Why do the Japanese like their buns askew (2026)

We should absolutely be fighting enforced adoption of AI products and the displacement of civic infrastructure with privately owned tech platforms, but amnesia about this as an ongoing political and economic project since at least the 1990s is not the way.

"Defunct startups are being liquidated for their Slack archives, Jira tickets, and email threads—operational exhaust that AI labs now treat as premium training data."

"This was an entirely unforeseeable usage of the tombstone functionality of Oregon Trail. If anything, our zeal for the social sharing of sensitive medical information was way ahead of its time."

yeah, seems like you're definitely headed for "transformation"

Robo-dogs and ghost sharks grab headlines—but they’re just the spectacle. Behind them lies a vast system of surveillance, data extraction, and automated exclusion shaping modern border control, writes Tech Policy Press Fellow Petra Molnar.

Sal Khan is spinning up an alternative post-secondary credential in concert with the largest tech companies in the world. Anyone who cares about human-centered education should wish it nothing but failure. Thankfully, Khan has practice at that. www.insidehighered.com/opinion/colu...

I love silent updates that introduce new features I didn't ask for, that pop-up all over the interface, and that I can't turn off.

OpenAI’s chief revenue officer sent a 4-page memo to employees on Sunday about the company’s strategic direction, emphasizing the need to lock in users, build a moat and grow its enterprise business. (It also threw shade at its longtime rival Anthropic.) www.theverge.com/ai-artificia...

I generally admire Ronan Farrow’s reporting, but I’m not sure I read anything in the Altman piece that wasn’t just more confirmation of Karen Hao’s in-depth analysis in Empire of AI, i.e., ‘Everyone he knows says he’s a lying prick who will say/do literally anything for power’ Did I miss the reveal?

I really can't believe how we don't talk about the incredible amount of bias baked into modern models nearly as much anymore, when egregious examples are _everywhere_.

Another really smart blog post by @nobugsnous.bsky.social that explains why abstinence-only sex ed is a shit metaphor for understanding generative AI refusal.

New post on Refusing Generative AI in Writing Studies! I wrote a blog about why abstinence-only sex education is a bad metaphor for generative AI refusal. refusal.blog/2026/04/07/w...

Just have fun with your kids, people. No need to get Meta involved.

If you find this story disturbing, and I do, you should feel similarly about using LLMs for editing advice or essay grading, etc. Both are fundamentally simulations of human response untethered from the way humans respond.

I don’t know what saving either the profession or the discipline means, if it doesn’t involve preserving pathways to full careers. If the institutions best placed to do so, however troubled, are not willing to make that commitment, then it is delusional to think those less well placed will step up.

Trillions of dollars have been spent to create an artificial workforce composed of the kind of employees who if they were human would probably be fired within a year or two of being hired. Imagine if that money had been spent getting humanity off fossil fuels instead.

GenAI being used to “fix” to systemic issues is a landlord painting over cracks in the foundation.

I did not write a new blog about tech journalists using AI but let me leave this from last summer right here www.404media.co/the-medias-p...

I wrote about this very thing today www.thehandbasket.co/p/refusing-t...

prescient

I think more than anything else AI feels like a miracle to people with questionable literacy because it can read and write for you, and a lot of us who are highly literate underestimate how many of our society's leaders, esp in business, struggle with literacy.

I do think efforts to make tech systems more politically transparent are generally good. But at the same time, I am so tired of pretending that computing technologies cannot or should not be regulated more directly. The responsibility for avoiding harm is always passed on to the users.

I’m fucking crying dude. Asking “pretty please do not introduce a security vulnerability”.

"The real threat is a slow, comfortable drift toward not understanding what you're doing. Not a dramatic collapse. Not Skynet. Just a generation of researchers who can produce results but can't produce understanding."

It is really, really wild how persistent and prominent this remains in common LLMs. There seems to be an air of "well they fixed that" and it is 100% not true: most of the energy just seems to be going towards 'being plausible enough to be accepted', ie style and not accuracy or truth

My university: "Copilot is Microsoft's AI-powered productivity service that uses large language models (LLMs) to help you create content, analyze information, summarize documents, and complete tasks more efficiently." Microsoft: LOL

"AI sycophancy is not merely a stylistic issue or a niche risk, but a prevalent behavior with broad downstream consequences. Although affirmation may feel supportive, sycophancy can undermine users’ capacity for self-correction and responsible decision-making"

part of the pollution that's happening to our information ecosystem is that it becomes harder to actually be responsible like somebody says "oh, i don't trust ai. i'm going to find a reputable source, like an actual medical clinic"

“But after tens of billions of dollars of school spending…studies have found that digital tools have generally not improved students’ academic results or graduation rates. Some researchers &organizations like UNESCO even warn that overreliance on technology can distract students and impede learning”

I think we need to celebrate the death of Sora a bit more. This is a technology that, just MONTHS AGO, we were being told was going to literally destroy Hollywood and Disney was going to give them a BILLION DOLLARS and NONE OF THAT EVEN REMOTELY HAPPENED

The main goal of Silicon Valley appears to be finding an ethical way to have a slave

What is marketed as AI “innovation” is, in practice, a consolidation of global labor hierarchies and a commodification of African knowledge, write Marché Arends and Kathryn Cleary, drawing on their year-long investigation supported by the Pulitzer Center and published by Africa Uncensored.

after readers & ai detectors flagged ai suspicions in a modern love piece, the author told me she "did utilize ai as a tool”—using five chatbots for inspiration, guidance & correction. 🤯 research suggests this is happening in big papers more than we realize. [GIFT LINK: @theatlantic.com]

No.

Meta and YouTube were both found liable by a California jury for harming a plaintiff's mental health with the design of their products. A massive ruling in a bellwether case that will open the floodgates for other lawsuits

You need to read this. Sweet sweet schadenfreude.

"If the data centers we rely on are built on the infrastructural remains of colonial and imperial ideologies, we must interrogate how these spaces and histories shape our engagement with the newest manifestation of these ideologies."

I know not everyone is guilty of using arXiv to avoid peer review, but... I guess its good they have some small standards there. (And yes, I know there's a lot of problems with academic publishing and peer review, but that's a separate discussion)

LRT: Got a paper coming out soon (written w/our incredible WPA, Ann Dean, who isn't on here) surveying student views of AI. We found students were pretty ambivalent about it & also their understandings of what it does are shaped mainly by marketing. Important to provide space for resistance/refusal.

An unaccounted for part of the economy is how much young people virulently hate AI, despite how aggressively it's being forced on them. They realize it's making their friends dumber and ruining the world and they want nothing to do with it. From the Penn student paper: www.thedp.com/article/2026...

"You can't put the genie back in the bottle"

This is the head of product development at Canvas essentially declaring that the choices Canvas makes will determine the shape of classroom instruction. This is not happening in partnership with professors or students or institutions. This is a profit-seeking 3rd party.

I encourage people to look at what happened to the profession as a whole through the process of adjunctification, a mechanism which degraded the quality and security of ALL faculty jobs, even those who are tenure track and see how AI automation is the same mechanism, only potentially faster.

The friction of actually trying to understand something is annoying and takes practice to push through, and a degree of introspection to say “ah man I do get it.” Learning is a skill! It takes practice and also is frustrating as hell.

Anyway, AI is a managerial technology that came out of fields where researchers act more like managers than researchers and this is not a coincidence. It’s fundamentally related to why certain people jumped at the chance to try to build certain kinds of systems that devalued (other people’s) work.

I have refrained from commenting on genAI as a scholarly transcription/translation tool because (1) it raises a different, if overlapping, set of issues than genAI as a classroom tool and (2) I’ve heard good things from credible judges. But this seems like a serious problem, hallucination and more

My fear: the work that will need undoing after this is unfathomable, and the resources to undo it will be gone. Consider how far a fake quotation spreads before it’s caught, and how rarely a correction happens even then. The legacy of slop scholarship is going to be several orders of magnitude worse

“It’s inevitable” “you can’t roll it back” “you just have to accept it”

If you follow edtech, please read this LinkedIn note from a Google Admin about Google forcing Gemini on K-12. CC @civicsoftech.bsky.social www.linkedin.com/posts/kate-s...