Darling, the open-source project working to run macOS software on Linux, has introduced rules for AI-assisted contributions and reported initial ARM64 support. Its latest development update frames both changes as a response to the scale of recreating Apple's platform with a small volunteer community.
The new policy permits contributors to use language models, but places responsibility on the person opening a pull request. Contributors must understand and be able to explain every submitted line, and maintainers expect questions to be answered by people rather than bots. The policy also incorporates the project's clean-room requirements: implementations should follow documented behavior, preserve macOS layering and clearly isolate undocumented quirks.
Darling says contributors must not use leaked material or restricted Apple source code. Its stated approach also prohibits using disassembly directly to generate matching code. Where local law permits behavioral research, the project describes a separation between documenting observed behavior and implementing it, so the coding agent does not see the disassembly itself.
Maintainers argue that AI changes the economics of a compatibility layer that must reproduce a large and growing collection of Apple frameworks. Darling has historically lacked the contributor base and commercial support available to Wine, while macOS applications represent a smaller target market than Windows programs. Several recent forks have used AI extensively and produced results the project now hopes to consolidate upstream.
The technical update adds early ARM64 support, an important step as contemporary Mac software increasingly targets Apple silicon. Darling plans official support for i386, x86-64 and ARM64. Maintainers cite development of the VibeDarling fork on ARM64 as a source of regular testing for that architecture, though the upstream integration is still described as work in progress rather than finished compatibility.
The short-term goal is to merge changes from the various forks. Darling says agent-driven merge tooling has already corrected bugs in incoming work and in existing project code. A separate experiment aims to run ARM64 applications on x86-64 hosts through Dynarmic, but the maintainer calls that effort speculative and warns users not to depend on it.
Beyond architecture work, planned priorities include stabilizing AppKit, improving interface appearance and adding native management tools such as a configuration utility and log viewer. Those applications would also serve as practical tests of the project's own AppKit implementation.
Darling's policy represents an attempt to use higher-output development tools without abandoning legal and technical discipline. Whether the approach can accelerate broad application compatibility will depend on the quality of upstream merges, sustained testing across the three planned architectures and contributors' ability to verify rather than merely accept generated code.
