You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There's been a large influx of so-called "vibe" coded mods using AI for some/all of the development. Since many of these mods are being created by people without development/debugging experience they may have serious issues and should most likely be, in my opinion, avoided.
Issue
Since there is a large influx of these mods it is difficult to sort through the DB without an option to broadly filter out AI coded mods.
Ideas for solutions
Rule to mark mods as AI generated, and an option to [show/hide] AI generated mods.
Ability to save/create a filter that excludes certain contributors (Similar to current inclusive contributor sorting)
An opinion of a modder that has both mods with AI assistance and without.
I don't think the technical side is the hard part here. Adding metadata and filters is relatively straightforward. The difficult part is enforcement and definition boundaries.
With current tooling, reliably determining whether a mod was "AI-generated" is not trivial, especially when many workflows involve partial AI assistance (Copilot, generated snippets, asset touchups, documentation, etc.), except those blatant obvious. Any strict enforcement policy would likely depend heavily on self-reporting and manual moderation since no automated tool is accurate enough.
The difficult part is enforcement and definition boundaries.
I agree, which is why as a fallback in the event that a collective moderation method cannot or will not be implemented, user-created disallow-lists or filters would help. Currently there is no way to hide anything at all from any search page without explicit allow-listing contributors or tags, and with the amount of contributors and tags, it would be an unrealistic goal for anyone.
Even outside of the scope of AI, let's say for religious reasons you wanted to filter out anything tagged "alcohol." Well, you are going to have to add every other tag to an allow-list, and now you are also going to miss anything without tags too and if something is tagged as alcohol and another tag, welp...
Let's say someone doesn't self-moderate/tag correctly, personally harassed you or a friend, or uploads mods that you know cause performance issues... it would be nice to hide that contributor. Like with tags, you'll have to instead add every person other than them to your allow-list, and again if that contributor simply adds someone else to their team it works around this workaround.
Any strict enforcement policy would likely depend heavily on self-reporting
Which just makes this a non-starter for most developers. The mere mention of ai gets one demonised beyond comprehension by people who don't know the difference between claude/codex and replit/cursor. So the only reliable marker would be the ai images, which is what initially started this whole obsession. Mods can be broken without ai, so mod just being broken is not a definitive check, regardless of what many commenters say.
I don't see any way for this to not be a nightmare to moderate with all the abuse and self-reporting requirements
Reacted by Liviu Cocianu, NiclAss and Tobias Dudziak
I have to agree with KnewOne. The tag system we already had failed miserably due to low interest in interacting with the system, and tag abuse with people just spamming ai tags. The only way i can see this kind of work is by ignoring users, but that would also mean you have to encounter that modder at least once then ignore him. So far, what we have is unreliable at best. I have no other idea to ignore mods and modders.
Activity
Zaldaryon commentedon May 7, 2026
An opinion of a modder that has both mods with AI assistance and without.
I don't think the technical side is the hard part here. Adding metadata and filters is relatively straightforward. The difficult part is enforcement and definition boundaries.
With current tooling, reliably determining whether a mod was "AI-generated" is not trivial, especially when many workflows involve partial AI assistance (Copilot, generated snippets, asset touchups, documentation, etc.), except those blatant obvious. Any strict enforcement policy would likely depend heavily on self-reporting and manual moderation since no automated tool is accurate enough.
QtKaitly commentedon May 7, 2026
I agree, which is why as a fallback in the event that a collective moderation method cannot or will not be implemented, user-created disallow-lists or filters would help. Currently there is no way to hide anything at all from any search page without explicit allow-listing contributors or tags, and with the amount of contributors and tags, it would be an unrealistic goal for anyone.
Even outside of the scope of AI, let's say for religious reasons you wanted to filter out anything tagged "alcohol." Well, you are going to have to add every other tag to an allow-list, and now you are also going to miss anything without tags too and if something is tagged as alcohol and another tag, welp...
Let's say someone doesn't self-moderate/tag correctly, personally harassed you or a friend, or uploads mods that you know cause performance issues... it would be nice to hide that contributor. Like with tags, you'll have to instead add every person other than them to your allow-list, and again if that contributor simply adds someone else to their team it works around this workaround.
KnewOne commentedon May 12, 2026
Which just makes this a non-starter for most developers. The mere mention of ai gets one demonised beyond comprehension by people who don't know the difference between claude/codex and replit/cursor. So the only reliable marker would be the ai images, which is what initially started this whole obsession. Mods can be broken without ai, so mod just being broken is not a definitive check, regardless of what many commenters say.
I don't see any way for this to not be a nightmare to moderate with all the abuse and self-reporting requirements
Shadeloid commentedon May 12, 2026
I have to agree with KnewOne. The tag system we already had failed miserably due to low interest in interacting with the system, and tag abuse with people just spamming ai tags. The only way i can see this kind of work is by ignoring users, but that would also mean you have to encounter that modder at least once then ignore him. So far, what we have is unreliable at best. I have no other idea to ignore mods and modders.
Sunconure11 commentedon Jun 20, 2026
Is there nothing that can be done in the way of some form of approval by the hosts of the site itself?
Shadeloid commentedon Jun 22, 2026
Check the most recent devlog on the site.