Posts by vivi@arff.archandle.net
(DIR) Post #AuZgDSfb3nMPZ1t9F2 by vivi@arff.archandle.net
2025-05-29T04:03:35Z
1 likes, 0 repeats
ready for sleep and biting
(DIR) Post #AuyCiSJLlm0nk7WvHE by vivi@arff.archandle.net
2025-06-09T23:46:47Z
0 likes, 1 repeats
snuggle ?
(DIR) Post #AvHkn4PcXIVHKlnW3U by vivi@arff.archandle.net
2025-06-17T21:00:11Z
1 likes, 0 repeats
Vibe versioning
(DIR) Post #AvWeUHlPckQMGBs6Yi by vivi@arff.archandle.net
2025-06-26T12:47:45Z
2 likes, 2 repeats
My instance is fucking dying please boost to finish her off :vivi_flat_ko:
(DIR) Post #Avc4lb9wKlDFw63m40 by vivi@arff.archandle.net
2025-06-28T17:12:56Z
0 likes, 1 repeats
fool. you do not go "viral" on mastodon. you go "woolly". you simpleton. you absolute station wagon
(DIR) Post #Avc4lgpFGcDxVeQKdk by vivi@arff.archandle.net
2025-06-28T17:30:54Z
0 likes, 0 repeats
Fun "Gargron" Notallowed ratio'd by Viridi "mosslet" Vix
(DIR) Post #Ax1a7vtBtfLfdI9K88 by vivi@arff.archandle.net
2025-08-10T04:19:54Z
0 likes, 0 repeats
@kevinrns @dotjayne how would you propose to poison alt text in a way that is still accessible to those that rely on it?
(DIR) Post #Ax1a7y6Nfb5UUmKsy0 by vivi@arff.archandle.net
2025-08-10T10:34:14Z
1 likes, 0 repeats
@kevinrns @dotjayne that's wonderful, thank you for these very nice articles,,, so let's get back to actually answering my question instead of diverting, how are you going to apply this to alt text which is a tiny stream of text characters which have to be chosen in such a way to accurately describe the picture and not just nonsense, in a way that will not impact accessibility?Image poisoning is a little different than a text stream since the human eye can handle a bit of lossy noise and still see essentially the same image, text streams are leagues different. Most image alt text is 100-200 characters, if you want to sufficiently poison a tiny bit of text on such small scale, for individual images, would absolutely require giving up any hope of writing an accurate written description of the image, unlike poisoning hundreds~thousands of alt texts over an entire sampleset. And what's your idea on how to avoid impacting screen readers which rely on local TTS models to read the alt text? These would just as likely be affected as well from poison streams and would incorrectly read aloud what the actual text says.Or do you see this fancy poison tech and think it works like a magic panacea solve-it-all tool and no issues would arise and it Just Works, just like how techbros think LLMs are a magic panacea solve-it-all tool and that no issues would arise and it Just Works? It doesn't, you're just gonna fuck up someone's accessibility on the web and not solve anything.I'll give you a hint, or maybe even answer the question for you. You don't need to fuck with alt text and pay the accessibility toll to achieve the same effect. Just poison the image instead and give genuine alt text. The LLM would then not be (as) able to associate the alt text to what is represented in the poisoned image. And one or two descriptive sentences describing a poisoned image that the AI doesn't even know how to correctly interpret is a nothingburger for training data.Advocating for making accessibility worse, even by just a little, by poisoning the data of the only asset some people will have to rely on navigating the web, is the wrong approach. Throwing your allies under the bus when confronted with fascism is truly a speedrun strategy to ensuring you have no allies left when later confronted by fascism.
(DIR) Post #AyLsypO0g6ShBHVJHU by vivi@arff.archandle.net
2025-09-19T03:23:06Z
1 likes, 0 repeats
anyone with a telescope in northeastern US help me identify a thing in the sky? from my position (i'll dox myself for science, Michigan) it lies pretty much on a straight line between Capella and Polaris, but very near Capella. like, I'm not good with polar coordinates, but start at Capella, turn left about maybe 5 degrees, and up 2 degrees. ish. i dunno, I don't have any equipment or knowledge about this stuff. Anyways, it's mostly white but also twinkly red sometimes, and it's not fixed in position like the rest of the night sky, but has risen every night with the rest of the stars but then around 1AM it drifts to a halt and then turns back the other way and sets near where it rises, so I think it's some kind of satellite that has a weird orbit interacting with Earth's rotation in a way that makes it move in that funky pattern. but i'd love to know what that thing is :vivi_heart: I have seen it every night for the past 5 nights in a row so it's a relatively constant feature of the sky now
(DIR) Post #AyzRLGFLH7DHsSNng8 by vivi@arff.archandle.net
2025-10-07T20:23:41Z
1 likes, 0 repeats
look at me ——— i'm an LLM ————— billionaires give me money please please please —————————— invEst in LLMoss ————————————————
(DIR) Post #Az3PQgNS7sZz5nTQUC by vivi@arff.archandle.net
2025-10-10T03:44:45Z
1 likes, 0 repeats
it's Fucking cold
(DIR) Post #Az3PQnGcVgH0SdKZJA by vivi@arff.archandle.net
2025-10-10T03:46:35Z
0 likes, 0 repeats
86F --> 44F (30C --> 7C) two day transition
(DIR) Post #Az3PQu4pBvzta4rkOW by vivi@arff.archandle.net
2025-10-10T03:47:57Z
0 likes, 0 repeats
I make this joke all the time but it's just real nowMichigan legislature decided to save money by simplifying the overly complex four seasons system, into just two, Summer and Winter. There is no Spring or Fall anymore
(DIR) Post #B0ZBCMwsbAQ8ARKEKW by vivi@arff.archandle.net
2025-11-24T03:43:49Z
0 likes, 0 repeats
i suspend instances on sight when i see that dead horse trick with the $INSTANCE_HOST or whatever, or other forged posts like those dated to 1970on the surface that seems inane and petty but i see it as, if your instance infrastructure is equipped to supply falsified post metadata or dynamic content in which the post payload is not the same from everywhere it is viewed, then there's nothing stopping that instance from using that for large-scale misinfo fabrication instead of just silly jokesit does make it very easy to spot when it's being used though since nobody is going to be discussing the domain name arff.archandle.net really for any reason, i'm not popufur enough for that :vivi_thumbs_up:
(DIR) Post #B0ZBCOOZDiA6ebHr7o by vivi@arff.archandle.net
2025-11-24T04:47:49Z
0 likes, 0 repeats
the crux of the issue is thisanyone can of course say anything on the internet at any time, and it's up to the reader to verify before fully believing whatever you read. not exactly shocking that misinfo does existlet's say there is a screenshot of your post that says it's from 2019 that says "I predict valve will announce the Steam Machine on November 12 2025"a quite overwhelming majority of specifically fedi users who are generally more versed in tech understand that screenshots are easy to forge (and regularly are) such as through browser devtools etc, and would be apt enough to know that this prediction is likely full of shit. there is already a baseline mistrust of misinfo surrounding screenshotsbut when you can forge posts or their metadata, you can send a "real" link for readers to locate this post with a "real" timestamp dated 2019 and this is a kind of receipt that is less immediately struck by the screenshot sniff test. in almost all webservices that a particular person has ever interacted with, forging this type of data is not as accessible as the F12 menu.there is then no possibility for anyone to fully disprove that this prediction was falsified like you can with a screenshot. because the distributor can produce a link to this post to substantiate the claimlike, if I had the ability to record a video of me saying that phrase "I predict valve will announce the Steam Machine on November 12 2025" and could upload that to Youtube or whatever and alter the metadata to say it was published in 2019, who could call it out as bullshit? i would have the receipt that the video upload date says 2019 as evidence, and opponents cannot cite lack of a link to that post as evidence, like one can with a screenshot.if the matter at hand was anything more serious than a party trick of predicting a product announcement, there is a serious issue here, because subconsciously we're trained to assume that links to these types of posts all but confirm proof - at least that's learned from typical use of centralized platforms.the one thing i always see in particular with the $INSTANCE_HOST gag is the replies are almost always, still to this day, for anyone who's never seen it before, being so surprised that maybe their niche instance is being discussed in that way, and only then in further-level replies would either the original poster or someone familiar with seeing this trick in the past, would that user ever learn that in reality it's just a quirk of their software that allows this.It's even muddier when you are supposed to be able to trust your own instance not to have falsified post content so you should be able to trust your local cache copy of the post, but the way that $INSTANCE_HOST placeholder works means it's falsified on the remote end and then sent in plaintext to one's local instance with no indication that string was originally a string template. when I receive that post it reads "arff.archandle.net" in plaintext without any indication it was ever a string template, ergo my instance has a different falsified copy of that post than every other instance, that i am supposed to be able to trust since my instance does not support falsified content.again, all of the samples here are just silly inconsequential fun, but the framework that exists for falsified content for silly inconsequential fun also innately means that the framework exists for falsified content for gravely serious matters too. what happens when a spiteful admin hates one of their users - hates them so bad that they don't think banning them is sufficient, they instead turn to forging a post from that user with some reputationally crushing activity like, i dunno, spamming slurs or soliciting minors or something? How does the user ever even recover from that? "I swear I didn't post that" doesn't generally fly as an excuse anywhere on the internet when someone can produce a cite hyperlink to said post, but with the way ActivityPub is designed, this fucked up scenario is genuinely possible.and if you're the admin in this scenario, and you're already well familiar with the infrastructure for falsifying post data for fun, then the option is there for evil too.yes, this is inane and petty, but i don't sit right with the idea of normalizing falsified posts, even just for fun
(DIR) Post #B0uS0bw2pd6EiF4XVg by vivi@arff.archandle.net
2025-12-04T15:50:25Z
0 likes, 0 repeats
@peertube US QWERTY support? ? and shift+? and shift+/ are all the same key :neofox_googly_shocked:
(DIR) Post #B269KXDBWqRFwZyvqq by vivi@arff.archandle.net
2026-01-08T23:53:19Z
1 likes, 0 repeats
just beat minecraft for the first time ever :neofox_floof:
(DIR) Post #B2mKv1b2jbSjnH9mM4 by vivi@arff.archandle.net
2026-01-29T06:48:05Z
1 likes, 0 repeats
fedi's great cuz there's just no requirement to be hinged
(DIR) Post #B2mKv7IpWESVUWgJnc by vivi@arff.archandle.net
2026-01-29T13:00:10Z
0 likes, 0 repeats
who's gonna ban me. god herself :vivi_sit:
(DIR) Post #B2tNjJImjvspCURSgS by vivi@arff.archandle.net
2026-01-31T21:15:00Z
0 likes, 1 repeats
@imbl who in their right mind would think it's a great idea to need to buy a $100 sticker to "prove" their project isn't AI. nobody owns the rights to "not AI" just write "this project is not AI" on it. it's free :neofox_googly_shocked: love to imagine the roundtable that brought about this racket.billionaire AI goons when AI was not as profitable as they first imagined: "...well if AI isn't making us money... why don't we try making money off of not AI instead?"and everyone in the room clapped.