Unable to load image

Bard spotting on Bluesky 16 November 2024 :marseyrandom:

Here we spot wild Bardfinn Bluesky activities.

Be valid and ping ! bardfinn for something worthwhile or create a new thread.

5
Jump in the discussion.

No email address required.

Dr Henry 🦃 (@henryhenryhenry.com):

valid. not sure the composable community mod systems have been really tested. like nation-state adversary tested. guess we'll see.

Sara Aniano (@coolfacejane.bsky.social):

My main concern about Bluesky right now is not that it's an echo chamber, but that the content hasn't gotten "bad enough" to adequately test its content moderation — in other words, if purveyors of disinfo and hate do come here, are moderators prepared to rise to the occasion?


Ms. Penny Oaken, SkyWitch (@skywitches.net):

My current concern is, "What effect does it have on the site if 30,000 accounts are added to a blocklist 10,000 people subscribe to". And - they can't disclose that (it would be like shouting "break us" if they did) & I can't test it just to test it.

But I see it approaching.


Ms. Penny Oaken, SkyWitch (@skywitches.net):

I think they'll be able to handle an influx of hate groups as long as user blocklists scale. An imageboard tried to "raid" harassment a month back & I'm told they script-made north of 100,000 sockpuppets, which T&S handled.


Ms. Penny Oaken, SkyWitch (@skywitches.net):

Hate / harassment on a nationstate level will either be AI-generated individuals, or they'll organize into "hives" the way /pol/ & groypers did on Twitter @ 2018, or they'll mass follow a "leader".

1 is expensive for an attacker; 2 & 3 are cheap to blockenheimer. Economics favour us in all 3.

Jump in the discussion.

No email address required.

Dr Henry 🦃 (@henryhenryhenry.com):

interesting. I get the sense that it's the non-standard BS-specific features—like you said, blocklists and such—that provide new terrain for organized frickery?


Ms. Penny Oaken, SkyWitch (@skywitches.net):

Blocklists here are almost certainly implemented with a Bloom filter, for space & computing resources efficiency.

They're not space-optimal, however, and they have to sacrifice more space to minimise false positives. Some providers that I know use them limit the # of elements, to prevent overflow.


Ms. Penny Oaken, SkyWitch (@skywitches.net):

Like, on Reddit, which uses Bloom filters for blocking, personal blocking is limited to 1,000 blocks.

There's also a compute hit for rebuilding them if elements are removed. There's ways to mitigate that, too, but no way to know if BlueSky has backend to do that.


Ms. Penny Oaken, SkyWitch (@skywitches.net):

I think they might have some experience / roadmap on handling these kinds of things, though, since there were people here early trying "block everyone" experiments.


Ms. Penny Oaken, SkyWitch (@skywitches.net):

We also need 2FA. We need 2FA last month.

Jump in the discussion.

No email address required.

Redshift (@redshiftsinger.bsky.social):

Reddit's blocking limit is fricking terrible for anyone with an account more than two years old who's both moderately active and part of a marginalized community.


Ms. Penny Oaken, SkyWitch (@skywitches.net):

I agree 100%.

Every month I start up a script to cull the suspended accounts from mine, which is the only reason mine has only hit the limit twice.

If I were as powerful a baba yaga as the bigots believed, every one of them would be perma'd

Jump in the discussion.

No email address required.

Link copied to clipboard
Action successful!
Error, please refresh the page and try again.