@antisomniac
Internet, YouTube, Wikipedia, TikTok, NYC, birds... Sr Research Fellow @ UMass Amherst/IDPI/Media Cloud. Best known as "some scholar who said something" from Hard Fork rhododendrites.com Rhododendrites @ Wikipedia/Instagram Antisomniac @ Mastodon.social
Official statement on the incident here: meta.wikimedia.org/wiki/Wikimed...
More or less cleaned up. So much speculation about attackers, and it turned out to be for a different target, years ago, implemented on Meta-Wiki accidentally.
So not an attack on Wikipedia, and all damage appears fixable, but sure did look bad for a minute. (and it's not clear what the call to the external domain was doing - hopefully addressed in a forthcoming statement).
It was an accident. Someone had a script to attack two projects unrelated to wikipedia, but stored a copy of the code on the Russian wikipedia. It took a mistake by a poor engineer who apparently loaded user-created js en masse, incorporating it into sitewide js on metawiki.
Wild events on Wikipedia this morning. Multiple Wikimedia foundation staff accounts compromised, doing a bunch of damage, spreading malicious JavaScript, and connecting to a .ru domain. Everything had to be put in lockdown. But it turns out...
I enjoy Bluesky & a lot of my favorite internet people are here but there is no denying it is the Park Slope Food Coop of social networks
Went down a bit of a rabbit hole of AI upscaling on WikiCommons following the now viral Alex Pretti image. Attached: that Pretti image and Francis "three-foot" Beynon that was her main article image until a few minutes ago. :/
Apparently I have a low bar for warm/fuzzy feelings about an AI company: Anthropic puts itself on the receiving end of government bullying by saying it doesn't want to be part of murderbots or a mass surveillance project. www.wired.com/story/backch...
DDoSing, not outing in this case. The doxing stuff is what some allege the DDoSed person was doing by trying to figure out who/what is behind archive.today.
New development: Wikipedia has deprecated archive.ph/is/today, and will eventually remove hundreds of thousands of existing links. Massive thread here: en.wikipedia.org/wiki/Wikiped... (and FWIW I just verified the malicious code is still active on the captcha page)
Not headed there today, but now I'm curious -- where in in the cemetery is the merlin tree? :)
Dear all capitulating universities and companies, why did you not realize all you have to do is tack on "this is not a diversity initiative"?
From Meta ending third-party fact-checking in the U.S., to YouTube loosening enforcement and reinstating previously banned creators, platform choices are setting the stage for a more volatile & confusing information space.
Sure seems like they already are for a nontrivial part of the population. Just asked Claude, and he agrees with me sooo...
20 years ago, Colbert introduced "Wikiality" - change consensus reality by editing Wikipedia. His viewers' edits were undone by humans, usually within seconds. We're now in the age of "scrapiality," where text is true by default if it exists on the open web. This time there's no human to hit "undo."
The original: arstechnica.com/ai/2026/02/a... and the context: theshamblog.com/an-ai-agent-... I hope they elaborate on the retraction, too, though I'm glad they acted quickly to take it down (and hope they put more effort into covering this terrifying story, since it's in their wheelhouse)
An AI agent published a hit piece on a python library maintainer after he rejected its code. Among the salient points: to humans the content of its blog is silly, but how do these reviews of human behavior influence AI decision-making down the road? theshamblog.com/an-ai-agent-...
The recently closed TikTok deal has some safeguards built in, but Chinese influence could persist. The fix: disclose licensing terms and let independent researchers verify what is happening on the platform, say Mark MacCarthy and Carl Schonander.
Here's a hidden secret of doorbell cameras, as revealed by Nancy Guthrie's kidnapping. They're recording even when you don't know it, and sending video to company servers even when you don't pay
The meaning of words is to be decided not by their use, usefulness, or history, but by their commercial benefit to the most powerful lobby groups. I have a recipe for almond milk in a cookbook from 1226. It has been used as a term in English for hundreds of years.
www.theguardian.com/business/202...
The Wikimedia Foundation is hiring a Principal Research Scientist. job-boards.greenhouse.io/wikimedia/jo...
Oh, the DDoS code is *still live*. If you've used archive.vn/is/today/ph today, you're DDoSing gyrovague.com. Screenshot from just now on archive.vn.
As someone on Hacker News pointed out, "Only somebody with some degree of lawlessness would operate such a project." We already knew it was sketchy: opaque, anonymous, operating an expensive service with money from *somewhere*... is it enough for people to stop using it? Probably not. 16/end
Vacillating: But it's an invaluable tool. Unfortunately, part of why it's invaluable is because it extracts value from journalists. For Wikipedia's part, I'm leaning towards affirming support of caution, even though it pains me from a free knowledge standpoint (not to mention linkrot). 15/?
Now Wikipedia is again debating a blacklist. I think first I have to think about my own use of it. It's such a wild betrayal that makes me wonder more about how, with zero transparency at all, material on archive.today might've been steered in other ways to align with the operator's goals. 14/?
Archive.today added code to its captcha screen so that anyone who saw it would repeatedly send a search query to Gyrovague. Yes, if you archived something last month, you participated in a DDoS attack on Gyrovague. gyrovague.com/2026/02/01/a... 13/?
Back in 2023, Gyrovague posted a summary of what it learned about Archive.today's operator. Some already covered above plus a pseudonyms, indications it's based in Russia, etc. gyrovague.com/2023/08/05/a... But then last month when Gyrovague declined the operator's request to take it down... 12/?
Then AdGuard shared a story about demands from a shady French entity insisting AdGuard block its users' access to archive.today domains. adguard-dns.io/en/blog/arch... In other words, more intense legal efforts targeting the site. 11/?
The recent story: FBI subpoenaed information about it from its registrar in November. arstechnica.com/tech-policy/... 10/?
Its long-term reliability is also unclear. Its domains are periodically taken down as it plays some extended cat-and-mouse. And there's a prominent rumor that it relies on use of botnets, but as far as I can tell that goes back to a claim by a Wikipedia user that didn't provide any evidence. 9/?
In addition to the operator being unknown (and it does seem to be a one-person project), it's grown to be a massive project but the funding model is also unknown. It displays a few ads, but it doesn't seem like nearly enough to pay for a huge archive like this. That's a big worry. 8/?
