LLM stance

[lanodan]

Came across the comment in choice_dialog_win32.c while reviewing the diffs between v1.4.5 and v1.4.6. Which while I'm pretty fine with it, made me wonder if it's usage would be restricted? Say dealing with proprietary software, and/or maybe other kinds of workarounds due to AI slop in general.

Thanks, and hopefully not too hard/annoying a question

[Akaricchi]

What exactly are you asking about?

If the question is about the legality of distributing that piece of code under the MIT license, then I don't really know, and I doubt anyone does, until/unless there's an actual legal precedent to establish that. I'm assuming that LLM-generated generic code like this is effectively in the public domain, but IANAL. From a pragmatic standpoint, I believe that if what I did there is not legally ok, then a lot of much bigger and more important projects (e.g. the Linux kernel) must be way more legally screwed than we are.

If you're worried about redistributing the code downstream, as long as you're not targeting Windows that file is not even compiled in, and you can literally just delete it.

If you're asking about the project's official stance on LLM-generated code (as the issue title suggests), we don't have an official one. My personal stance is that AI-generated code is generally discouraged, but may be acceptable given the following is true:

• The scope is clearly isolated.
• The LLM is not making any high-level design decisions.
• The code has been reviewed and understood in its entirety by the prompter.

Is this a sufficient answer, or did you have something else in mind?