[00:01:05] She's a mix of golden retriever, putbull, boxer, and American bulldog [00:02:18] [1/2] I think this shows off her size a little better [00:02:18] [2/2] https://cdn.discordapp.com/attachments/615786602454581249/1527103044591030282/20260406_184112.jpg?ex=6a59710a&is=6a581f8a&hm=ad7321e0c887c6e020555e1618d076f37fad6e32090e6194b87d28c26596e44d& [00:13:44] aww, cute! [00:14:20] it's hard to tell in photos the size of a collar. we have a mini-dachshund with a similar looking collar, but would be way smaller if you saw it in person 🙂 [00:38:09] Awww, reminds me of our 100lb rott/doberman mix that passed last month, 100% sweetie but a very big lapdog [00:46:58] They sound like they were very sweet, I'm sorry for your loss [00:50:59] [1/2] Yeah, absolutely gigantic but incredibly gentle. [00:50:59] [2/2] https://cdn.discordapp.com/attachments/615786602454581249/1527115295595167854/AP1GczN7LbyxKfwew-KHBz2NcjSlwmNpBn78hCbaEzJ2i_eIykdblJQJ4gjfAAw1125-h945-s-no-gm.png?ex=6a597c72&is=6a582af2&hm=aed106c9765275e0da805fd56d6af5c39b323c419ba530df4be8de73577fa23b& [00:53:22] sorry to hear about that...losing a pet is one of the worst experiences 😢 [01:05:25] we should make a remade Miraheze landing page look like it's glued together [01:05:57] kind of like how the Baldi's Basics Wiki does it [01:07:06] lol [01:07:32] it would look cute [01:07:54] save it for a rainy day, why not [01:09:36] any style that's heavily different from the corporation slop-like site design is a bonus [01:10:37] ngl we really need to decide on a mascot if a redesign for the landing page comes to reality [01:13:42] maybe it could have a book-style design [01:13:44] Large dog [01:15:12] Large dog [02:12:12] Large dog [02:12:31] gasp Wicker didn't do her perfect grammar™! /j [02:16:06] Scandalous!! [02:18:11] I must be sent into the wiki mines and do 250 hours of nonstop editing! CARPAL TUNNEL, HERE WE COME! [03:29:23] Lorg Dorg [03:29:58] Dorg Lorg [03:47:08] may they rest in peace, dog are truly one of the sweetest creatures i've ever seen [03:48:30] [1/2] honestly, i could see the miraheze mascot be a bee [03:48:30] [2/2] why: honeycomb-like brand miraheze does have [03:48:42] like how a bee would make an honeycomb [03:48:56] Bee [03:49:14] There is a thread of mascot concept designs in #general and I am still in love with Xena’s design [03:49:37] it also does really fit the wiki community, as many bees contribute to it at the same time [03:49:38] they have double square bracket antennas [03:50:15] unsure where to see it exactly, but would love to [04:09:51] check archived threads in general [04:37:16] Lang Dang [04:53:05] https://cdn.discordapp.com/attachments/615786602454581249/1527176223535665152/HNTfNbUXkAALwzi.png?ex=6a59b531&is=6a5863b1&hm=b9e54b46dce2c91a58c3191da0556759fc45d00b53357abda71ba35cd1624f98& [05:05:58] ANCIENT ALIENS [05:06:13] A second goal has hit the Big Ben! [05:06:30] 😭 [05:06:35] omg that was actually funny [05:07:05] Next Sunday, let's see if I can post the same meme with Messi goaling Sagrada Familia [05:09:48] 😭 That made me lol hard. [05:15:22] I'd rather my rulers be the bees than Fand*m [05:15:30] At least bees give you freedom! [05:16:48] FOR MOTHER BEE! [05:20:00] Being a Fandom slave is horrible [05:34:31] It's worse than anything else. [05:50:12] Fr [05:50:31] But hey, what hurts you makes you stronger [05:51:29] And now I have wikis under Gamepedian (wiki.gg), Mirabee, and RuneScapian (Weird Gloop) rule :p [08:01:02] Hell yeah. [13:58:26] I'm personally a fan more open source LLMs, so this is good news (https://x.ai/news/grok-build-open-source). But the really important question is, does anyone use it and is it any good? [14:02:16] ive heard its okay. hard to see it as anything more than a racism and nsfw printer though [14:02:32] [1/2] the bar is fairly high; you need a certain class of hardware to get anywhere for which I barely have the resources to even try, you need to come in with some realism what you can expect and a practical need to solve or its just fluffing around, your skills interacting with it must grow pretty quick to get much return relative to models that run on bountiful hardware which will always [14:02:33] [2/2] be faster and probably just better. But maybe if you can find a good enough compromise you will have a reasonable return on investment [14:03:24] and there are real functions to it ie some kind of search or summarizer for a local library if that's something you want to put together [14:03:44] you might manage something that can produce relatively usable scripts a little faster if you have starting savvy to address pitfalls [14:04:16] these niche cases are where I see the most opportunity and have been lightly tinkering with local/more openly sourced ai myself [14:04:43] but more than half of that is just curiosity and not because I've found or expect to have some revolutionary benefit [14:19:36] raidarr, agree with that. I still haven't reall explored the new features in Google's AI Studio. It would be interesting to see how Grok Build compares to that at least [14:27:31] I've gone fairly out of my way to avoid those models myself as they are the apex of the industry I do not wish to support, which yes, means I put myself on the back foot [14:54:03] Bruh, that's just the tool, they didn't give you the AI model, aren't they? [14:57:14] Open source LLM = publish LLM models you can download like Gemma 4, Deepseek V4, Qwen 3.6 or something like that [14:57:32] raidarr, fair. I can appreciate that. But I can appreciate that being so large, they also present an existential risk to the global economy to not f*ck up in a major way. [14:57:32] the ones I'm most troubled by are closed models peddled by their proxies, Anthropic and OpenAI [14:58:13] max20091, agree. I know, and I prefer the open models, too, unless they outright suck like Llama / Meta lol [14:59:02] haven't tried Qwen yet, but it's received some good reviews / demonstrations on YouTube [15:00:24] Yea, the modern small-med models are very smart [15:02:11] [1/2] I saw somewhere people saying closed models are made dumber over time. [15:02:11] [2/2] These open source models aren't. [15:16:26] qwen is cool because it rejected the proprietary american establishment and said ok here's all of it [15:16:58] including other ingredients like weights that us megacorps are shy about because they have a vested interest in people not getting too competent with their local solutions [15:17:38] as far as I gather its also the most useful selfhostable option for code and is competitively competent in other uses [15:18:40] llama I've seen have moderate popularity as a 'budget' llm for writing related things but meh [15:33:00] > qwen is cool because it rejected the proprietary american establishment and said ok here's all of it [15:33:00] nice :) [15:33:26] > including other ingredients like weights that us megacorps are shy about because they have a vested interest in people not getting too competent with their local solutions [15:33:26] heh true [15:34:26] > llama I've seen have moderate popularity as a 'budget' llm for writing related things but meh [15:34:27] heh maybe. I'm just mildly against it because of some very bad Star Trek crossover image generation memes I've seen on social media :P [15:34:51] which is kind of sad how much we're spending on gen AI to basically produce memes [15:35:06] it was both bad in term of intelligence and having AI geofencing [15:35:24] we're = civil society [15:35:40] Banned in EU if I'm correct [15:36:04] AI geofencing meaning a jurisdiction has banned it? [15:36:42] No idea, the license prevent anyone in EU to use it [15:36:57] max20091: ack [16:17:16] https://www.reddit.com/r/sysadmin/comments/1uy08tz/linus_torvalds_puts_the_foot_down_against_antiai/ [16:25:52] +1 - that's my stance as well. It's a tool. Contributors are still responsible for the code they submit, and code still needs to be reviewed by human volunteers [16:27:13] as usual, he doesn't pull any punches, either. [16:27:14] "But the solution is not to put your head in the sand and sing 'La La [16:27:14] La, I can't hear you' at the top of your voice like some people seem [16:27:14] to do." [16:27:52] "Because it's not like natural intelligence is always all that great either." [16:27:52] lol [17:55:28] kinda my thing sometimes, I'm not always convinced what AI can come up with is worse than what typically ensues when you throw the usual expected human at the matter [17:55:56] heh [17:55:59] my problem at scale is almost purely the big corporate evil factor + it being slathered into places without justification which is really just the first factor extended [17:57:01] and likewise with my foundational quibble of how it was cobbled together, which has proven better than anything else that copyright is a hypocricy that supports the wealthy, not a law with useful applicability to the common peon [19:09:25] honestly i dont think LLM's exist ofr like everyone [19:09:30] it can be very harmfull [19:09:46] also if u use it for "art" thats just disgusting [19:09:59] and the enviromental impact [19:10:04] its just not worth it [19:15:40] i do def support it for like detecting cancer early and research [19:16:05] but thats not really generative ai [19:18:43] [1/2] Yes, though, the costs are coming down significantly in that regard, which will force the LLM providers to lower their prices. Unless they're also a hyperscaler, this will exacerbate existing pressures on their margins. [19:18:43] [2/2] The probability of Anthropic or OpenAI having an insolvency proceeding is not zero [19:19:26] Thats fair [19:19:56] Its just not worth it, even if the costs are coming down imo [19:21:19] True. I haven't tried any image gen models from some of the open model providers from China. Do DeepSeek and Qwen's parent companies have an image gen model yet? [19:22:07] image gen is just harm, [19:23:00] [1/2] - ruins the point of art [19:23:00] [2/2] - can be used for illegal stuff [19:23:05] I don't know that I agree in all cases... all compute costs money, whether it's done by an image gen tool or by a human in a cloud-based environment. [19:23:52] I will say it's kind of funny when companies use AI generated images of music in their promotion, thereby making that creation public domain (at least in the U.S.) [19:24:52] seeing a corporation or someone use ai for art is just fucking disgusting. especially a corporation, they couldve just hired an artist ;-; [19:24:55] Are you sure that's the law in this case? My understanding is the works are still attributable to the human/corporate entity creators? [19:25:41] True. But maybe they did and the artist used Gen AI to assist in the creation? :ThinkerMH: [19:25:48] shoutout to code reviewers who have to now sift through line after line of ai sludge code [19:26:00] lemme pull up smt. hold on [19:26:35] https://cdn.discordapp.com/attachments/615786602454581249/1527396045913325658/ypxri0hv4n6h1.png?ex=6a5a81ea&is=6a59306a&hm=43dcaf480f1b3cf74b7a4c58c9ec7481083ef180709ae424afa3306b38762c19& [19:26:45] here [19:27:01] at the very least the use of ai makes artistic copyright extremely complicated and if a well-resourced entity called a corp on it they could have a problem. But because legal systems are money games that's the trick [19:27:08] Iunno all the details but AI images like that Coca Cola ad are updated to Wikimedia under PD. [19:27:19] funny, it does complicate the origins of things for sure [19:28:31] if its fully ai outright my understanding is the copyright on the thing is void [19:28:34] I saw a Chinese kernel dev complain about how incompetent frontier models are when dealing with his coding tasks. And when he tries the models on webdev the difference is night and day. [19:28:35] not worth it, enviromentall impacts [19:28:45] i read wrong [19:28:46] mb [19:28:46] that happens a lot of times because Wikimedia Commons has an aversion to Fair Use licensing (since Fair Use is a U.S.-specific creation). Could be a lot of misapplied licensing templates [19:28:50] don't think its been tested with what you know is vibe coded slop baked into proprietary software [19:29:21] Interesting. I'll have to review it more. [19:29:22] and fair use anyway is basically a speculative court defense and I imagine wikipedia dislikes hedging its bets on something incredibly expensive to assert in reality, on top of it being a US particular thing [19:29:29] this is why LLM's shouldnt be avaliable to the public [19:30:01] Under U.S. law at least, AI images can't be copyrighted [19:30:03] Diffusion models (which existed for a long time) are still pretty good afaik. There's also models that can generate and revise images based on user chat but I'm not sure whether these have good open weight alternatives. [19:30:13] that is good [19:30:30] oh goodness, yes, all this arguing over the strength of one open-source license versus another may be moot when you consider things like this [19:31:07] oh yeah you know the vast majority of real world ai is just corporate models with every disadvantage in full force being used for every toxic purpose imaginable [19:31:14] real world llm usage [19:31:33] verses the technology explicitly not llm which is driving research and creates value [19:31:42] And yeah AI image generation is just an excuse not to pay artists what they're worth [19:31:55] Terrible use of the technology with no upsides [19:32:10] some argument at the fringe exists if those people might have bothered to pay an artist in the first place [19:32:28] that's why I believe governments need to get with it and adopt Universal Basic Income, so artists and other creatives can survive and continue doing what they love [19:32:32] Drawing stick figures instead would be more worthwhile [19:32:33] for corporations there's no excuse, for my budget, hm [19:33:08] Make the hyperscalers pay for it with a new corporate tax [19:33:30] for corps its the zero sum unregulated capitalist game, if they can do it cheaper and the consequences do not outweigh the benefits why would you not do it [19:33:54] https://klipy.com/gifs/marx-karl-marx-3 [19:34:00] Also why they're firing swathes of their coding teams right now [19:34:13] Choosing to believe in vibe coding instead [19:34:17] and then they realize oh shit we probably could use some of that expertise when ai keeps making shit up [19:34:29] Yeah but by then it'll be too late [19:34:33] indeed [19:34:47] dont wanna be too political but god i hate current capitalism (like in the U. S.) [19:34:50] they are all hell bent learning these lessons the hard way [19:34:59] yeah I am very tired of the US model [19:35:14] I am, glancing at wwr, very weary of it [19:35:16] that's not political. that's a statement of fact of macroeconomic environment [19:35:22] fair [19:35:54] its pretty evident when looking from anywhere not in certain corridors of the us tbh [19:36:05] i heavily doubt the models used "as a tool" while coding are not these same corporate AIs tbh [19:36:41] you have to go very out of your way to be using an ai that isn't a corporate ai with the same origin of content theft even when locally hosted [19:36:51] at which point you're behind the curve doing a lot from scratch [19:37:09] i hope the ai fad goes away soon enough [19:37:20] I think it's global problem. There are country-specific issues and poor leaders, for sure, but I think because of lack of direction from governments everywhere, the best days of capitalism are behind it [19:37:20] Back in my days, if you want to make something and it's free. You actually have to LEARN how to do it. [19:37:24] I think it will only have a dot com moment but never go away [19:37:27] Not paying some corpo money for AI "generation". [19:37:31] Which is mostly plagiarism. [19:37:55] Besides, AI can't do anything right. [19:38:00] Not code, not art, not movies, nothing. [19:38:09] i can only hope whichever linux project i switch to is human made and remains so [19:38:11] it will lose hype, then become a standard part of toolkits in the domains it has some value add [19:38:32] and meanwhile power the next generation of worldwide surveillance but that's a whole thing itself [19:38:48] Also when actually making something it makes you feel very accomplished. [19:38:48] and evidence fabrication which is one of my main worries with it [19:38:55] [1/4] The only thing they do right is that they're either: [19:38:55] [2/4] A - "Free" (They're selling your data) [19:38:55] [3/4] B - They cost money that is "cheaper" than people but not in the long run. [19:38:55] [4/4] C - The thing they're being told to do is very simple. [19:39:11] Like, I can expect Claude AI to code a website, which. Honestly. Who the hell want to code HTML anyway? HTML is garbage. [19:39:20] er [19:39:28] But anything else... Ehhh. [19:39:32] Also, we're not talking like. [19:39:35] Very high-quality websites. [19:39:42] Most corporate use dogshit websites. [19:39:44] So they can use AI. [19:39:45] I mean I like nice handcoded 90s inspired sites tbh [19:39:46] 🤣 [19:39:48] -# i like messing with html [19:39:56] lol how so? there's a lot to be said for static websites in terms of performance and load times [19:39:58] The issue is how long it'll take corporations to realize this. And by then a lot of jobs would've already been lost, and a lot of knowledge along with that. [19:40:07] Do you go to corpo websites? Even their handcoded websites look like shit. [19:40:11] And work like shit. [19:40:15] oh we'll have a full on recession, worldwide fallout etc first [19:40:21] So I think AI is slightly better than their coders. [19:40:24] not really no, I go out of my way to not go on them [19:40:28] Mostly because they probably forced a intern. [19:40:36] but yeah ai can slop out the entire corporate world and then some, and it already does [19:40:37] With bare minimum coding knowledge to code a website. [19:40:37] no...:bombasticsideeye: [19:40:45] Because all their websites are dogshit. [19:40:47] Hell, [19:40:50] Government websites. [19:40:55] Imagine them not CRASHING every 5 seconds. [19:40:58] Literally impossible. [19:41:29]