I think the vibecoding tag has gotten out of hand
38 points by anex9d
38 points by anex9d
I was reading this post and noticed the vibecoding had been applied. I’m not sure if this was applied by the lobste.rs team or by the author, but after reading the post it’s pretty clear that the only vibecoding involved is mentioned in an appendix separate from the actual meat of the post, which was well written and the product of sole-human effort and care. This tag feels like a black mark on what are otherwise interesting posts and has the potential to filter out genuinely human content. In other words, I think the tag does not capture the nuances of actual LLM use and is perhaps being applied too broadly.
edit: it now seems the tag has been removed on that post. I still feel this should be a point of discussion
Anyone thinking to post here should also read https://lobste.rs/s/qgy6ak/can_we_stop_tagging_every_thing_as , which dedicates 135 comments to the same request for discussion, 2 months ago.
This is valuable context but I think it risks acting as a soft dismissal of current / ongoing grievances (not to say that this is your intention, I think you’re just being helpful). The point is that this is a continual issue for some (such as myself), not something that is addressed by visiting the thread.
The thing is that re-litigating the issue constantly is not going to help, either. Take it from someone for whom vibecoding on lobste.rs has been an big issue for much more than two months ;-)
You're getting pushback because you opened with a broad statement supported by one example, which turned out to be the very opposite of what you claimed. The submitter included the tag, and the community suggested its removal.
If you go through the modlog, you can see which stories have had the tag suggested and applied. That's step 1. Step 2 is to try to ascertain whether the tag was suggested in error of the guidelines. This can be subjective on your part.
Step 3 is to present this as a case, ideally with some sort of quantitative evaluation, i.e. "of latest N submissions, I believe X% were incorrectly tagged by the community", and concrete examples.
Right now your argument is based on... vibes (sorry couldn't resist).
The tag was applied by the submitter, then removed by user suggestions, according to the moderation log. Once user suggestions modify a post, no further modifications can be suggested without moderator intervention.
Someone suggested that maybe a "trigger warning" type of tag (or "content warning") similar to the media type tags would be a better match for how people tend to apply the vibecoding tag, and I think I mostly agree with that. My ruby/rails chops are not good enough to propose that kind of change personally in an expedient manner, though.
The tag was applied by the submitter,
Moreover, the submitter is the author of the post, so talking about a "black mark" is really not adequate.
I think it is important to highlight that even if OPs linked post might not fit the problem stated, there is (in my opinion) still a prevalent issue despite the misleading post example.
Mostly saying this so that the thoughts are not disregarded simply because of an erroneous assumption in the opening post. OP states that he read that post and started thinking about [vibecoding] (the tag), not that the problem is exclusively for posts exactly like that one.
Also, even when the tag is applied by the author — does it really reflect the contents (besides the imo poor definition of the tag itself)?
I don't disagree: it's worth highlighting and discussing.
Also, even when the tag is applied by the author — does it really reflect the contents (besides the imo poor definition of the tag itself)?
That's an interesting question. I am not an expert in cognitive psychology. I've read a lot about it, and thought a lot about it, but I'm at best an educated and curious person. That said, I'd estimate that for people who are attempting to participate in good faith (as opposed to attempting to market something or disrupt the site) nearly all of them assign tags based on what they think is helpful to the community first and strictly accurate about the content second. Most of the time that overlaps a great deal, but I think this specific tag is more challenging in that regard.
I've had similar feelings during the past few weeks, where [vibecoding] is applied to submissions that discuss [ai], even if they are not necessarily making use of it, nor are in the category "churning out code with help from their favorite LLM".
For me the expectation when I see [vibecoding] rarely match what I actually read on the other side.
From my point-of-view, [vibecoding] implies that a project was primarily/significantly driven forward by interaction/usage of AI — and I suspect more than me would appreciate a clearer distinction between "discussing something tangent to AI" vs.. well, vibecoding[1].
From my point-of-view, [vibecoding] implies that a project was primarily/significantly driven forward by interaction/usage of AI
From the tag definitions: https://lobste.rs/tags
ai: Developing artificial intelligence, machine learning. Tag AI usage only with
vibecoding.
vibecoding: Using AI/LLM, coding tools. Don't also tag with
ai.
Based on this, my interpretation of the tag's intent is different from yours: if an article is about how to use AI, or other general topics related to LLM use, tag it with vibecoding. There's not a lot of controversy or room for interpretation there. "Project that was vibe coded" or "blog post that's clearly AI generated" feel like gray areas that aren't covered by any existing tag but in practice get tagged vibecoding pretty consistently. I think there would be value in splitting the tag.
I think a lot of the resistance to the vibecoding tag comes from the fact that it literally uses the term "vibecoding" which is somewhat pejorative. Personally I have zero issue with that and am just looking at the intent, and I think this tag is too broad.
Based on this, my interpretation of the tag's intent is different from yours:
The tag definition is part of the problem in my opinion.
Everywhere™ else "vibecoding" is closer to my personal definition than what is currently the tag's definition. Sure, the definition could be anything and we could accept that "on lobsters tag X means Y"— but are we really satisfied with a tag which one can truly understand only after reading the definition?
Read the above with a grain of salt, I am not saying that my interpretation is the only correct one; but I would assume I am not alone given how the phrase/term is used elsewhere.
I think there would be value in splitting the tag.
Very much agreed!
The title is overly dramatic. Paging through /t/vibecoding to see what's been getting tagged in the past few weeks I don't see too many (or any?) false positives. If anything vibecoding seems to be one of the more carefully curated tags since many people are sensitive to it.
Compared to a few months ago, I actually feel like it has calmed down. I do not like LLM discussions / articles but I have seen less and less of actually good submissions that even slightly mention LLM not being tagged as vibecoding. Maybe it is because I filter the vibecoding tag, but even on the articles I post, it's not very often that it is suggested.
However, I'm having my meta meta moment: Meta discussions have been invading the yearly top. The literal top 5 are only meta discussions and it's getting kind of annoying to read about the same issues each time.
I get that them not being handled is frustrating, I truly do. I wish this PR was actually merged and not just closed as it would've set clearer rules on the site than there are now.
Wish we could discuss more on what makes us return to this site often, articles about computing that might improve our programs or deepen our understanding.
I also attempted to make a PR, based on the discussion, https://github.com/lobsters/lobsters/pull/2023
In the months since, I've been told that there's a discussion about it coming Real Soon Now(tm). I personally don't care if there are techncial enforcement measures; a statement of values would be a good start.
There hasn't even been removal of content for being ai generated, as far as I can tell. There's been no communication on policy. The only thing has been a quiet expansion of the useless flagging that doesn't actually lead to any moderator action.
I think at this point, I don't expect any action on the topic; the only venue people on this site have is yelling louder about the kind of site that they want to exist.
Yeah, I think so too. I don't filter the vibecoding tag and yeah, I think meta is more annoying now (edit: not the meta tag in general; the meta discussions about LLMs)- I just see three vibecoding stories on the front page right now.
Actually, I find #lobsters on Libera more annoying wrt. to LLMs- I actually parted a few days ago because it was very tiring.
(I find both sides of the LLM discussion nearly equally exhausting, BTW.)
Personally, I greatly dislike the fact that the community uses the tag of vibecoding to mean “any programming with an llm”. To me, vibecoding has a very clear colloquial meaning of “coding only with an llm, often where the author barely knows how to code, or they know how to code but they’re putting in 0% human effort”.
Ie, vibecoding implies “I don’t care about this, this is all the machine” and that simply does not cover every use of an llm to write code with.
You can disagree with my characterization I guess but I think my description of how the word is used is more accurate then how we’re using it here.
Agreed, LLM-generated articles should be deleted on sight and projects consisting of primarily LLM-generated code should be deleted on sight. There shouldn't be a tag, just utter rejection of slop.
I would rename the tag to 'AI Programming' or 'AI Assisted' to remove the bias of the term 'vibecoding' which is normally reserved for using AI to code without understanding or care.
A lot of people here are refusing to accept that programming with AI is a sliding spectrum of human effort and understanding.
I took a look at the modlog looking for recent posts with [vibecoding], and most of the tag changes are in the following format:
Action: changed tags from
$tagto "$tagvibecoding"Reason: Automatically changed from user suggestions
Not to wear a tin-foil hat but... is it perhaps a fairly low number of users responsible for this almost mass‑retagging behavior?
It is oddly similar to certain users in #lobsters who take every opportunity possible to disregard something because it's "100% vibecoded", or arbitrary metrics trying to disprove something because it "feels like LLM" (like overily eager focus on usage of emdash), etc.
Perhaps we as a community should bare responsibility, if many of us do not agree with the tag usage — perhaps we should step up and show our opinions through flags.
Edit:
<whatever thing would allow us as a community to say "hey, this added tag is not ideal>"
It doesn't take many users to get a tag or title change to take effect. Last time I looked, the quorum was two, though I haven't been paying attention lately and I'm aware that work was in progress to have different quorums for specific tags.
Thanks for the info!
I kinda suspect this is a severe case of "those who yell the loudest"; if it only takes two votes to change the tag that feels a little optimistic in terms of how well those two represent the rest of the community.
Perhaps an alternative would be that votes-required is somehow based on the number of upvotes of said post, but.. that's probably more tricky to get right in practice than I can envision currently.
Yeah, this tag does seem to break the general model. If two users say a thing deserves an "emacs" tag, it almost certainly does. There's considerably less consensus on how the vibecoding tag should be applied, as far as I've observed.
votes-required is somehow based on the number of upvotes of said post
I don't think it's an issue for already big posts, it is mostly harmful for just posted ones, as false vibecoding will hide it from over 600 users, but acceptable threshold also varies between users
If you filter a tag, you're trusting the community to do the right thing with regards to the application of that tag. If you filter "vibecoding" and something that is clearly related to GenAI in software development, that submission is more likely to get a suggested tag change by people who notice it's gone through the filter.
I think adding to the quorum requirements per tag has been implemented, but I am honestly not sure. Sometimes a quick Changelog to stuff like this would be good to have.
Have you considered that the people supporting the use of the tag as it is used currently (I am part of these people) may be larger than you think?
And as such, people railing against its uses are the vocal minority.
That would line up well with every survey data we have about LLM sentiment in the software community. And it would align with the FOSS maintainers community use of LLM (minimal, 5% average with an heavy skewed distribution). Which is a community I would expect to be a good proxy for people on this site.
perhaps we should step up and show our opinions through flags.
I don't quite understand what you mean by this. Who exactly would you like to flag when tag suggestions are anonymous?
poor use of term, I meant "perhaps we as community need to <somehow> make an effort in.. well, making sure we do not apply the tag". I don't know how this would work in practice though, as far as I understand once a tag line has been edited once you can't do it again?
I apologize for the usage of "flag", I should have used some other term.
Ah, I see.
In my opinion, if this is indeed a small group of users, then I think it's not a community, but a moderation / rules issue. With how loose the rules are around AI (compared to practically everything else on the site), it's a lot harder to argue about the good/bad faith-edness of any single action related to the tag / content falling under it. If the rules are tightened and this small group is still continuing their behavior (despite it now falling under misbehavior) then moderating them becomes a very straightforward story.
If it isn't a minority, but rather either the majority or the majority is fine either way, then there is not much to do.
I personally used to filter the tag, until I realized it occasionally filters out stuff I would be otherwise interested in, so nowadays I just use it as a warning sign, that works most of the time, but not all.
I personally only suggest the tag "vibecoding" when "ai" has been incorrectly applied.
This is basic housekeeping. As per the tag guidelines, "ai" is used for developing and implementing AI, including GenAI.
Anything downstream from that is, currently, "vibecoding".
It can be hard for new or non-active submitters to know the difference, which is why the community curates the tags.
I don't suggest the tag when the content itself is the result of GenAI. That's a case for the "spam" flag.