Add stuff about LLM-generated code to CONTRIBUTING.MD and PR template

[dfsek]

I made the language in the PR template very final, but specified in CONTRIBUTING.MD that using LLMs for minutiae is acceptable.

[solonovamax]

tbh, I would also probably add smth outside of just the pull requests section and to the bug report/enhancement sections saying smth along the lines of

Do not use an LLM or another "AI" tool to help you diagnose the errors you are experiencing and taking what it says as the truth.
Often times, the "fixes" they will give you will be outright wrong.
If english is not your native language, you don't need to use machine generated translations in order to communicate with us, as we will do our best to attempt to help you.

also, as a side note, imo that pr template (and also the issue template) really needs to get completely redone bc it's rather long, though that's a problem for another time

[neutrino84]

Question, I recently built from the ground up a PaperMC platform plugin for Terra v7 (Folia and PaperMC compliant, 1.21+ Minecraft support) And its working pretty well -- I built it with the help of AI/LLMs obviously, AND I very much understand where you're coming from, with all the AI generated SLOP going around right now. I went over every single line of this plugin, and am currently debugging it (every line has been looked at by human eyes)

What is the policy in a case like this, should I not make the PR?

[duplexsystem]

Question, I recently built from the ground up a PaperMC platform plugin for Terra v7 (Folia and PaperMC compliant, 1.21+ Minecraft support) And its working pretty well -- I built it with the help of AI/LLMs obviously, AND I very much understand where you're coming from, with all the AI generated SLOP going around right now. I went over every single line of this plugin, and am currently debugging it (every line has been looked at by human eyes)

What is the policy in a case like this, should I not make the PR?

What you have provided is an example of what we very much want to allow. As long as you're submitting in good faith then it's all good.

[neutrino84]

Excellent, I plan to PR it in the coming weeks, will need experienced eyes to help perfect it

[dfsek]

Chiming in too - it sounds like what you've done would fall under this requirement in CONTRIBUTING.MD:

Using an LLM to automate small, tedious tasks (regex and other fiddly things like it) is acceptable, but submitting a low-effort, completely LLM-generated PR will result in a ban.

We don't care if you use an LLM to generate fiddly repetitive bits (things like platform delegates which are very repetitive and I'd guess what you used an LLM for would fall under this), as long as you wrote and understand the meat of it. These rules are specifically designed such that when someone types "fix the issue <pasted issue>" into ChatGPT and then lobs the result at us without so much as looking at it (#548, PolyhedralDev/BiomeTool#12), we do not have to engage and can just close it immediately.

[dfsek]

Another metric is, "if you're just going to copy/paste our code review feedback into ChatGPT, we don't want to bother reviewing your code"