Clojurians Log v2
ai-skeptics 2026-07-30
Right, I don't think mods need to do anything at this point. Having a dedicated channel ends up being the right compromise. I'm still interested in figuring out what norms would make sense to adopt, like as an individual, that would then reflect back on the community. My learnings are: • Human written announcements • Responding to questions genuinely, no-AI • Human readme if possible • Articles/blog posts best if human written • AI disclosure section in readme that describes your level of involvement Just doing these seems like they'd go a long way gaining trust and not asking others to parse through slop.
What is your take on the AI2027 report published by this ex-OpenAI, who is intervieweded in this podcast? Too pessimistic? Realistic? https://podcasts.apple.com/dk/podcast/the-diary-of-a-ceo-with-steven-bartlett/id1291423644?l=da&i=1000776541328
What I mean is, he is saying a 70% chance of the worst-case scenario happening, i.e. human extinction.
I think that interview in itself is deeply unsettling. Even the positive scenarios are quite dystopic in my opinion.
He basically says that he doesn't think his children will ever join the work force, because of how AI is developing.
I find it really hard to take this kind of rhetoric seriously. The whole American AI industry is based on a deeply unsettling ideology, it's basically a cult, with people being absolutely convinced of the most outlandish sci-fi stuff. It's rooted in libertarian anarchist x californian hippy x techno-utopian doublethink. It's also a great screen to cover the capital and power accumulation, and it helps to accompish regulatory capture. It should be treated the same way doomsday cults are treated. See also https://www.realtimetechpocalypse.com/p/skeptical-of-the-tescreal-acronym
This ideology is itself the reason that we are now in this mess. People justifying spending ludicrous amounts of money and resources to trying to create artificial minds, based on the premise that AGI/ASI is inevitable, and if they don't do it then someone less benign will. Meanwhile any concerns like environmental damaged are waved away, cause once they achieve AGI it will figure out how to instantly fix all of that.
But this stuff seems to be the air people breathe in silicon valley. People disagree about the details, but the basic premise of this cult ideology are widely taken for granted.
It's bollocks to me. People working on that stuff are way too deep in some weird echo chamber. It reminds me of the South Park episode where they make fun of Silicon Valley tech folks who are all smelling their own farts and loving the smell lol. There are some possible issues with AI. It's economic reckoning, using it for manned weapons, and dispatching more and more autonomous agents of it. But that thing is so extreme in my opinion, and it's always predicated on this weird idea that AGI is just around the corner. Yet every time they show a new model after hyping it up, it's just the same model slightly better. And I really hate the completely ungrounded belief that it's so much how our own brain works.
I've wondered if it's a symptom of that AI "research" is actually a bit like working on Frankenstein. You just trial and error, a bit blindly, hoping you get some function back that approximates intelligent outputs. And you start to wonder if you might eventually find something that "awakens it", the function will do more than that, intelligence will emerge. And each time you try it seems you can get a bit better and you really start to believe it's just a matter of trying another time.
Thanks, I needed to be reminded of the Silicon Valley ideologies. They had totally slipped my mind when I was thinking about what his motivations for publishing that report are.
I would say that all these bombastic and "scary" predictions are more marketing then something they believe in.
Coming from Anthropic and OpenAI I agree, but from someone no longer employed by companies building the AI’s, I was confused about the motivations. The ideology angle fits well with that. He might of course also be making money on publishing predictions that get press attention, it is unclear to me how his company makes money.
The motivation is simple: attention == $$$. He's using his authority as "ex-OpenAI" to draw attention to himself and his product. It's a tale as old as time.
I want to say that it probably isn't just a "loud minority", and that characterizations like that can stiffle discussion as well. I know I was put off from stating my opinions on that thread initiated by cjohansen after how the mods responded to it.
Let's consider the opposite - suppose it's a majority with a small loud subset, with the rest being silent, not even leaving any reactions anywhere. What do we do with that information? We can hyperbolize and hypothesize all day long, but in the end decisions must be based on something observable.
In other words: if you're unhappy about something, make yourself and the reasons for that unhappiness known. There is little reason to avoid doing that, especially when someone has already made a thread and all it takes to make yourself known is click a single already existing reaction.
> after how the mods responded to it And this is also important. So please do describe what was wrong, if there is anything at all that can be made concrete. I, as a chronically online person that likes to read and write, often re-read such threads and see where I could've phrased something better. I've done it more than once with those threads, and while I'm obviously biased, all I keep seeing is me asking concrete questions and only getting non-answers in return.
Thanks @p-himik for your elaborate message earlier. I've let it sink in and I better appreciate your position now, you feel like you're expected to moderate, in which case you need objective ways to decide how to act in a certain case. I get that now. Up to now however all this repeating of "bUt HoW cAn YoU tEll" really read to me like a load of concern trolling, because I don't actually care that much about moderation, to me that's a secondary concern. I don't want to saddle people with the duty to do policing and vigilante work. That's not a lot of fun to do, especially as a volunteer, and too much of it isn't healthy for a space either. It also comes across as dismissive and argumentative. "We can't objectively tell so what are y'all even asking for??" What I'm personally hoping for are cultural norms. I'd like to see the people who are in charge of this space (and so have a certain power as role models) be clear about the kind of space they want, to create clarity for people so they know what the expectations are of how to behave around here. You can say something like "we expect messages in announcements to be written by humans, and if a project is largely AI-coded it's generally expected that you mention that" (just as an example, y'all have to decide what stance you do or don't want to take). And then see what happens. Let people sort it out from there. Reserve moderation actions for cases of actual abuse and spam.
I don't want to go into that thread and re-read it again, but I think too much was read into cjohansen's post. From the get-go, very early on in the thread, he was characterized like his (and those who might support him) intention is to marginalize people who vibe-code. Very hard to go back to a constructive discussion after that.
And in general, too many 'slippery slope' arguments for my taste. Yes categorizing degree of LLM use is difficult, but that shouldn't be a precondition to start a discussion on this.
My sense is that cultural norms are already pretty in place here, cf. the ways in which people pin emoji reactions on posts they think are not human enough, or the "crickets" that follow in announcements when certain projects are discussed which people generally have a good reason to think are entirely or substantially LLM output. I see plenty 🙏🏻 and 🎉 tagged on posts about libraries that I assume most of the said taggers aren't even using, they are just being pleasant occupants of a digital space
My concern is that the "crickets" will follow for serious projects as well, whether hand-coded or serious vibe-coded projects (meaning with maintainers who are serious in maintaining the project).
That seems unlikely to me, while there are a large number of people in this slack relative to many programming social communities, there are only so many people who are working on serious projects and posting about it. I've been in this workspace for about four years now and I see lot of the same names pretty often.
I really must say that I think this is a text/subtext problem and some of the "loud minority" (a description I think I agree with, fwiw) are posting things where they are either failing to see the fairly easy subtextual interpretation or are simply denying the possibility of its existence
> I don't actually care that much about moderation The longest and arguably the first most impactful thread here started exactly with an ask for moderation: "Can we have some LLM disclosure rules for #C06MAR553?" And this very thread started with the word "curated" in the OP. If rules are desired but enforcement is not necessary, if curation is mentioned but is, in fact, also not necessary - why use those words in the first place? Of, if neither enforcement nor curation is expected from the moderators - why not say so from the get go? We've asked multiple times stuff like "who would be enforcing the rules", with zero clear answers. Of course, different people in these threads want different things - that one of my main gripes that such people still agree with each other, thus making the landscape for the moderators significantly harder to navigate. > It also comes across as dismissive and argumentative. "We can't objectively tell so what are y'all even asking for??" How can the same sentiment be phrased in a way so that dozens of participants all understand it, don't find it dismissive and argumentative, and don't feel that it's patronizing or condescending? > I'd like to see the people who are in charge of this space (and so have a certain power as role models) be clear about the kind of space they want We already have it, it's all in the CoC. It's been mentioned multiples times already, two times in this thread. Sean has explicitly said this very thing, although it might've been in a different thread. > You can say something like "we expect messages in announcements to be written by humans, and if a project is largely AI-coded it's generally expected that you mention that" It's not in the CoC, so we do not expect it. Now if somebody has a problem with CoC and wants to see it changed - please create a new thread with concrete details on what and why. > From the get-go, very early on in the thread, he was characterized like his (and those who might support him) intention is to marginalize people who vibe-code. I just went and re-read the thread, at least the relevant bits. And maybe my brain simply refuses no notice such things anymore, but I didn't notice any such characterization. Could you please point to the exact message? > Yes categorizing degree of LLM use is difficult, but that shouldn't be a precondition to start a discussion on this. Without such a precondition, there would be an action without consideration. What would that achieve? In any case, it's all been addressed in that thread already - the way things currently are, nothing of what's been suggested is actionable by the admin team here.
An example of what I meant @bhlieberman93: let's say someone has been working on a clojure on go dialect for a long time, but they haven't published it yet in #C06MAR553. Normally I would be inclined to be excited by that (as it's probably a passion-project and they are likely to commit time to it) but I might mistakenly consider it as similar to another "vibe-coded in 2 weeks for a particular need project " that someone posted the previous week.
so "crickets"
I suppose that's a risk one runs alas
> but I might mistakenly consider it as similar to another vibe-coded project that someone posted the previous week. This type of generalisation really boils down to I do not like vibe-coding projects doesn't it? I mean, who knows what the quality of that project is (unless it is reviewed somehow)
Crickets would necessitate a union of thought, which doesn't exist. I'll mention it again - this channel is 2% of #C06MAR553. Of course, not everyone who doesn't want to see AI-assisted projects in #C06MAR553 is a participant here. But even so - most definitely not everyone on the server cares so deeply about something being AI-assisted to whatever degree that they would completely ignore some announcements. In other words, there's little reason to be afraid of "crickets".
I mean the very label of "vibe-coded" carries a lot of subtext that most people implicitly regard as being pejorative, does it not?
@alexander.vanelsas that's what I don't agree with, it doesn't boil down to that.
let me know which is the non-pejorative term, I'll use that going forward. I don't think spec-coded has caught on.
> I mean the very label of "vibe-coded" carries a lot of subtext that most people implicitly regard as being pejorative Maybe in this channel, maybe. But definitely not out there in the large.
it does indicate to me that it is more likely to be a throwaway, or less likely to have a committed maintainer.
> let me know which is the non-pejorative term, I'll use that going forward. I don't think spec-coded has caught on. Please keep using "vibe-coded", we don't need a language war here. It's an ambiguous term, but it at least helps with the direction. Anything else by now would only muddy the waters and require additional explanations.
I think there are a lot of people in this community who, regardless of comparisons to other programmers or even "knowledge workers" generally, set a high price upon "ownership", "craft", etc. (I put those words in quotes to emphasize them, I'm not being snarky.) In that context, I'd say calling something "vibe" or "slop" coded is definitely the opposite value
I'll just add to Eugene's point about the CoC—we are also supposed to operate under the Clojure Community Etiquette (on http://clojure.org) that was basically written by Rich and has been referred to several times by the core team folks. I'll also just note that when I opened my inbox this morning, there were two posts from the Clojure/conj team: one inviting me (and I assume everyone on their mailing list) to submit lightning talks about using Clojure and AI together, and one announcing that there will be an afternoon workshop the day before Conj, showcasing Clojure and AI together, based on those submissions. When Conj was in Alexandria a few years ago, there were (I think) two talks in the "unconference" the day before Conj about effective AI usage, and several talks during Conj about effective AI usage—including talks from folks in Nubank, who are using AI in all sorts of ways to support software development. Whilst Rich may not like (some aspects of) how AI is being used, the company that powers Clojure and its main conference are not shying away from AI...
it does indicate to me that it is more likely to be a throwaway, or less likely to have a committed maintainer.In my (limited)
(Disclaimer: I know I have no concrete proposals on enforcement below) #C06MAR553 doesn't seem that bad these days, but there was a while where I felt that some announcements would be more suited to some lower traffic channel like #show-clojurians or whatever it would be called, for people to show off their cool vibecoded stuff or even cool handcoded 'experiments', i.e. things they don't necessarily commit to maintain or think are mature enough to be dependencies or building blocks for other libraries.
Perhaps the creation of #C068E9L5M2Q gave that an outlet? I'm in that channel but have to keep it muted since it is pretty high traffic.
I wonder if we might have had different expectations of things. I never expected curation from admins/moderators. The curation is partly by the name and description of the channel and partly by the members, and certainly by a person asking a question. Just as I wouldn't expect constant talk about datahike being better in a channel on xtdb. I'm not sure I see it any differently than that. This channel clearly isn't the right one to have these kinds of discussions. The overall tone of it sadness, anger, and "debate me". This is what prompted me to ask if anyone here wanted a new channel. I wanted a different vibe. This channel is about ai, and I wanted a channel that wasn't about AI. Early signs are that I seem to have gotten that. I think I understand a bit more of where you are coming from @p-himik, I think initially I took your responses as rules lawyering, but now I see that you are in a difficult position and I'm sorry that channels like this put you there. You would like to do right, you feel you have a role, but you are being asked to do things that you think are beyond what you consider your role to be, and that appears to me from some of the messages upthread to frustrate you https://clojurians.slack.com/archives/C0B7Q4XRM1U/p1785419258110829?thread_ts=1785244731.366779&cid=C0B7Q4XRM1U and https://clojurians.slack.com/archives/C0B7Q4XRM1U/p1785416782575709?thread_ts=1785244731.366779&cid=C0B7Q4XRM1U and https://clojurians.slack.com/archives/C0B7Q4XRM1U/p1785416062275139?thread_ts=1785244731.366779&cid=C0B7Q4XRM1U and https://clojurians.slack.com/archives/C0B7Q4XRM1U/p1785318191537589?thread_ts=1785244731.366779&cid=C0B7Q4XRM1U
the post that appeared like rules lawyering in the other channel is here: https://clojurians.slack.com/archives/C0BLJS63J2Y/p1785272565993109?thread_ts=1785271764.632969&cid=C0BLJS63J2Y
I'm sorry that my wording appeared to put extra burden on you when what I was hoping to find was people who wanted to stop complaining here and wanted to talk about how to build things in another channel with a different vibe
Do you have specific examples of this "anger, sadness, and debate me"?
@bhlieberman93 are you asking me to debate you on whether or not this is a debate me channel?
I think moderation in slack can be quite difficult due to the lack of filtering and blocking (this isn't meant in response to you @bhlieberman93)
No
moving things to another channel is about the only affordance we have to do this, and community curation in that channel (via norm setting etc) is what we have
Yes, moderation can be extremely difficult. We have the CoC, the Community Etiquette document, and a few rules about job posts and unsolicited commercial DMs. Moderation is based on those "rules" (guidelines). That's why we (mods) have asked repeatedly for guidance about how we would "police" any rules around AI usage. Now I think it is clear that y'all don't actually want the mods to take action (good—thank you for the clarity around that), which means we won't be updating the CoC, etiquette, or workspace rules. Which is fine (and I think is the correct choice). But that leaves us in a place where some people are unhappy with what they see in the community but nobody seems to know what, if anything, to do about that. We (mods) don't want to see people unhappy—hence our concern here and continual pushing for guidance. Which we will stop doing (and likely disengage from the whole topic of AI, just keeping an eye out for CoC violations, and assuming that some channels don't really want the etiquette guidelines applied because those channels are very specifically for "opinion pieces and diatribes", to use language from the etiquette doc).
@seancorfield two questions: do you see this channel for opinion pieces and diatribes and do you see #C0BLJS63J2Y as a place for opinion pieces and diatribes or something that is attempting to get closer to what is in the clojure etiquette document?
Yes, this channel is definitely for "opinion pieces and diatribes" (and not covered by the etiquette doc), since it is specifically a "meta" channel. I would prefer we don't have too many of them here. I do not see #C0BLJS63J2Y as being a "meta" channel, so I do still consider it beholden to the etiquette doc. And it doesn't feel like "opinion pieces and diatribes" so far so good job on that (since it's "your" channel and you set the tone well).
while I do enjoy an off topic channel, this doesn't feel like the kind of place I need. Unless I'm needed here I'd like to follow the Law of Two Feet "if you're not contributing and you're not learning, use your feet to take you somewhere where you are doing one or both of these things." Tho, virtually obviously
I love and respect the mods here, but I think they got defensive too quickly this time. It poured oil on the fire and made it feel like criticizing the use of AI, especially in that channel, wasn't allowed. The line between opinion, diatribe, and technical argument is blurry. The impact of a technology on users, people who make things, the environment, or the UX it produces can all be part of a technical argument. You can't cleanly separate the technology from what it does and how it is used. The community should be able to discuss how to deal with AI use. What to expect when someone announces an AI-assisted or vibe-coded project, how to evaluate which libraries are worth using or promoting and can be trusted, and what idioms should develop around AI use. That discussion can still be inclusive. I use AI and didn't feel attacked. In fact, I learned that an AI written README can be more off putting than AI written code, which I genuinely hadn't realized. I even used AI to generate my last two announcements, thinking it could do a better job at grabbing people's attention without realizing it likely doesn't. And this happened in ai-skeptics. Some wariness of AI should be expected there, not treated as inherently hostile or inappropriate. I understand not wanting these discussions in the mainstream channels, it's touchy, it can get a bit more emotional, and it's also unsettled, people thinking out loud, as we and everyone come to grasp with the change AI is bringing. But here it feels appropriate. @seancorfield The relationship between this Slack, its moderation, and Nubank should be clear. Companies have strong incentives to push AI adoption. That's exactly why a community space needs to allow honest criticism of it without people worrying about retaliation. So if there's any policies driven by Nubank it would be good to know?
> I use AI and didn't feel attacked. I did not feel attacked either. As previously noted, several times, we were seeing disruptive, drive-by "anti-AI" sentiment being dumped on people who used AI for their projects (and sometimes their announcements). That behavior was not acceptable based on existing community guidelines. After some discussion with the folks expressing those sentiments we created a space for those discussions and publicly announced that space and why it was created—and guidance to avoid future drive-by criticisms. We would do the same for any similar behavior, based on any criteria. I don't think we (mods) got "defensive"—we were responding to what seemed to be calls to change the community rules/guidelines, and we asked what those changes should be and got no useful, actionable answers. When folks start saying "X should not happen here" that is inherently something that mods are going to engage with, and try to figure out whether/how the rules should change. We eventually got to the answer: the folks saying "X should not happen here" clarified that they don't actually want rule changes. Great. That's what we (mods) were trying to figure out. We're happy to engage with the topic as community members now, rather than mods trying to figure out what, if anything, y'all wanted. As I've said, I have a lot of sympathy for the dislike of "vibe-coded" projects for pretty much all the reasons that have been discussed in this channel. Outside of "vibe-coded" stuff, I'm genuinely interested in where folks stand on much smaller AI involvement—and that aspect of the discussions has been very enlightening (and, I admit, a little frustrating at times!).
(and, again, I will note that the worst offenders of the drive-by criticisms left this channel very soon after it was created—and have contributed nothing to these discussions)
GCC's now established an AI policy. Contributions of 15 lines or fewer can be AI-generated, but no larger, due to litigation and copyright concerns. https://lwn.net/Articles/1086041/
It's interesting that the rationale seems to be solely around copyright.
> The policy does not forbid use of LLMs for research, analysis, bug discovery and reporting, patch review, etc. as long as the output is not included in contributions. The committee says that it expects the policy will evolve and will be revisited periodically.
Yeah. Makes sense, given that GCC is part of GNU, which is the most Stallman-aligned organization, in terms of maintaining copyleft software and being strict about licensing.
There's been a dust up on eMacs devel and tangents mailing lists about a change landed that included a coauthor credit to Claude
Outcome seems to be no issue with Claude usage for commit message and other non code related tasks. Commit was removed and patch will be reapplied. Was a bit of controversy over reverting or force pushing without it
As per the next post in the channel, this seems more relaxed than the OpenJDK stance 😄
there didn’t seem to be any official guidance on commit messages and that was more off-the-cuff of the maintainer. Reasonable take for a non-lawyer i think. And the openjdk stance here about commit messages, emails, etc is a bit more guarded and perhaps more correct. Don’t know
I guess GCC thinks that if some small 15 or less LOC was found to "infringe" they could easily rewrite it, but not for larger changes
that might be below a threshold where it’s not enough to copywrite
like even a human couldn’t maintain that
Ya fair enough, like how do you prove you own the rights to a 15 loc change
Just for fun, take a 2k LOC PR, cut it up in 134 smaller than 15 LOC PRs 😛
I wonder how they got to the number 15 and if they had anyone on the legal side of things review it
15 lines is also the limit for contributing to GNU/FSF projects without signing the copyright assignment form (plus having your employer sign it, and whatever else is involved). So the number isn't new.
Yes though I suspect what matters most for them is IP over the JVM