We should have an academics who love @tonytula.com fanclub @nathanihoff.bsky.social and I am interested.

Melanie Walsh@mellymeldubs.bsky.social · yesterday

”Which is why I loved Twitter: an open, rhizomatic forum where you could aggravate existing mental illnesses, shop for new ones, violate your Miranda rights, and get fired.“ —Tony Tulathimutte, Rejection

And now this: the winner of a major prize for Spanish literature. (This is my translation of the first paragraph of the article; it doesn't seem to be up in English.) elpais.com/cultura/2026...

Vauhini Vara@vauhinivara.bsky.social · 2w ago

In case English Bluesky missed it, El País found that tons of self-published Spanish books — including a bestseller by the alt-right politician Alvise Pérez — seem to include AI text. So do a much smaller proportion of traditionally published books.

I get 5+ emails *a day* from prospective PhD students. I have seen some advice going around about what to write in these emails. Here's my piece of advice: don't waste your time. Or your tokens. #phdadmissions Or, if you really insist...

So now the Public Library Association — part of the American Library Association — is "partnering" with Microsoft on "AI literacy and adoption."

Microsoft and the American Library Association expand AI literacy

Microsoft and the Public Library Association launch free AI literacy training for library workers and patrons, helping trusted local institutions lead on AI.

blogs.microsoft.com

Science is inherently political and when our government is actively hostile to science as a whole, publications should be able to write about it without being censored. Alas. www.thenation.com/article/soci...

The Fall of “Scientific American”

When the beloved science publication got too political, its corporate owners stepped in. Now its future is uncertain.

thenation.com

Voters who Googled how to donate to Talarico got an AI answer that led them to give to a different senator. A network of fake news websites seeded messaging about state & local candidates that AI answers picked up on. My story on how the AI midterms are going: www.theatlantic.com/technology/2...

America Is About to See What Happens When Voters Turn to AI

Chatbots are adding a new layer of chaos ahead of the midterms.

theatlantic.com

Would love to see our American free speech organization stick up for Sally Rooney... Alas.

In case English Bluesky missed it, El País found that tons of self-published Spanish books — including a bestseller by the alt-right politician Alvise Pérez — seem to include AI text. So do a much smaller proportion of traditionally published books.

How much AI is in Spain’s bestselling books?

EL PAÍS analyzed 175 popular titles using the Pangram detector. There was no trace of AI in 2019, but now one in three self-published bestsellers is written mostly with artificial intelligence, includ...

english.elpais.com

Really interesting research that gets at some of the challenges to AI detection that I covered anecdotally in my Businessweek profile of them!

Kevin Ren@kevinren09.bsky.social · last wk.

Excited to share my first PhD paper, “Hitting a Moving Target: Test-Time Adaptation for AI Text Detection under Continual Distribution Shift” (accepted to #NeurIPS2026)! We frame AI detection as a continual learning problem, robustifying models to shifting human/AI writing. (1/N)

Very glad to have @kevinren09.bsky.social 's first PhD paper out. AI text detection (aka Pangram) is very impressive. But when do we expect it to work? We warn that distribution shift is [still] a major issue -- (1) performance suffers in adversarial settings and with new LLMs

Kevin Ren@kevinren09.bsky.social · last wk.

Excited to share my first PhD paper, “Hitting a Moving Target: Test-Time Adaptation for AI Text Detection under Continual Distribution Shift” (accepted to #NeurIPS2026)! We frame AI detection as a continual learning problem, robustifying models to shifting human/AI writing. (1/N)

"The 'shut up and let this happen' of it all could give you whiplash." My latest rant about AI is up at Lit Hub. lithub.com/why-are-publ...

Why Are Publishers Suing Big Tech With One Hand and Embracing AI With the Other?

Last weekend I had a little stagefright before I was set to appear on a panel about authorship and AI for an event called Whose Truth? Writers and Readers Forum presented by the Authors Guild. I ha…

lithub.com

I’m giving a talk at Kenyon tomorrow — come one, come all, by the way! — and the reading material at the hotel is ISSUES OF THE KENYON REVIEW. I AM BASICALLY ROOSTER. This is the best.

I wrote about how the winner of another major literary prize appears to be AI and why we need to do something about all this. 😢

Yet Another Literary Prizewinner Seems to Be AI

The prestigious Caine Prize, announced last week, went to a story with many red flags.

theatlantic.com

🆕 📝 New paper officially out in @nathumbehav.nature.com! In 2023, Change.org built an AI-assisted writing tool into their platform. We offer causal evidence that access to the AI tool changed petition text (word count, language)... but may have actually hurt outcomes (signatures, comments).

This is a really fascinating research demo, from the lab of Mohit Iyyer at the University of Maryland: It predicts the likelihood that the ideas contained in a piece of text are AI-generated versus human-generated. A couple of examples in this thread.

IdeaLens: is this text human or AI?

We check a text's ideas and prose separately.

ideadetector.ai

Over the weekend I was on a panel about AI in books and I said that publishers have always been so careful about fair use -- you can use maybe a line or two from a song but never any more -- and then the tech bros were like, we are entitled to all books. They belong to us now. Fuck you.

Brilliant article from @maris.bsky.social . The first thing I read this morning.

Maris Kreizman@maris.bsky.social · 3w ago

I am a cynical person but I am pretty earnest when it comes to literary prizes. Books need all the help they can get these days, and prizes are one way for unheralded books to get their due. lithub.com/in-defense-o...

I visited this school as part of my reporting last year for a Businessweek story, which also revealed that Alpha is actually a subsidiary of the tech company Trilogy. Here's what I found.

Aaron Rupar@atrupar.com · 4w ago

Linda McMahon: "AI is definitely a tool used correctly. I've seen it in action. I have visited the Alpha School in Austin, Texas, where they use AI for the first two hours every single morning. The students have their headsets on, they are watching their computers ... it is incredibly effective"

Some aspects of the methodology and strength of the results are disputed (see thread), but it's not plainly true that filters prevent extraction.

Tuhin Chakrabarty@tuhinchakr.bsky.social · 2mo ago

Anyone interested in verifying is more than welcome to go to our demo and see for Coraline ( book with 85-90% overlap) how much each paragraph is regurgitated across every 100 generation. Demo here cauchy221.github.io/Alignment-Wh...

In response to that putrid Bret Stephens column, brilliant @reluctantlyjoe.bsky.social made a whole Google Doc with buy links for all of the @nationalbook.bsky.social Prize nominees. docs.google.com/document/d/1...

Bret Stephens is mad at these books, you should buy them and read them

Bret Stephens is mad at these books, you should buy them and read them, not just the SCARY PR COPY that triggered him like a snowflake. No, I won’t link this fuckshit here, because I won’t drive tra...

docs.google.com

Of course they all fucking knew! www.publishersweekly.com/pw/by-topic/...

Unsealed Files Show Open AI/Microsoft Knew Copying Was Illegal and Could Hurt Authors

Newly unsealed comments made in briefs filed by Open AI and Microsoft reveal that executives knew they were illegally training their products on copyrighted books and that AI generated books would lik...

publishersweekly.com

Weird, this in-depth profile of Pangram doesn't focus at all on its founder's suspicious burping and food-microwaving habits. Instead there's just a bunch of stuff about how they built it, how it works, whether it works and why it matters www.bloomberg.com/news/feature...

The World’s Most Reliable AI Detector Has a Human Problem

Armed with a near-perfect detection tool, Pangram is fighting to convince the public AI cheating can be caught.

bloomberg.com

Great profile of the leading AI detector, Pangram, and the people behind it—including a former English major who wrote a thesis on Ishiguro (me too). Fascinating snapshot of our cultural moment with AI, and a sober reflection on the future of the internet, teaching, and human communication.

Vauhini Vara@vauhinivara.bsky.social · last mo.

I profiled the startup behind Pangram, the AI detector everyone’s talking about. It works really, really well. I wanted to understand how they did it. and also: why people still don’t trust them. (also, news: Penguin Random House is using it.) gift link! ✨

I profiled the startup behind Pangram, the AI detector everyone’s talking about. It works really, really well. I wanted to understand how they did it. and also: why people still don’t trust them. (also, news: Penguin Random House is using it.) gift link! ✨

The World’s Most Reliable AI Detector Has a Human Problem

Armed with a near-perfect detection tool, Pangram is fighting to convince the public AI cheating can be caught.

bloomberg.com

I trust no one but @vauhinivara.bsky.social to come up with something so good. It’s brilliantly researched and engages thoroughly with the space of AI detection. There is a funny bit in the article that talks about how I got introduced to @pangram.com and the rest is history !!

Vauhini Vara@vauhinivara.bsky.social · last mo.

I profiled the startup behind Pangram, the AI detector everyone’s talking about. It works really, really well. I wanted to understand how they did it. and also: why people still don’t trust them. (also, news: Penguin Random House is using it.) gift link! ✨

Francesca Mancino talked to me for this piece, along with some very smart thinkers on the book, its history, and its digitization @mmvty.bsky.social @mkirschenbaum.bsky.social @leahprice.hcommons.social.ap.brid.gy

The New Yorker@newyorker.com · 2mo ago

Anthropic is trying to “destructively scan all the books in the world.” How worried should we be? www.newyorker.com/culture/the-...

Brilliant read in todays @newyorker.com essay on utility of books in LLM pre-training with quotes and inputs from @mmvty.bsky.social and @mellymeldubs.bsky.social

BMC is a pooled metric should not be surprising ? It was explicitly mentioned. k=5 is deliberately sensitive and does contribute to BMC but central extraction doesn’t depend on 5 word matches. Some books have 24-69% of entire text covered by exact spans > 50 words and single generations > 100 words

A. Feder Cooper@afedercooper.bsky.social · 2mo ago

Prospective copyright plaintiffs have started asking me my opinion about citing "Alignment Whack-a-Mole" in litigation. It reports that fine-tuning makes frontier LLMs reproduce up to 85-90% of copyrighted books. I don't think the headline results hold up: afedercooper.info/whack-a-mole/

👇👇👇

Will Oremus@willoremus.com · 2mo ago

When I wrote about Daggermouth being likely AI-assisted, I used Pangram as one piece of evidence among several, and I stand by that. Even then I was clear it can't be proven — just highly likely. One of the characters is named Elara for god's sake www.theatlantic.com/technology/2...

It's wild to me that the Daggermouth author and publisher are still standing behind that one.

Will Oremus@willoremus.com · 2mo ago

I kept reading this article excitedly waiting for it to deliver the goods on why we can't trust Pangram and it turned out to be mostly that the founder seemed socially awkward? www.wired.com/story/pangra...