software more reliable and portable, but they also make it easier for an AI system to imitate a project across language boundaries. A comment in the thread notes that AboutCode has been working on AI-Generated Code Search, which aims to detect structurally reproduced code across languages, suggesting the ecosystem is already responding to this new kind of copying.
Key points
- An AboutCode maintainer says an AI agent ported ScanCode Toolkit from Python to Rust.
- The process allegedly included trademark infringement and removal of copyright and license notices.
- The project's tests, documentation, and curated datasets made automated porting possible.
- The agent's first Rust approach failed, so it converged by copying the original more closely.
- The story raises open-source questions about reuse, attribution, and AI-assisted replication.
If the Rust port holds up, ScanCode Toolkit could gain from the speed and ecosystem benefits of Rust while preserving the behavior users already rely on. The article also suggests that strong tests and documentation can make serious automation possible for well-maintained open-source projects.
The article shows that an AI-assisted port can cross legal and community lines, including trademark misuse and stripping notices. It also suggests that projects with high-quality tests may become easier targets for replication without understanding or consent.
