vriska
@vriska.bsky.social
1.4K followers 270 following 1.8K posts
https://vriska.dev
Posts Media Videos Starter Packs
Pinned
vriska.bsky.social
if you actually want to see me post things follow the bridged account at @leo.60228.dev.ap.brid.gy
Reposted by vriska
jonbois.bsky.social
i've never been inconvenienced before and i hope i never will be. it sounds terrible
vriska.bsky.social
i don't like that it isn't going to be properly integrated into the labelling system though, it's just a !takedown global label
vriska.bsky.social
bluesky is working on doing this but only for third-party clients, fwiw
vriska.bsky.social
my hot take is that making hard bans impossible would be a bad idea but is also fundamentally necessary for stackable moderation to solve the problems it sets out to
Reposted by vriska
ryannorth.ca
In video games, you will often find a fully roasted chicken abandoned on the ground, and eating it extremely quickly makes you feel better. What's less known is that this is also true in real life
vriska.bsky.social
this means you'll give them to zoomers for free right?
vriska.bsky.social
is SVP status neutralized if the company goes bankrupt
vriska.bsky.social
i have difficulty imagining a scenario in which this is actually, like, a problem, though
vriska.bsky.social
wait, IETF makes localhost being 127.0.0.1/::1 a SHOULD and not a MUST??
vriska.bsky.social
my posts get more attention on mastodon than here i think you all might just be bad at posting
vriska.bsky.social
bluesky was explicitly founded as "a small independent team of up to five open source architects, engineers, and designers to develop an open and decentralized standard for social media [...] for Twitter to ultimately be a client of"
vriska.bsky.social
i mean. that was the whole point, twitter was going to be the flagship implementation of bluesky's protocol
vriska.bsky.social
rust has this feature too it's needed to express some complex control flow
vriska.bsky.social
this is not something i'm remotely interested in actually having an argument about i was just assuming they were unaware of all the research showing how it's worthless
vriska.bsky.social
"We can have this conversation" (locks replies and makes claims i already posted counterevidence to)
lol
707kat.bsky.social
Hello there. If you want to have this conversation. We can have this conversation. But what you are saying is not true.

Also linking to an arxiv article from 2 years ago in AI development is a major red flag.

It's not snake oil. Can it be bypassed? Yes. Trivially? No.
The people who bypassed it-
vriska.bsky.social
the glaze authors' response was basically just "okay it doesn't actually work, but really it's the thought behind it that counts"
vriska.bsky.social
(nobody remembers this, because it was resolved very quickly.)
vriska.bsky.social
i'm pretty sure they did?? i remember this exact discourse happening in 2023 and ending the exact same way
vriska.bsky.social
the glaze authors' response was basically just "okay it doesn't actually work, but really it's the thought behind it that counts"
vriska.bsky.social
this is the most direct paper i'm aware of, tldr it's trivial to bypass and barely does anything anyway ("we found that Glaze (Shan et al., 2023a) performed significantly worse than claimed in the original evaluation, even before actively attempting to circumvent it") arxiv.org/pdf/2406.12027
arxiv.org
vriska.bsky.social
glaze and nightshade are snake oil, researchers in the field have criticized the methodology from the beginning and more recent evaluations have shown that they don't work at all
Reposted by vriska
myrrlyn.net
every day i go in to the office and collaborate with a graphical UI designer and every time he says he’s updated figma for me i do not say “figma balls” back at him. i would like a fifty thousand dollar bonus for this effort please
vriska.bsky.social
can't you just appeal it's not like they'd manage to take you to court anyway