While I can’t speak for anyone else, I’m sticking with the original definition for vibe coding (at least, to the best of my understanding of the original definition), because ai think it still has utility as a separate concept from AI-assisted coding in general.
Under this definition, vibe coding entails offloading both the writing and the reading of code largely or even entirely to language models. If you take the time and effort to thoroughly read and/or write the code yourself, that’s still AI-assisted coding (and still has the potential to qualify as slop), but it isn’t vibe coding.
Reading the discussion here is really enlighting; and its very respectful whats is, sadly, unusal these days. So already thanks for that, thays why Ziggit is a great community.
However, we’re discussing much the pros and cons of LLM usage, but the thing for this forum is not so much about the usefulness of LLM (it surely can be useful in the hands of an experienced dev if ignoring the other downsides). Its about the community.
Personally, I don’t want to spend any part of my sparse time looking on code which has not been written by a member. If I put energy in it, its only for stuff someone else (of flesh and blood) also has put energy in. Code quality doesn’t matter here. I think thats what e.g. @andrewrk or @whitehexagon with his nice furniture comparison wanted to say too.
This might be impossible to be properly checked, but I hope members are fair enuogh to put a valid disclaimer on it, be it “LLM-generated” or “no-LLM”
I might be a bit too policy minded in thinking there can be a UI / workflow way to solve this issue, but I think there is real value is requesting users posting code to describe the LLM. At the end of the day, it is a community of people, so not reporting, mis-reporting, and describing accurately or not sends a message.
I don’t know how the moderator can act on that, maybe asking to self report better.
I like handmade tag, with allowance to what was not handmade e.g. zig product code is mine, some tests are from LLM, LLM was used to update the html shell.
What I’ve enjoyed the most about this forum is how knowledgeable, kind, and helpful everyone is. The few cases where this doesn’t hold true are almost always in discussions related to this topic, so I’m glad that the moderation team is thinking about how we can improve the situation. I don’t feel that ad hominems or attempts to push ones ideology on others should have a place here and I hope that we can be open minded enough to get along despite having very different opinions on certain topics.
Personally I would prefer a level of disclosure that both includes the extent (or lack) of LLM usage as well as the users experience level. Unfortunately I don’t think this is practical or reliable in the most important cases: people being deceptive can simply lie and inexperienced people may not understand how much/little they are relying on an LLM to understand/create.
Ultimately I hope we can find an approach that doesn’t drive away too many of the amazing people that frequent this place regardless of their stance on this topic.
This. In fact, I have been describing myself (within professional circles) as a “Software Luthier”, just to make this point. Not many care, to be honest, but this is important to me too.
This a nice point, because established luthiers typically have helpers in the shop, doing the nitty gritty work under their supervision. And the final result still has the luther’s name on it. It works like this in many crafty areas. Even in the classical art pieces, what’s associated with the artists name is typically work of multiple people, realizing the artist’s vision. I called myself a programmer even though I barely touched any code myself and that was before LLMs existed. My vision was more important and other people were happy to realize it.
I realize this is off topic here, because the real problem is inexperienced people using LLMs and thinking they achieved something, for that it makes sense to label work where LLM were used in any way.
My experience is that generic LLMs are poor translators. They lose the plot after a while. A specialised tool like Google translate does a much better job. It preserves the original intention better, even if it’s translations can be awkward at times.
LLM drafted responses are not wanted. Ziggit is for communication between humans.
Showcase posts are the primary issue.
Self declaration of of LLM usage is a generally accepted as desirable
Concerns exist about where the line is.
Relying on moderation to re-tag posts isn’t really practical
People would trust a “handmade” declaration more than the reverse.
My suggestion (pulling from a number of commenters):
Showcase posts should require a tag from a set of LLM-usage tags. This forces a declaration of some kind.
As the bar is difficult to precisely define. the set of AI tags should be more than two levels (a bad example: “Handmade”, “10% LLM”, “50% LLM”, “90% LLM”). Four is probably a maximum though.
I think if you give people a middle option or two, people are less likely to use “Handmade” for “I’m not zero usage, but I’m below the line I think others care about”. Handmade should be kept for zero usage. That will the people who want to skip all AI happier. They can only look at “Handmade”. The other groups might be more fuzzy and people might low-ball but they’re only of interest to people with some tolerance of LLM usage.
I wouldn’t go this far, context matters here. This isn’t work, where you need to get things done, it’s a community where people share an enthusiasm for craftsmanship and human interaction, it’s normal that a lot of people don’t want to have to deal with AI assisted/generated code, and I think it’s only fair that we try to settle on a good compromise for everyone to be happy
Ultimately, I’m looking for non-fascist and non-fascist-adjacent spaces to invest my time in. If I don’t feel a space meets that, I move on. This is why I largely exist in the real world now, rather than participating online. I am uncomfortable in a community that takes a blasé approach to using tech created by fascists for fascist means and ends.
Let me clearly state that I believe using Codex for any reason is immoral. Using LLMs for any reason is immoral. Using agentic coding for any reason is immoral.
And it has nothing to do with copyright or “how good it is” and everything to do with the cost: mostly to the environment, my environment, my home which data centers are currently ruining and trying to ruin more. In terms of electricity, water, warming, and land use size at the least.
By using these systems you say that your benefit is worth more than my cost, and fundamentally, I am morally and philosophically opposed to externalizing costs on others, and I live and change my life in directions that reduce that. Not increase it.
I don’t feel safe in a space where moderators are okay externalizing costs onto other members of their supposed community.
I’m also in the camp of “as little LLM as possible; ideally zero” and really glad to see I’m not alone here.
One concern I saw from mods above is that it will be difficult to catch any bad actors lying about LLM usage. Two things on that: one, I think it will be impractical for bad actors to keep up any kind of lie forever; it will get discovered eventually. And two, I think enforcement doesn’t need to be perfect. This is a social policy for humans, and humans make mistakes. I bet there are existing uncaught policy violations on the forum today, but it remains a nice place here but the vast majority are handled appropriately. Mods and users all working together to keep the space nice.
We all agree that we want this to be a place for humans, so we don’t expect anyone to meet non-human standards, including mods.
Unfortunately the world has opened up Pandora’s Box and AI is here to stay even though I’m not in favor of it (at least for agentic coding, it already started being slowly implemented at the company I work at professionally and it’s spreading through the industry as well). I’m fine with AI for deep research or to ask it questions about cooking/gardening haha. I’m also not rejecting that AI is very powerful and can be a good ally when it comes to coding as well, but there are other reasons I would prefer to leave the AI out of the code for the most part, but I won’t get into that in this post. However, all we can do now is “ask” for disclosure and transparency on if AI was used and to what extent. Regardless of the policy, people can always lie and say they wrote it all themselves. Lying isn’t something new for humans. Individuals will need to make their own personal decisions on how they judge/perceive the code or actions of others as usual, and perhaps following the rule that if you have nothing to say then don’t say it would be wise here as well.
That quote represents the culture this community understood and built. AI is not going anywhere, it’s a beautiful and scary thing in regard to what’s in store for us, but it is not human. That simple. As new of a member that I am to this community, even I understand that. No disrespect to AI and it’s users, it’s not about bashing AI nor its use, but I believe it deserves it’s own place. Not Ziggit.
What would a library become if all of a sudden they allowed audiobooks to be blasted throughout the space or simply loud conversation? It’ll ruin the culture. The library isn’t forbidding the access to knowledge and a quiet place to study, nor is it forbidding you from listening to your audiobook or enjoying a conversation, but it does have policy that will not change due to its culture. It makes for a more pleasant experience for everyone.
“Hey, this space is for and by humans, we invite you to join “ZiggitAI” != “Get the F*@k out of here you slop”
This thread explicitly establishes considerable leeway for expressing how you feel about and relate to this topic. I encourage other users not to rise to this bait: nothing useful will be accomplished.
In normal Ziggit life, no one is going to be given two opportunities to call fellow users fascists. That is entirely beyond civilized discourse, you’ll need to get your kicks elsewhere.
If you can handle that though, stick around. I genuinely don’t care about your political convictions, they’re none of my concern. Keeping Ziggit from turning into a hellscape, is, and that’s all that lies down that road. We won’t be traveling it.
You caught part of it, there’s another part: we don’t want to. We do want to make it as easy as practical for those who don’t want to see LLM content, not to see it. We do not want to adjudicate disputes between users when one of them used two of the tags and another insists it should have been three.
If I’m going over someone’s Showcase with a fine tooth comb, contemplating whether asking ChatGPT if the right ‘disclosure level’ was chosen will even help, I’m going to feel like a crazy person. We’re all volunteers here.
I think that having a ‘handmade’ category, which is reasonably strict and well-defined, creates precisely none of those problems. It also nudges code which the user thinks is “almost compatible” to state how they used LLMs.
We haven’t had any problems with lack of candor about tool assistance. Broadly speaking, those who are comfortable using the clankers don’t feel guilty about it. Mandatory disclosure in that direction would change the dynamic, that’s my concern.
I haven’t talked with the others, but personally, I’m willing to consider the hypothetical “used the handmade tag improperly” scenario a problem to be addressed. I don’t anticipate much if any of it.