New research on Twitter (right before rebranding to X) shows right leaning accounts got outsized visibility from the algorithm & suggests this bias stemmed from different content production techniques (more sensational, outrage-bait) and from direct interaction w/ the new king of the platform: Musk.

Also this: marcusarvan.substack.com/p/anthropics...

"ai is inevitable" is a high accuracy indicator of "I am a sucker"

I've been overwhelmed this year by education news. I've hidden from some of it by focusing on one of the newer biggest threats to public ed: AI in K-12. But a whole lot of ish happened. I went month by month to look at key stories from the year and it ain't pretty. We've got work to do in 2026.

“AI companies use the future to justify the present, as though speculative benefits should outweigh actual, ongoing damages. This temporal shell game, destroying the world to save it, would be comedic if the consequences weren’t so severe.” — @jamesosullivan.bsky.social #agi #superintelligence #ai

Part of Lectures on Tap, an event series that turns bars and venues into makeshift classrooms. Attendees hear thought-provoking talks from experts on wide-ranging topics. Tickets sell out "within less than an hour". Would love to a revival if "communiversities" fostering liberatory education.

Today, DWI publishes a powerful essay by @mgasia.bsky.social, who works impersonating an AI sex companion: data-workers.org/michael/ This piece offers a rare glimpse into the psychological and economic realities behind one of the fastest-growing sectors: AI-assisted intimacy.

Happy Stylish IKEA Monkey Day to all who celebrate

My cat overlords go through 8 cans of food a day Who wants a Purina Peace Prize?

At "Autonomy Visualized" at a Tesla store in Miami, see a five second video clip in this story, where the remote operator for an Optimus robot takes off their headset, mindlessly reproduced by the Optimus robot followed by falling over backwards. Oh dear. electrek.co/2025/12/07/t...

Question for folks rolled up in the Anthropic content theft settlement: How reliable is the database linked to in the email? There is a book listed in the database that was authored by my internet doppleganger but not my books that were pirated on LibGen.

It’s that time of the semester when you pause to consider: Is one party-sized can enough to get me through the next week?

The university-surveillance component of this piece is !!! I had missed the Qualtrics shift and acquisition and wow. We really need an ed-tech key players power-map, does anyone know if this exists?

I’ve summed up and contextualized all of these critical AI pieces & more in one place — “From AGI to Workslop: What I Read About AI in 2025.” mail.cyberneticforests.com/what-i-read-...

My Huguenot ancestors arrived in Haarlem, New Amsterdam in 1650 after being driven out of Amiens, France due to religious persecution. They were not citizens nor were they naturalized immigrants because there was no USA in 1650. Would ICE question my claim to citizenship? Seems like it.

NEW - The Data Doesn’t Lie: How ProPublica Reports the Truth in an Era of False Claims As federal data becomes less available, our editor-in-chief @sengelberg.bsky.social explains how our journalists are providing readers with precise numbers.

One might think 52 years after the infamous Lenna Playboy image was included in a dataset for testing image processing algorithms, that this 🤬 field would learn something about consent, sexism, and abusive displays of power over research "subjects".

AI Is Not Inevitable join AAUP for a conversation with educators, educator unions, and the Collaborative Research Center for Resilience zoom.us/webinar/regi...

Familiar ed-tech pitch on GenAI: we will save you time to focus on what really matters, which is, of course, the relationships you build with students. But that pitch is cover for what's happening in classrooms, and as this excellent 🧵 makes clear, these products are damaging at multiple scales.

Told ya. Thread here: bsky.app/profile/kier...

h/t @robertfreundlaw.bsky.social holy shit, an accurate legal critique of LLMs. LLMs don't reason because they're just stitching together plausible-looking sentences indifferent to the content

“This technology that is ruining education has incredible potential to save education, if only I could figure out how to use it better to improve my students’ education.” Somehow missed this @mcsweeneys.net gem from September. ht: @brianrbowen.bsky.social

With a logic so circular I swear ChatGPT wrote this. We need AI literacy so we can assess AI literacy. But no exams! Bleep bleep bloop. www.nature.com/articles/d41...

An excellent and brutal article. I don't think the situation is quite as hopeless as some of this suggests, but the challenges are enormous. Our only chance, as ever, is to hold together: bring faculty, well-meaning administrators, and especially students together: human learning for human beings.

Up next in our series on drivers of the affordability crisis: technology. I wrote about AI bots are colluding with one another, anticipating consumer choices, and transferring wealth to the businesses that employ them. "That's why we're building all these data centers." 🧵

This is a terrible story--both because of the trap that was set (either in advance or ret-conned to seize a moment) for this graduate student but also because the student's department failed her. The 1st problem is easy to see, but I want to explain the 2nd one. 1/x

Re-upping this today, citing the concluding paragraph we wrote as I just responded to a frustrating survey about integrating GenAI into humanities education at my institution.

Come for the drug kingpin oligarchy pardon* of the ex-Honduran president Stay for the techbroligarchy sovereign network states in ... wait for it ... Honduras I feel like I'm taking crazy pills . . . *The massive drug trafficking case was prosecuted by none other than then-DOJ lawyer Emil Bove

Here's a (scholarly) rant about A.I. and the seeming paradox Asian Americans as complicit with it that me, Michelle Huang, and Tara Fickle wrote as part of a DISCOnetwork writing jam. Is A.I. the next Iron Chink? medium.com/@hatlabnorth...

The supposed use of AI (a proprietary tool) is for public opinion measurement, and as someone who follows that side of the AI industry closely, I simply do not believe that it can do what it purports to do at least based on the claims here

Elite tech industry mania for throwing $100s of billions (or trillions?) at so-called "AI" was a high-risk bet (which doesn't seem to be panning out) that more & more & more $ would make an AI superintelligence to solve all humanity's problems & create utopia. It's like "The Rapture" for techbros.

it is amazing how so much tech gets unremittingly worse year on year at whatever you once used it for and less pleasant to use and more bound up with creepy politics as well as personal and environmental harms and yet also still less productive of actual goods and these are somehow great visionaries

I call this a technology of coercion because it is almost entirely an upper-elite phenomenon being forced on unaccepting organizations, which frames how I approach “might as well learn how to use it” discourses

This blog post by Helen Toner (well-known by some for a stint on the OpenAI board) is *really really* interesting because it is the first thing I've read by someone in Frontier AI that starts to get to grips with sociotechnical concerns open.substack.com/pub/helenton...

The data rights hill that I will die on: If content (computer-generated or otherwise) is predicated on theft, the resulting output is not ethical, legal or moral. Full stop.

I joined @cwarzel.bsky.social on the Galaxy Brain Podcast to explain some of the reasons why America, and a lot of other Western democracies, are sliding into authoritarianism and why it's probably inevitable www.youtube.com/watch?v=HvFh...

Introduce yourself with five concerts you've seen — Part 2 The Bridge School Benefit 2010: Neil Young Buffalo Springfield Elton John Leon Russell Elvis Costello Grizzly Bear Jeff Bridges Kris Kristofferson Modest Mouse Neko Case Pearl Jam Ralph Stanley T Bone Burnett

Introduce yourself with five concerts you've seen — Queen (Night at the Opera tour!) Talking Heads Sarah McLachlan Boz Scaggs KD Lang

How come the policy obsession with "evidence-based" policy and technology in education of at least the last 20 years has completely evaporated with "AI"? ...

don’t cheapen your slides! there *are* places to get free, human-created, rich, visual content on this bitch of an internet, some on this list: livelaugh.blog/posts/non-ai...

I don’t know if anyone else notices or cares, but when I see a presentation in which the speaker uses obviously generated-AI images to illustrate their slides, it makes me immediately less confident in whatever other content they’re presenting.

I know a lot of people like to dump on Facebook. But let me make an IRL argument that, in these particular times, the platform is a counterculture goldmine. Come along for a story about how a basic community page can shift to hyper local news sharing and mutual aid-centered to reach everyday folks.

Friendly reminder to you Black Friday shoppers: AI is for surveillance Chatbots are for surveillance Wearables are for surveillance Your data/clicks/behavior/content is collected and analyzed to be used against you in decisions re: hiring, insurance risk, credit cards, loans, and law enforcement.

I teach computational data science, not history, but I found the examples helpful and extensible to my own class prep and formative assessments. Thanks for sharing!

“When we critique AI, we should do so with intellectual honesty and in a principled way. Utmost care is needed to avoid ethics washing, greenwashing, and generally — what we dub — critical washing.”

In light of record submission rates and a large volume of AI-generated slop, SocArXiv recently implemented a policy requiring ORCIDs linked in the OSF profile of submitting authors, and narrowing our focus to social science subjects. Today we are taking two more steps: /1

Not one mention of eugenics.

The problem of returning a "single answer" is that an AI overview is unable to grapple with fundamentally human ways of knowing: uncertainty and ambiguity. Smoothing out these necessary frictions with Gen AI chatbots outputs and AI overviews from search retrieval is dangerously reductive.

BROOKINGS: Our survey results lend themselves" to some conclusions: ".. professional AI use is far from ubiquitous and many respondents expressed skepticism that it would be as revolutionary as some experts expect." @brookings.edu www.brookings.edu/articles/how...

When is the last time that you saw a hypothetical situation described in a Terms of Use agreement? Doesn't this admission provide evidence that OpenAI knew that people were seeking self-harm information from ChatGPT *and* that the LLM would return dangerous outputs with no apparent guardrails?