YouTube removed a channel that was dedicated to posting AI-generated videos of women being shot in the head following 404 Media’s request for comment. The videos were clearly generated with Google’s new AI video generator tool. 🔗 www.404media.co/ai-generated...

Some initial thoughts on the requirements for AT's private data model

Very excited that my paper with @katakeith.bsky.social is now out in @polanalysis.bsky.social. We investigate whether LLMs actually follow the instructions/definitions provided in codebooks, propose some diagnostics, and release a new evaluation dataset. www.cambridge.org/core/journal...

#NewPaper is out! With #NucleoMEPOP coauthors Claudia Lagos and @karengheza.bsky.social, led by the amazing Isabel Pavez in @jmcquarterly.bsky.social ❤️ journals.sagepub.com/doi/10.1177/... Thread below 👇

If This Then AT:// (IFTTA) is now open source under the MIT license! Build powerful automation workflows triggered by any event in the ATProtocol ecosystem. Lexicon-agnostic and ready for the decentralized social web. GitHub: github.com/graze-social/iftta

not loving how fast the r-slur came back from the grave where it belonged

some thoughts on explaining atproto by comparing it to news wire services like AP or Reuters (h/t @seabass.bsky.social)

New preprint! Crowdsourced Context Systems (CCS) like X's and Meta's Community Notes are popping up on various social media platforms. How can we better understand, critique, and design such systems?

Nearly twenty years after the first igraph release, igraph 1.0 has finally arrived! 🎉 This release focuses on providing a stable and more consistent interface that users and downstream maintainers can rely on with confidence. For more detail, see igraph.org/2025/09/20/i...

Thank you to the 14,000+ people who shared feedback on our community guidelines! We’re updating our rules for clarity, strengthening enforcement, and introducing product changes for healthier conversations. New guidelines kick in Oct 15. bsky.social/about/blog/0...

Inside a tumultuous summer at xAI, in which Elon Musk's demands to make Grok less "woke" and more viral led to chaos: www.nytimes.com/2025/09/18/t...

Feeds thrive on context - and we just added a ton more. Graze now lets feed builders filter content based on quote posts, parent posts, author & mentioned user metadata, *and* text transcriptions from videos! More freedom to our feed builders, and way better feeds for the ecosystem! Lets see a few:

I think it's good that people could migrate to Bluesky when they wanted to leave X I think it's hard that ATProto, the protocol Bluesky is built on, still needs a lot of development to reach full maturity, particularly around safety and privacy It's hard to build a plane mid flight

This adjustment means if you’re hosted by blacksky.app (myatproto.social, or cryptoanarchy.network) + using blacksky.community you should be more insulated from bsky moderation in the future (Without us doing a ton more work) Our roadmap is unchanged. Just a heads up.

Losing my ever-loving-mind watching the same people who were just clutching their pearls claiming censorship over mean emails from WH staffers to Twitter about COVID misinfo are now HAVING THE FCC CHAIR openly threaten broadcast licenses over a joke about the president AND THE BROADCASTERS CENSOR IT

Clay Higgins sent a letter to Bluesky and other social media sites threatening investigations if they don't censor speech critical of Charlie Kirk. They don't even know who Jay is because they addressed her as "gentleman." clayhiggins.house.gov/wp-content/u...

Please share! We're trying to crowd-source a dataset of post-1929 novels with maps.

I saw this yesterday and can't stop thinking about it. To see "let's just murder homeless people, that'll solve it" drop *unchallenged* on one of the most popular shows on TV and there's no consequences whatsoever... why is this our world? Disgusting, shameful, and alarming.

I would simply do social network moderation correctly. I can't provide any examples of this happening in the geological record, but I'd pull it off.

🚨 New paper alert 🚨 Using LLMs as data annotators, you can produce any scientific result you want. We call this **LLM Hacking**. Paper: arxiv.org/pdf/2509.08825

ATProto question: I understand the PDS, relay, app view distinctions, but how do relays actually know which PDSs to read from? Is there a way relays search for all PDSs? Do PDSs "register" with a relay so they're broadcast through it? I understand relays aggregate PDSs, I just don't understand how

📣 Yale workshop, Oct 16-17! 📣 How could/should content ranking work? What's new in content moderation? How can platforms promote civility? Hosted by Yale's Institute for Foundations of Data Science (FDS). Great speakers! Submit posters by 9/22! Spread the word! yalefds.swoogo.com/socialalgori...

Public transit is (or *can* be) efficient and sustainable and delightful and popular!

When you publish a website that posts the names and photos of scholars who have written or said things you disagree with and encourage people to surveil and harass them, you are not, in fact, a “free speech” organization.

The bizarre liberal eulogies of Kirk by Klein, Newsom, & others overlook his relationship to violence & that of the white conservative movement more broadly. If we're serious abt not wanting to live in a society where ppl are subjected to this sort of thing, we absolutely can't normalize this stuff.

My feed is almost entirely people condemning violence and telling others not to celebrate this.

It’s not that it’s a bad taxonomy of TTPs, it’s that systematizing human information behavior like this constructs a narrow perspective on what harms and abuse ought to look like and how they can be addressed (hence problems with operationalization); imo this is the problem with the cybersec mindset

It's tiring how many people come to trust & safety from cybersecurity that claim there are so many unknowns with how to keep people safe online and how to build systems to do so, when in fact it's just that they're not familiar with the space and their expertise doesn't immediately port over

NEW: On June 11, Meta announced new measures to prevent AI nudifiers from advertising on its platforms. Since that date, the company ran at least 4,215 ads for these tools of online abuse that reached 359,547 users in the EU and the UK alone. And that's not all.

I don't really care if the future is short videos. I like words. Words have gotten me this far and I'm gonna stick with 'em for the duration

Here's something I've learned since this. This code collected from Twitter and inserted into a database Writing to a database could be slow for streaming data, so I had one process stream + insert into a queue, and another read the queue and do inserts It worked, but complicated and hard to debug

almost hard to remember when the center and right were having a years-long meltdown about free speech on campus because students protested Milo Yiannopoulos speeches

Four (categories of) metrics that can get you SO far in text analytics: pointwise mutual information, entropy, Pearson and Spearman correlation, KL and JS divergence. These are mostly one-line Python calls or just rely on counting and dividing things.

Can't count the number of times I've cited this and then followed up with "As @foucaultwelles.bsky.social and @sjjphd.bsky.social show, there's not a single networked public sphere..."

I was thinking about writing a Bluesky data collector for researchers (like the old Twitter API) There's no "search posts" endpoint, but I thought you could replay the firehose / jetstream starting from a certain time and filter from there. Looks like it only goes 72 hours into the past though?

v1.108 is rolling out today 🚚 Now live, at long last: Bookmarks, aka Saved Posts. For all those posts you'll definitely plan to come back to! Update the app and give it a try. The button is right down there 👇

🚨 UPDATE on the current COVID vaccine access situation in the US. Please note that this is based on state laws but different things may be happening on the ground, and some pharmacies are defaulting to federal guidelines instead of state laws.

At long last, I can post my team's summer project: applied modules to teach how ML/AI tools are changing social science and humanities research: ubcecon.github.io/praxis-ubc/ Highlights: LegalBERT to analyze anti/pro-immigrant sentiment in 19th c. BC law: ubcecon.github.io/praxis-ubc/d... 🧵1

Three weeks to go until the Trust & Safety Research Conference #TSRConf at Stanford! I'll be moderating a lightning talks session on day 1. Looking forward to connecting with experts across research, policy, and industry. 🎟️ Last chance to register: sto.stanfordtickets.org/the-2025-tru...

Bluesky, which is a very tiny dying network that no one cares about and is just a liberal bubble, is so powerful that it's going to destroy Democrats' chances of retaking power. - by Nate Silver

Please provide your input on @blackskyweb.xyz Community Guidelines (use of the feed, moderation, app, as well as developer/network services). Anyone using our feeds, subscribed to our mod service, on the PDS/relay Provide input via the following link 🗳️: assembly.blacksky.community/6carwc4nzj

Great news! @graze.social is delighted to announce that we now support show more / show less! You can use this *right now* to threshold your feed by what readers are actively flagging, and access a summary of reader actions taken on articles to learn what's working and what isn't!

Tech Policy Press is now accepting applications for our 2026 fellowship cohort. We're looking for journalists, researchers, and policy professionals to pursue independent reporting or analysis on urgent questions at the intersection of technology and democracy. Learn more below!

can't believe this was 9 months ago and nate's still at it

A lot of public sphere theory is based on voice, but the necessary counterpart to that is listening I'd love to see more work on "networked listening publics," especially empirical work. It's hard because "listening" signals are often on the side of the platform (who viewed what)

A lot of the big complainers about Bluesky are basically describing Threads but their real issue with Bluesky is that they specifically dislike around 12 people who they could ratio on Twitter if it still existed. It’s basically niche influencer drama pretending to be about policy

The prevalence of cross-platform meta discussions ("That site is dying," "It's not worth posting on that platform," etc) from the post-X fracturing is fascinating Touches on aspects of deliberation, community, identity, performativity, growth, safety, so on

I am perfectly happy to exist on a much smaller social media site and not have to deal with Grok being MechaHitler or my feed being inundated with constant garbage from Nazis

atproto Subreddits atproto Facebook Groups atproto Usenet Groups Lemmy for atproto These are analogies you can think of for what I’m trying to do with Blacksky Algorithms long term.