Assumptions galore in that, maybe ask my meaning before reacting to your inference.
1 Like
You’re not alone. I would also appreciate knowing if LLMs were used for a project or not.
Aside from code quality, there are also enough ethical issues surrounding LLMs, and people should at least be able to make an informed decision about the software they install and the projects they support. Are there ways to use LLMs to create good software, with enough care? Probably. Are there ethical ways to use LLMs given the current state of the industry and politics? I think that’s much more difficult.
But before we start another lengthy debate here, please: This is only about transparency, so people can make their own decisions before installing something.
3 Likes
Please explain me where I am wrong then?
Or: what was your meaning
This is definitely the bottom line of what I want. I just want upfront transparency so people know what they’re getting into. In terms of ethicality: as somebody who has spent months poisoning LLM trainers and has done everything to keep the tech out of my life and all of my codebases, it would be inherently hypocritical if I decided it was totally okay to use it to make apps myself or use apps 100% made by a LLM.
It isn’t too hard to do due diligence thankfully and my policy is currently checking contributors and if I see a lot of commits made by a LLM I simply click off and don’t use it. All I am suggesting here is a small self-directed label to let people know but there are some decent points against it in this thread. I understand everything about how it saves time and makes it accessible to more people but I don’t think it’s unreasonable to want to stay away from it. 
5 Likes
Hard pass on that, will wait and see how the dust settles around this one.
Clearly some folks have some investment in the outcome. Said me piece, offered some thoughts.
Not interested in forcing my opinion on others or having others push theirs in me.
I am out of this one. Will wait for the dust to settle.
Just adding the thought that if this is a concern, you don’t need to worry about plugins/themes, because you can’t use Discourse at all anymore. We’re all using AI heavily these days to develop it.
8 Likes
I know, and in an ideal world I’d not use any LLM-coded software at all. Unfortunately the world is not ideal and I don’t always have a choice. All the more reason to preserve that choice where it’s still possible.
6 Likes
Just to make it perfectly clear here
I develop Discourse using my agents, I have not written more than a handful of lines of code in about 10 months I have written significant features this year and repaired numerous flaws
If you are looking for a project with a ban on agentic coding I am afraid you are going to need to look elsewhere
7 Likes
I didn’t ask for a ban, and neither did the OP. The topic is “Require LLM-generated themes & plugins to be tagged as such”.
2 Likes
Why do themes need it when we get a free pass for core?
1 Like
I suppose the reasoning is: we can’t get around core being made-by-AI, but we can have a choice for the remaining pieces?
I do understand what’s being and asked (and why) but I think it’s more efficient than to do what’s been suggested earlier and make an AI-free tag, to be used by those who feel the distinction is important.
4 Likes
Then the question is what constitutes as AI-free? I do ask Ask Discourse or Claude on specific things when developing stuff, but I don’t vibe-code the whole project. So to what extent would such a name cover the use of AI? Literally 0%, AI chats for clarification, or full generation? I think this distinction should be made clearly.
4 Likes
To me personally, AI-free would mean:
“I wrote all the code by hand and if I get asked about any random line of code in this project, I know what it does – or at the very least I can figure it out after a quick read if I need to remind myself.”
But yes, the definition would have to be clear and I’m sure not everyone will have the same.
Then again, if it’s about ethical/environmental concerns, then even asking AI questions is part of the problem…
Can of worms.
4 Likes
A simple start would be “no LLM-generated code in this project”. You probably can’t make this 100% exact, it’s mostly a statement of intent. I often see parallels with veganism: if you take it to an extreme, eventually you’ll end up with contradictions and impossible choices. But that’s not the point. The point is minimizing needless harm.
I see a lot of harm coming from LLM use: to long-term code quality, to our environment, to our society, to our own mental health. When I look at a software project that I’m going to depend on, I want to know if the developer tries to avoid that, or if they’re all in. If they consider the issue important enough to label their project as “AI-free”, that’s already a helpful statement for me.
2 Likes
It is unfortunate it has to be like this because I still have strong ethical concerns with how AI companies basically just steal code from the Internet, destroy books, and otherwise run abusive crawlers that steal everything without respecting the host server (even when you look past the unnecessary DRAM crisis!) all to make it out first in the LLM race that nobody wanted. I’ve already stopped using a lot of apps because the quality of code has degraded so much, like VS Code, which became slow and memory-inefficient and generally buggy. Formbricks’ upgrade packages stopped working, my music player got slow and bloated (once adding features becomes so easy you begin to lose project scope and add everything), and stopped playing music half the time.
But the issue is I like Discourse so I don’t want to stop using it. It’s not like an IDE or music player where I have no attachment or even my OS because I already hated Windows before I stopped using it in 2024. Unless the software gets extremely buggy or slow as a result of LLMs it’s unlikely I’ll decide to stop using it.
I’m not looking to be entirely “black or white” about this whole situation but I can’t get over the ethical or privacy concerns that come attached with using a LLM.
2 Likes
I share some of your observations, and as someone who values privacy, I understand that to avoid using AI, you’d have to stop using the clearnet.
Regardless of the technology itself, it inherently has commercial ties with those who technically enable the use of Internet: Cloudflare, Amazon, Google, Meta, Microsoft.
We are facing a digital era shift, which will eventually continue to deepen both its light and its darkness. As long as we can still browse without an ID, I understand that we can choose the middle path, that is, to use AI consciously.
And I think that’s what Discourse does, or I do on my own even though I’m part of the group you mentioned that designs plugins or components without knowing how to do it. I do it privately, I test them, and I’m always learning, because I work on an alternative, decentralized, niche project that hasn’t received funding in years and I haven’t found anyone willing to do it for less than thousands of dollars.
There are different nuances to analyze, and I don’t believe a simple tag solves the issue at its root or center. In my life philosophy, doing is better than saying, so I’d like to know in what way and in relation to this, you really act in the end.
3 Likes
Not trying to be nitpicky, but do you think the use of AI is technically an issue, in general, for well-etablished software made by competent developers? And I’m only talking about the apps efficiency stability, security.
I’ve seen plenty of apps go slow, bloated, malicious or broken way before LLMs existed.
I use many applications in my everyday life and so far I have not noticed any technical issue allegedly caused by the use of AI in their code. And I’m quite sure nowadays 95% if no more of those apps are made with AI to some extent.
If I use an app made by incompetent devs, surely I will have a bad time, AI or not.
Like anyone I’ve seen dramas like “look at this famous app made by a famous dev which broke because of AI”, I believe it happened (and will happen again).
I also think this is rare occurence, I think talented devs didn’t wait AI for their project to fail at some point for whatever reason.
I’m concerned (and sometimes disgusted) by AIs for many reasons including some of those you stated, but not really by “my music app is now bad because of AI” (and again, not saying that doesn’t happen).
6 Likes
It depends on self-regulation. Having AI do menial tasks such as writing tests is not really the same as building entire apps (or plugins, in this case) and publishing them as your own. I’m sure if I woke up tomorrow and decided “what was I thinking, LLMs are super ethical and can really help my workflow!” then yes I would be able to get over burnout and develop stuff faster like some other, more experienced developers do. But personally, that takes all the fun out of programming and development. I enjoy programming as a hobby and a job—offloading it to a LLM takes any sense of enjoyment out of the equation because you aren’t really coding; you’re just typing prompts all day.
That has always been a problem (windows 10 released in 2015), AI becoming popular and “good” at programming only accelerated it because it made it so easy to add bloat to your app. Let’s say I had a navigation app and suddenly I decided one day I wanted to add a built-in FLAC player with tons of integrations. Obviously, this doesn’t need to exist and is far out of scope for a navigation app and would be a waste of my or my developers time. But instead if I just go to Claude and tell it “add a music player to my app with x, y, and z”, then I can just add it and ship it. What was a “this might be cool but is useless and a waste of time” just turned into a real feature that my imaginary users now all have which adds to the app’s footprint, making it slower and heavier on RAM/storage. This is just imaginary at the end of the day but the scenario absolutely isn’t and is happening to many apps.
With or without LLMs, it’s the responsibility of the developer to make a good app. A careless developer, AI or not, will either make a bad app or make a decent one but get carried away and ruin it. Let it be known I’m not strictly against LLMs the way I am with image and video generation models, I just believe they have a place and that is not coding for the better majority of people who cannot responsibly make use of the technology.
3 Likes
I haven’t caught up with the conversation, and am too brain-dead to look up the proper references, but I’ve come upon a couple of « labelling systems » or « disclosure frameworks » that might be worth looking into if they aren’t on « your radars ». The question today is less about if AI was used or not, and more how. There’s a huge difference between giving an LLM a paragraph of instructions to spit out a plugin and then tell it to « test and fix errors », and retaining overall direction and supervision of development and using the LLM to do its work in a controlled manner. For me, asking people who provide code to self-disclose would be something to consider.
(PS: I’m more familiar with these issues in the field of writing, rather than coding, of course.)
3 Likes
Yeah, I share your opinion on this, it’s a concern 
1 Like