I would be shocked if all of the projects on your list genuinely adhere to requirements as stringent as "no contributor has ever googled an issue and accepted a working AI suggested solution". How would they even enforce that?

I'm starting to suspect the way I asked my question wasn't clear and you (and BaroqueInMind) aren't answering the question I meant to ask. Were you thinking my question involved LLM-generated code? I intended for my hypothetical to involve only wholely-human-authored code written with knowledge originally gained from LLM output.

(If you could read my first edit on my response to BaroqueInMind in this thread, I hope that'll clear up what I was trying to ask.)

(Also, I hate calling anything gotten from an LLM "knowledge" because implies LLMs are reliable sources of accurate information. I could simply call it "information", but that makes it sound like I'm talking about copying output from the LLM into the contribution itself when I actually mean information ingested into the human's brain and considered in the process of authoring code themselves with their own human brain. Please excuse the lack of better ways to language it.)

If you are saying that in order to qualify for inclusion on noailist.org a project must forbid the use of knowledge gained from generative AIs, I of course want to know that as well, but yeah. I just want to make sure we're not talking past each other.

The last part where you mentioned Google AI suggestion is still literally using AI.

Well, yeah, but I'm trying to get a better understanding of where the line is exactly. Not risking using knowledge you don't specifically remember where you learned for fear it might have been from an LLM? Not using LLMs for the duration of the process of constructing your contribution? Does it require changing your search engine from Google to something without an AI overview, or is just scrolling past the AI overview without reading it enough? Do the project core maintainers need to go research the background of the contributor for signs that they may use AI for contributions to projects not related to yours, or is it enough to review the contribution to your project and ask the contributor to promise there was no AI used? If a contributor finds a bug or vulnerability using LLMs, confirms the issue themself without using generative AI, and files a ticket without using generative AI, am I allowed to fix the bug, or must I reject that contribution because the information contained therein would not (yet) have been found had LLMs not been used? Can I use dependencies with LLM-generated code in them? Can I use dependencies which do not have LLM-generated code but which have their own dependencies with LLM-generated code in them? What about dependencies that don't have any explicit policy on LLM-generated code that I don't know that they contain any LLM-generated code? Can I accept code contributions authored in a text editor with a generous generative-AI-generated code policy (but with no features for users to integrate with LLMs from the editor)? What about with the use of a compiler that has LLM-generated code in it? What about on an operating system kernel with a permissive generative AI policy such as Linux? Most importantly, is the language in my project's contributor policy sufficiently strong to justify inclusion on noailist.org? (From OP's answers to my question, it sounds like maybe the answer is "no".)

If I were to contribute something to another project on noailist.org, I highly doubt it would be an issue for the project to accept it. I don't use LLMs (or Stable Diffusion or Dalle or Suno etc) at all when I have a choice. I use noai.duckduckgo.com for search. My employer requires me to use Jira and Confluence, and both have LLM features. Confluence in particular shows summaries without being asked. I never read them, so I (purposefully) never learn anything from them, so making contributions based on information I'd learned from an LLM wouldn't be an issue. (I'm lucky my employer doesn't impose token usage quotas or anything.) That said, actions I take on a regular basis as part of my job result in the prompting of an LLM along the way without any specific intention to do so on my part. Assuming I'm right that noailist.org wouldn't have any issues with contributions I make under such circumstances, my question is whether all contributors must be similarly strict for my project to be included on noailist.org. If not, I'm wondering what I'd have to change (aside from my employer, of course) to be a noailist.org-approved contributor. (Both so that I can contribute to projects on noailist.org and so I can make informed decisions if/when I get contributions to projects I maintain.)

When I have time, I'll probably go read some of the no-AI policies for projects on noailist.org and see just how strongly they're worded and how my contribution policy would have to change to be similarly strict.

Edit: Oh wait, I think I misread your comment. My example didn't involve any LLM-generated code. (Or at least I meant for it not to.) Just getting knowledge (or at least "information") from an LLM and then writing code themselves (not LLM-generated code) using knowledge they'd acquired from an LLM.

This bit here:

Did a human write that output? No? Oh it was generated by an LLM? Yes? And you used it? Yeah, that’s fucking called using AI.

makes it sound like you thought I was talking about LLM-generated code.

But now that I see you didn't interpret it the way I meant it, I'm wondering if OP might have interpreted it the way you did.

Like, let's say:

• I authored and published a project under a FOSS license.
• I was the only contributor to the project.
• I authored it before ChatGPT existed and haven't contributed to it since, so there's no question that no generative AI was used in its creation.
• After noailist.org came to exist, I added a contributor policy to my project that noailist.org's rules would agree is sufficiently strict to warrant inclusion in noailist.org's list. I submit my project and noailist.org accepts it and includes it in their list.
• Then one day, someone we'll call Sam wants to contribute a new feature to my project that involves calculating the length of the diagonal of a square.
• They go and ask an LLM "how do I calculate the length of the diagonal of a square?"
• The LLM indicates that the length of the diagonal of a square is the length of one of the square's sides times the square root of two (which happens to be correct). The LLM outputs no code (and was not asked/prompted to). Just an answer to the question asked.
• Sam closes the LLM and, without involving any LLMs any further, writes code themself using their own brain that invovles, among other things, calculating the length of the diagonal of a square. One line of the code is float64 squareDiagonalLength = squareSideLength * sqrt(2);, but that line of code itself was not written by an LLM (nor was any other line in the contribution). It was authored by Sam using knowledge they'd learned from an LLM.
• Sam then opens a PR on my project with the code they wrote (including code which calculates the length of the diagonal of a square).
• If I accept/merge the PR after code reviewing it knowing that Sam had asked an LLM how to calculate the length of the diagonal of a square (but also knowing that no LLM-generated code was included in the PR) have I broken noailist.org's rules?

I hope that better illustrates the question I meant to ask.

But whatever the case, there has to be a line somewhere, right? I think of Stallman deciding that it was acceptable when authoring Free Software to do so on a proprietary operating system when no Free Software alternative existed. Much of the first version of the GNU project was authored on proprietary OSs. (With the blessings of RMS.)

Edit 2: Well, it looks like my go_lemmybot project with the exact contribution policy I linked eariler in this comment was accepted. It now appears on noailist.org. :shrug: I must have just been really imprecise with my earlier comment and everybody thought I was referring to LLM-generated code.