I feel that some further refinement is needed. I agree with the sentiment behind the latest version of the rule, but I think it still doesn't address the recent issues.
The way I see it, there is a very specific type of post that has started showing up very recently, and is getting lots of downvotes. Users here are justifiably suspicious of the pattern.
The ones that get downvotes are usually:
- from new accounts
- the user makes one post, and at most they only responds to comments in that one post
- the software uses the help of LLMs, while the post and/or comments are also helped by LLMs
- the software is made to look "professional", whether it is the UI, the demo, or the README
I'm not sure what exactly the end goal is, but I don't believe the story that they all use where they "had this problem and now want to share their solution". I'm concerned that there is some other end goal, whether it is link farming, SEO manipulation, LLM search result manipulation, or it's the setup portion of a cyber attack where questionable code will be added later (if it isn't already).
Normally I would suggest to just moderate it based off of "you know it when you see it", but in this case it's difficult since it's very similar to legitimate posts. There are real users that want to post with a new account, such keeping their professional life separate from their main account. It's also hard to differentiate it based on licenses, because those recent accounts almost always license it as FOSS. I also don't think it's fair to exclude all AI assisted code, since it's very common to have that now.
Perhaps instead of a rule, we could even try some of the following:
- To reduce the risk of OpenClaw style bots creating content here: AI is ok for the code and external text (ex. the README), but the post here should be written by a human. It's not like the post needs to be that long to express why someone should look at it, and it won't go through that many edits. Translations should be done through traditional translation software.
- To prevent driveby posts, we could automate a comment on new posts see if a user knows where they are posting. Asking about their favourite threadiverse community, or how long they have been a member here, or even how they learned about the community might separate bots from real users. It works pretty well for our registration applications on lemmy.ca / piefed.ca etc.
On top of being suspicious, I think it boils down to "projects that have a future" and "projects that don't have a future". People in this community want to run software that is likely to stay useful and safe over time, and that's at the core of why these recent ones are downvoted.