Paint.NET enables Linux support via AI-generated Direct2D rewrite

Paint.NET has released an experimental 5.2 alpha build that allows the popular image editor to run on Linux via WINE. The update solves a long-standing technical barrier by using an AI-generated rewrite of the Direct2D graphics API.

To achieve this, developers used the AI model Claude to create a clean-room, reverse-engineered rewrite of Direct2D, contained within the 'PaintDotNet.Windows.Direct2D1.Managed.dll' file. Over a three-week period, the AI generated approximately 180,000 lines of code to bypass the proprietary API hurdles that previously prevented the software from functioning on WINE. The implementation is currently a proof of concept and is described as highly experimental. Users can enable the Linux support via WINE using a specific '/wine' trigger. The software has been tested on Ubuntu in both native Linux and WSL2 environments, running on .NET 10 win-x64 and WINE 11.14-dev.

The Direct2D Hurdle

Paint.NET is a closed-source Windows application that has long relied on Direct2D, a complex Microsoft graphics API. For years, this dependency made the software nearly impossible to port or run on Linux, as WINE's own Direct2D implementation was insufficient for the application's specific requirements. While users previously relied on open-source alternatives like Krita or Pinta, many preferred Paint.NET for its specific user experience and speed in basic image manipulation.

The Rise of 'Vibe Coding'

This release serves as a high-profile example of AI-driven engineering, where a large language model was used to solve a compatibility problem that had persisted for a decade. A Paint.NET developer noted via Hacker News and community forums that Claude made more progress on this issue in three weeks than had been seen in the 12 years since the software began using Direct2D in 2014.

However, the scale of the AI's contribution introduces significant technical risks. The developer explicitly described the implementation as "vibe coded," explaining that the 180,000 lines of code have not been thoroughly reviewed by humans due to their sheer volume. This "trust me bro" style of development highlights the tension between the rapid acceleration of legacy software porting and the stability risks of integrating massive amounts of unverified code into a codebase.

What's Next

Because the build is an alpha proof of concept, it remains buggy and is not intended for production use. It remains to be seen if the developers will perform a full human audit of the AI-generated DLL or if the "vibe coded" approach will be refined through iterative testing. For now, the build stands as a demonstration of how LLMs can be used to reverse-engineer proprietary APIs to unlock cross-platform compatibility.