Preloader

Community

Join WIRED@Night for an Uncanny Valley Live Recording on Women, Tech, and Power

What is the state of women in Silicon Valley?

In a word: shaky.

This fall, WIRED is making a special issue devoted to gender and power. For the live-audience recording of the Uncanny Valley podcast in San Francisco, WIRED will dig further into the fascinating and nuanced state of affairs. First up: a special conversation between Paulina Borsook, tech journalist and former WIRED writer, and WIRED global editorial director Katie Drummond.

Borsook, whose prescient 2000 book Cyberselfish is being rereleased this fall, has long been a skeptic of the tech world—which might have been to her detriment during the growth mindset of the past few decades. Come for a lively conversation about WIRED, then and now, and what Borsook thinks of the current trillionaire class.

Then, the podcast panel, featuring executive editor Brian Barrett and contributing editor Zoë Schiffer, will bring on WIRED senior correspondent Lauren Goode to talk about her quest to find the most powerful woman in Silicon Valley.

EVENT INFO
WIRED’s Uncanny Valley, featuring Cyberselfish author Paulina Borsook. Copresented with KQED.

Thursday, October 8 | 7 pm (Doors at 6:30 pm)
The Commons at KQED in San Francisco

General admission: $25 $15 for WIRED readers. Buy tickets here.

Subscribers save $10 per ticket—plus all the other exciting benefits of a WIRED subscription. Use the discount code “WIRED” at checkout.

Not able to attend but want to stay up-to-date on WIRED’s next event? Sign up here.


Source: Culture Latest

‘I Felt Super Violated’: Men Wearing Meta Glasses on Dates Is an Instant Red Flag

Courtney McAnuff was out of a year-long relationship in the summer of 2024 and starting to date again when she agreed to grab drinks with a potential suitor.

They met at a bar in midtown Manhattan. He arrived wearing a baseball cap pulled low and sported a pair of black-framed glasses, she says. As conversation flowed, from discussions about navigating life in New York to past relationships, he mentioned that his glasses were Meta Ray-Bans.

“I wasn’t familiar with them,” McAnuff, a 28-year-old project manager who now lives in Toronto, tells WIRED. It was her first time seeing the smart glasses up close, which can record images and video from the wearer’s viewpoint. He let her try them on and even showed her a couple of videos that he’d taken previously, she says. “That was it. We stopped talking about it after that. I figured that was the end of the use.”

The next day, checking her Instagram, McAnuff noticed she’d been tagged in one of his story posts. To her surprise, it was footage of them at the bar and on the subway together, en route to a basketball game uptown. In the video, McAnuff is captured from an above angle, sitting across from her date who is holding a drinking glass; WIRED has seen screenshots of the footage. “I felt super violated,” McAnuff says. “Like, what if I don’t want everyone knowing what I’m up to? He felt so free to film me and post it. There was no conversation about it.”

Meta sold more than 7 million pairs of the Ray-Bans in 2025, as its smart glasses have infiltrated public spaces thanks to pranksters with dreams of viral fame, some of whom even film people while they’re at work. The phenomenon, which has also spread to pickup artists hitting on women, has earned the hardware the nickname “creep glasses.”

Among singles, and particularly women, Meta’s AI-powered eyewear has become a recurring nuisance, as people share a spreading unease about being covertly filmed on dates.

“Ban those glasses completely,” posted a young woman on Bluesky in August, writing that her date “leaned over to show me pictures he took of me I had no idea he was taking.” One TikToker claimed in a video after a date that she “immediately felt uncomfortable,” questioning why people are “so comfortable videoing everything. It’s a terrible way to portray yourself, wearing a fucking camera on your face and you’re sitting in my face.”

In a Reddit post from August, users debated the ethics of the trend, wondering if it was OK to continue a date when someone showed up wearing them; in a separate thread, one person said they had ghosted a guy after realizing he’d worn the glasses on their date. “I didn’t notice until toward the end of the night. I kissed him while he was wearing the glasses,” they said. In July, a man was put under investigation in Seoul, South Korea, for allegedly filming four women he had gone out with and later uploading clips to social media, according to The Korea Herald.

In many regards, dating has been reframed by our dependency on technology. Apps connect people. Chatbots are used for all sorts of things these days: as wingmen and for role-play. Smartphone cameras capture important moments together. But while a phone being held up to someone’s face gives them a clear opportunity to object, smart glasses largely erase that opportunity. (Meta’s glasses do have a front-facing LED that switches on when the glasses are recording, but the light is the only warning, and it can be disabled by hacking the hardware.)

For McAnuff, the trend is about more than privacy instructions. She was worried about what would happen with the footage afterward. “He didn’t ask if I wanted to be recorded or if I wanted my face uploaded. And now he has this video file of me forever. It’s definitely a big trust issue,” she says.

In a 2025 YouGov survey of US adults focused on factors contributing to successful relationships, trust ranked the highest, with 94 percent of respondents categorizing it as “very important.”

Gary Lewandowski, a psychology professor at Monmouth University whose research focuses on the science of romantic connection, says adding covert cameras to the mix will only make building trust harder. He says beyond posting for the ’gram, a number of factors could be driving men to do this. The most logical, non-creepy reason may be tied to post-date analysis, in that the glasses can help men identify blind spots around their own behavior.

“A lot of men out there are eternal optimizers. They want better, more efficient ways of doing things, so reviewing game tape allows them to learn from mistakes,” Lewandowski says. “But I do wonder if some are doing it to build a case instead of a connection—why they’re right and the other person is wrong—or to help better identify red flags.” That type of surveillance “makes the date less about getting to know each other and more about fact gathering,” he adds.

The trend has also reanimated discussions around consent when it comes to dating. Yue Xu, cohost of the Dateable podcast, believes new social norms are needed to keep up with our evolving tech. “It’s bringing in a third party, which is disturbing. It just shows that this person lacks respect for the integrity and privacy of the date,” Xu says, adding that she fears that the glasses are triggering a future “where people have no more anonymity.”

Concerns have intensified as issues about the device have continued to make headlines. Meta and EssilorLuxottica, which owns Ray-Ban, are facing a class action lawsuit alleging that they captured and stored users’ private data without consent. This month, Semafor reported that Meta disabled the recording function for thousands of users who tampered with the light that notifies people that the Meta Ray-Bans are filming. Following the addition of an unreleased face-recognition feature called NameTag to its eyewear earlier this year, revealed by a WIRED investigation, Meta scrubbed the code from its app in June and is facing another proposed class action over allegedly breaching Illinois and California privacy laws by collecting biometric information from users’ photographs without their knowledge.

“This lawsuit is without merit and misrepresents our work,” a Meta spokesperson tells WIRED in reference to the NameTag lawsuit. “We’ve been transparent about how we use people’s information to build and improve our AI products. As for NameTags, nothing has shipped to consumers, and no final decision has been made on what to do here, if anything. If we do decide to roll something out, we will take a thoughtful approach and do so with full transparency. One decision we can be clear about—we are not building a universal face database.”

EssilorLuxottica did not respond to a request for comment.

According to Meta, if people tamper with the recording light, including placing blockers over it, the camera will not turn on. Additionally, the company says it removes accounts and content that violate its policies when it becomes aware of them, including content meant to degrade or shame others. In August, Meta expanded its public awareness efforts through a social media campaign focused on helping people identify when someone is recording with the glasses.

“We don’t want people to be surreptitiously taking videos of other people and harassing them and then posting them on our platform. So we’re trying to fight that every way we can,” Meta’s head of Instagram Adam Mosseri said in July.

That hasn’t stopped regulators around the world from pushing to enact harsher privacy laws for consumers. A California bill just passed both chambers that mandates visible recording indicators on recording devices and bans users from turning them off. Likewise, Australia’s eSafety Commissioner is asking tech companies to automatically blur people’s faces recorded with their smart glasses, writing that “as smart glasses evolve and become more common, they may become increasingly interconnected with and create new ways to perpetrate different types of harms, such as image-based abuse, adult cyber abuse, cyberbullying of children, or the livestreaming of violent or other harmful content.”

When McAnuff realized footage of her had been uploaded without her approval, she confronted her date over text. He apologized for his actions but she says that “he didn’t think it was that serious.” She hasn’t seen him since.

For McAnuff and many others, Meta’s smart glasses have simply become the latest privacy issue hiding in plain sight.

“Being in each other’s physical space is such a beautiful experience in itself. We don’t need to have gadgets around,” McAnuff says. “I don’t think that every moment needs to be captured.”


Source: Culture Latest

If the US Collapses Soon, Don’t Blame Emily St. John Mandel

If the United States collapses in the near future, you’ll no doubt be tempted to point a few fingers. Just don’t blame Emily St. John Mandel.

As Covid tore across the planet, Mandel became something of a pandemic prophet for writing the hit 2014 novel, Station Eleven. Journalists and everyday Twitter users alike looked to her for guidance as they hunkered down at home, desperate for someone—anyone—to tell them what was happening, and what might come next. Now, Mandel is out with a new book, Exit Party, that documents the interwoven lives of people living in two different, near-future versions of the US: A violent, post-civil-war free-for-all, and a totalitarian state where surveillance runs roughshod over a paranoid populace. She started writing the book in 2022, when polarization and political extremism were already surging—but well before the era of Trump 2.0, widespread ICE raids, and tech-infused surveillance we’re living through today.

Exit Party is a tremendous, riveting, deeply human novel—one of the best I’ve read in years. And despite being strictly fictional, the two futures that Exit Party imagines also feel eerily, uncomfortably close to reality. But as Mandel reminded me when we sat down at WIRED’s New York studio, the book is also infused with plenty of hope, and forces readers to ask themselves important questions about what kind of person they’d be if it all went to hell. She was also, of course, quick to point out that “it doesn’t actually take any particular clairvoyance,” to see where this moment in US history might take us. So no, Mandel can’t help you much if the militias roll through—but Exit Party might offer a few lessons in how to show up for your neighbors. (And throw the best post-collapse party.)

You can read our full (light on spoilers) conversation below, or listen to Uncanny Valley anywhere you get your podcasts. Exit Party is out today, September 15, and I highly recommend you pick up a copy.

This interview has been edited for length and clarity.

KATIE DRUMMOND: Emily, thank you so much for being here. I will get this out of the way at the top and acknowledge that we are both Brooklyn moms. And sometimes when you’re a Brooklyn mom, you become friends with another Brooklyn mom, and we are actually friends outside of work.

EMILY ST. JOHN MANDEL: Because you seek out the best mom you can find at the playground.

We met at the playground in the latter stages of the Covid-19 mask era. I’ve known you ever since. But regardless of that, and I’m saying this very, very genuinely, I read Exit Party—I was lucky enough to get a galley ahead of publication—in a single day. I highly recommend it.

Thank you so much. It’s such a pleasure to be here.

For those who didn’t get an advance copy, could you explain what it’s about?

God, the elevator pitch on this book has been so challenging.

I can understand why.

I suck at elevator pitches generally. But I think of this as a story about crime and resilience. It’s about trying to be a decent person when things are falling apart, and it’s about who you’re going to be at the end of the world. I don’t mean to imply that the collapse of a country is the end of the world, but what I do mean to imply is that it feels like it.

I do see this as a hopeful book. It’s about what comes next and trying to be a new person in a rapidly changing and difficult landscape.

It feels very timely, in troubling ways. When you think about what’s actually happening in the US right now and what’s happening in the US in the book, in what ways are those two places different?

The difference is that in the country that we live in, there’s still time. The United States that opens this story, that country ran out of time. That country fractured. I wish that it felt more fictional. You know, I started writing this book in the spring of 2022.

I was wondering about that.

We could already see the division that was coming because Trump was obviously so much on the horizon for a second term. But I wish it felt more speculative. I’m about to go out and do a 42-city tour where I talk about this absolutely fantastical speculative leap of imagination in which the United States falls apart.

Yeah, imagine if that happened. That would be wild.

Yeah, no one would’ve been able to see that coming. I don’t mean to be glib about it. It’s a terrifying prospect. A big part of the beginning of this book for me was working through my political anxieties.

In 2022?

Yes. I read the news. I’ve always read the news. It doesn’t actually take any particular clairvoyance to look ahead and see where the incredible division in this country could potentially lead. Of course I hope it won’t. But a way of thinking about that for me was, well, what might come next?

Suppose the worst were to happen. Suppose there were to be a second American Civil War. What might the aftermath be like? Which is not dissimilar to the project of Station Eleven. You know, imagining the aftermath of a very different kind of catastrophe and collapse.

This sounds ridiculous given everything I just said, but it’s actually a hopeful book. The book opens at the end of the world, but we’re going to a party. That’s kind of the spirit of it.

I have a memory, it was maybe about a year ago, we were having dinner with your wife. I knew that you had this book coming out. I didn’t know anything about it, and she was very stressed because she was like, “Book publishing is so slow. Look at everything happening in this country right now. This book that Emily has written is so timely right now. I can’t believe it’s not coming out for another year.”

Knowing that you started writing it in 2022 is interesting to me, because you have now watched Trump be reelected. You watched the ICE crackdowns happening at the top of the year. This encroachment of surveillance into different aspects of our lives, the ongoing fracturing of this country, the polarization. How has that felt in these last four years, and particularly once the book was finished and you knew what was in that book?

It’s funny, I remember that dinner. I remember my wife, Laura, her concern that the country might collapse before the book came out.

She’s like, “They’re going to ruin the book.”

Right. That is a spoiler. [Laughs] No. God, we have to laugh because otherwise it’s devastating.

Well, Donald Trump is going to ruin Emily’s book!

Yeah, stakes. Seriously.

No, it’s been a strange experience because, I don’t know if this is too big a spoiler, but the book envisions two different versions of the United States. There’s the vision we begin with where there has been a civil war, the country is fractured into somewhat contentious city states and small new republics and splinter groups. It’s a bit chaotic.

There’s another version of the country that is a totalitarian police state, and my experience of writing this book was that when I started writing it, I was just looking at the polarization, and I was thinking, “Well, obviously we’re in that first universe, the one where the country fractures into these little incompatible small new republics.” But by the time I finished the book, the ICE raids were happening, and my totalitarian vision of a country where citizens are disappeared by masked secret police wasn’t fictional anymore. So there was this feeling of struggling to stay ahead of the news cycle through the entire writing of this book.

Now, the protagonists in the book, Ibari and Ari, they’re the same person. But they also aren’t at all.

Right.

How did you think about the idea of identity in writing the book?

I’ve been fascinated for a really long time by the idea of the counter life. That’s the life you didn’t live, where you didn’t immigrate from Canada, or you went to a different school or took a different job or married a different person. These inflection points that would have made us, over the passage of time, into somewhat different people.

I think the reason I’m so interested in that idea is that the life I’m living feels a little bit improbable to me. I’m from Canada and I’m trained in contemporary dance, and I don’t have a high school diploma, so I do have moments of thinking, “This actually feels fairly implausible.”

I was thinking of identity in terms of the people we could have been. But then in this book, it was also interesting to me to think about, well, what is the counter life of the state? Because that’s a really interesting idea. I was going to say in any democratic country, but honestly in any country.

I was thinking about, well, if we take that idea of the counter life of the individual, these different people we could have been, how is that affected by the counter life of the state? Because we are so influenced by the environment we live in, the people that surround us. I think that I’m a different person having lived for 23 years in New York City than I would’ve been if I had stayed in, say, rural British Columbia where I grew up.

Oh, for sure. I think about that too all the time.

Yeah, you’re from Canada too.

I think a lot about, what if I had just stayed in Calgary?

Yeah. Canada is a little more chill.

It is a little more chill.

See, now here we are, both extremely un-chill people in New York City.

Yes, thriving.

Living our best lives. Extremely stressful lives, but here we are.

Ari and Ibari, they grow up in different versions of a country. They are so profoundly shaped by that country. I saw Ari as somebody who’s got this character flaw where she’s fundamentally drawn toward crime. Maybe that’s not even a flaw. Maybe it’s just an interest, because there’s a version of the world where she does the obvious thing and becomes a criminal, but there is another version of the world where she channels that into becoming the police. It was interesting to me to think about that, the way we might be drawn in one direction or another depending on the environment we’re raised in.

I almost hate to ask this, because both versions of the US that you portray in the book are really frightening in their own respects. But did you find yourself drawn towards preferring one over another? Is there one version of this that you would rather live in?

No, they’re both really hard. They’re both awful. The version I want to live in is Canada, candidly. It’s like, do you want to live in chaos, or do you wanna live in a totalitarian police state? Hard pass on both.

That’s an idea I’ve been thinking about a lot: no version of a country lasts forever, and there’s both terror and hope in that idea. The totalitarian police state version’s not going to last forever, and neither is the period of total chaos. What I wanted to suggest, and this is kind of a deep cut for people who have read a lot of my work, but in Sea of Tranquility there’s a futuristic section where an author’s on a book tour in the year 2300, and she’s traveling between the Republic of California, the Atlantic Republic, and the Kingdom of Deseret, and it’s chill enough to support a book tour, you know?

But yeah, I kind of wanted to suggest that this is the founding story of that space.

I wanted to ask you about your multiverse. You have a history that your diehards, your fans, really glom onto, which is you pull from previous works and you bring people and worlds into your new work, and that continues with Exit Party.

You’ve talked about it as a multiverse, a cinematic universe. The New Yorker called it, quote, “The Mandel Cinematic Universe,” which I think is pretty incredible. Is there a master plan here, or is it just ad hoc as inspiration strikes?

I wish I could say I had the kind of brain that could hold an entire multiverse. I should just say yes, but, uh, no. There is no plan. That’s the baseline for individual books when I start them, but also for The Mandel-verse, as some people have also called it. I don’t have any kind of master plan that I’m working towards. I just like the opportunities for connection. I guess I have some abstract longing for order in a chaotic universe, you know? I like finding the moments when books can be drawn together, or sometimes I just fall in love with a particular character, and I want to spend more time with that person.

Like Miranda in Station Eleven, I really liked her. It was really interesting to get to see a completely different facet of her life in The Glass Hotel, to see her as a shipping executive as opposed to an artist. Then I guess with Exit Party, I did lay the groundwork in Sea of Tranquility for a world where the United States no longer exists, so I guess you could make the argument that that implies the origin story.

One of the things I love about your books, several of them at least, is that they’re set in the future, but they don’t foreground the technology. They foreground the people and the human stories and the lived experience. Technology is often present, and it’s certainly present in Exit Party, but how do you think about introducing technology and science into your work? How much research goes into that, and how much commitment to factual accuracy as opposed to just letting your imagination run?

What I’m aiming for is plausibility in the context of a literary novel. Which is going to be a disappointingly low bar for your average WIRED podcast interviewee.

But I think that you’ve touched upon a really interesting choice that writers of speculative fiction need to make every time we embark on a new project, which is really just how deep into the weeds you want to go on your science fictional technology.

There are writers who go pretty deep. I loved Cixin Liu’s work, The Three-Body Problem, and the rest of that series. He will go so deep into these science fictional technologies, which to my totally untrained eye, feel totally plausible.

I did flirt with that a little bit when I was starting out on Sea of Tranquility, which has a time-traveling detective, because we were all deranged during the pandemic, but yeah, I was thinking about “Well, how would this actually work?”

I was reading about the quantum blockchain and trying to understand what the mechanics could be. But then at a certain point I realized, you know what? It’s just transport. Like, if I’m not explaining how your car works in the 2020 segment of the novel, then why do I care about the mechanics of the time machine?

There is a part in your book from Ibari’s perspective. She’s working as a detective in the secret police in one of these United States. She has this thought, and I’ll quote it, “I’ve spent a long time thinking about the nature of life in a surveillance state, which is perhaps obvious given my work, but there are observations I’ve made that I share with no one, and one of them has to do with the way in a place like this, reality splits into two.” That line, “reality splits into two,” is great. What does it mean to you?

In the context of that passage, she’s a secret agent with a cover story, basically. But the cover story is that she’s a professor, and she actually is teaching real classes. So it’s not false that she’s a professor, it’s just that she’s also secret police.

She has the most obvious kind of surface experience of that idea of living a double life. I don’t want to say it’s exactly modeled on East Germany, but the Stasi apparatus, where civilians were constantly reporting on each other, which everybody thought was overstated until the country collapsed, and it turned out to have been even bigger than anyone knew. Everybody really was reporting on each other. That does create this awful kind of doubleness where there’s the part of you that’s a friend, but then there’s a part of you that’s reporting on the friend to your handler, and both of those ideas are true.

But I think that also living that kind of doubleness would do a lot of damage to a person. It sounds so literary and pretentious to say damage to your soul, but I think that might be the situation.

In that surveillance state, there are some very stressful passages, at least for me, where you realize that none of these people can really fundamentally trust each other. And then in the other universe, there is this baseline notion of trust eroding because everybody is just looking out for themselves in the chaos. How were you thinking about that as you were writing the book? Did you have any revelations about trust in this country that we could all learn from?

It’s not specific to this country, but this actually leads us into AI for me. Do you remember when we were kids and a video was proof? It was like, “Oh, there is video evidence. Case closed.” How terrifying and damaging and ridiculous that our children are being raised in a world where a video means absolutely nothing. It’s incredibly hard to authenticate, and that’s what I find myself thinking about when I think about trust.

Sometimes I lurk on Threads, just recreationally. It is such a trash fire. It’s like Twitter in, I don’t know, 2012 or something. But something real that’s really struck me is the degree of betrayal that people feel when it emerges that an author has been using AI, and I share that sense of betrayal. It offends me in some incredibly deep way, and when I try to parse why that is, I think trust might be the answer.

I was going to ask you about AI. You write novels. You also work in Hollywood. And so you have exposure to what’s happening in the book world around AI, and also what is happening in the entertainment industry. Do you use it at all?

I don’t. Maybe this makes me an outlier at this point. I haven’t found it reliable in search, and so that made me distrust it completely.

Every now and again, I like to search and make sure that my daughter is not on the internet, basically. So I searched “Emily St. John Mandel daughter” I think about a year ago, and AI is so confidently wrong. It confidently introduced me to my son Leo, born in 2016, and said, “Emily St. John Mandel does not have a daughter.” And it’s so confident that you have a moment of just like, “Do I have a son?”

I do not have a son named Leo. I have no idea where that came from. I searched again a month later, and it had revised that, and my daughter remained unnamed, which is what I wanted to confirm.

But it made me distrust it fundamentally from a search perspective. I know that it can be used to streamline administrative tasks. I haven’t used it as such. I can write my own emails.

I was going to say, you’re still writing your own emails.

Yeah, exactly.

So am I.

I read your interview with Matthew Belloni talking about the use of AI. It was interesting to me because without refuting anything he’s saying, I’ve just had a different experience, and I haven’t heard of people using it in my writers’ rooms in Hollywood.

I suppose AI is an argument for the in-person writers’ room. You really can’t use it if you’re sitting around a table and there’s a chalkboard. Probably people are using it without my knowledge, but I haven’t personally encountered it.

When you look at the effect it’s having on book publishing—we have covered some of this—but it is chaos out there.

How does that look from where you sit? Do you worry about AI slop books becoming a mainstream commodity, taking up a bigger share of the fiction market? Or do you think people just don’t have the taste for that?

I’d like to think they don’t have the taste for it, but I think the slop will get less sloppish with time. What I really worry about, actually, is less the work, and I worry more about writers. Because for me, I started publishing in 2009, and I have a consistent voice across books.

My process is such that it would be very easy for me to produce iterative drafts of what I’ve been doing if I were ever forced to prove that it’s not AI, which it absolutely is not. But not all writers have that process. Some are just, you know, changing a Word doc over a period of time. Not all of us keep written notes. Something that’s emerging that alarms me is that it’s very hard to prove sometimes that your work wasn’t AI.

Oh, like to prove a negative.

Yeah, to prove a negative. The reputational risk is enormous. That’s the end of a career, or at the very least, the end of a book deal.

I had a weird experience a while ago where I was just curious, so I found some free AI detection thing online, which probably stole my work, and I pasted the first chapter of Exit Party in, and what it came back with was there’s a 20 percent chance that AI was used on this. At first I was completely beside myself. I was like, “How does it think that?” And then I realized, oh, because Anthropic stole five of my novels to train their LLMs.

But to your point, I mean, the idea of being a writer for a living right now, and you’re doing the work yourself, you’re doing it with integrity, you are painstaking in your process, and one accusation or one result … And these tools are extremely flawed …

Yeah. They’re very inaccurate.

… But that can end a career. The stakes of this are very, very high. I actually did the same thing. I think anyone who writes for a living right now has this paranoia. I fed some of my articles into one of these tools because I was like, “I just want to know what it’s going to say.” And I got back zero percent.

Congratulations.

Thank you, but it was such a relief because there is this sense, this cloud of someone can point a finger. And the people who are using AI, I would argue, in creative pursuits, are ruining it for everybody else.

I would agree with that, and I don’t have any respect for them. It’s like if you can’t write, then maybe do something different. There are a lot of better jobs.

I wanted to spend a bit of time talking about you and your career. You grew up in Canada, you were in Toronto. Objectively speaking, where you are now is improbable relative to where you were then. Tell us about those early years.

I’m going to digress for a second back to the book. So much of this book is about the condition of living one life after another. And this is so many lives ago that it’s like looking back at a stranger.

I grew up in rural British Columbia, Vancouver Island, and then Denman Island, which is, just for context, about the same size and shape as Manhattan, but with 1,000 people. It’s forest and deer and a general store. I was homeschooled. I did a couple semesters of community college without ever getting my GED. I auditioned successfully for a conservatory program in contemporary dance in Toronto, and I was a professional dancer after that for, like, five minutes.

I went pretty far, but I never really joined a company. I danced for a couple of independent choreographers. I loved that experience, but I never got into one of the big companies, I think in part because I’d realized by then that I no longer loved it. There’s this phenomenon, I think, where if you’ve been dead set on one thing your entire life, like since you were six, then you can fall out of love with it. People change.

In my early 20s, I realized that dance was feeling like more of a chore than anything else, and I kinda resented it, and I didn’t love it anymore. But this is kind of spectacularly poor planning on my part. I’d taken out student loans to go to the School of Toronto Dance Theatre, not a degree-granting program. So I had a mountain of student loan debt, no degree, no high school diploma. It would feel overwrought if you wrote that in a novel. My editor would probably do, like, question marks in the margin.

Right. Many poor choices.

I had a totally destroyed credit rating in Canada due to student loan debt. So I moved to the United States. That was the solution to that problem. I’d always written just as a hobby. It was just something I loved and something I did a little bit compulsively. Like, if I went for a long walk, I’d bring a notebook because otherwise I might start writing stories on Starbucks napkins.

I was starting to write in Toronto. I met a guy who lived in New York and we were dating, and I moved to New York to be with him. Moved back to Montreal, broke up, and then I was kinda stranded in Montreal working. In winter, working retail, trying to figure out what to do with the rest of my life.

And it’s a cold city.

It is a very cold city, which actually played into my writing process. My roommates and I were dead broke, so we kept the heat turned down pretty low. I would go to bed at 7:00 pm just because I was so cold. It was easier to be asleep. I would sleep until about 1:00 in the morning and then naturally wake up. I’d get up in the middle of the night. I would dress to write. I had this feeling that I was doing something formal and important.

I would put on, like, a sweater and a tie and my warmest coat and go out with my laptop to the 24-hour cafe two blocks away, and I would write until morning, and that was the book that ultimately became Last Night in Montreal.

I was able to find a way to move back to New York, with, I don’t know, $400 and a 10-day sublet in the West Village when I was 22 or so. I’ve been here ever since. I started writing Last Night in Montreal in Montreal. I finished it in New York, and then I just started cold querying agents, and the 13th or 14th one found me in her slush pile and took me on.

There was a lot of luck and also an incredible amount of hard work and literally cold nights in Montreal.

And you had a day job?

I did. I’m from a very working-class background. And what that came with for me was an absolute terror of quitting the day job, because there is no safety net. I was working retail in Toronto and Montreal. I became an assistant in New York. I had a really great job for a long time at The Rockefeller University. I was an admin in a cancer research lab, and I loved that. It was such a pleasure to be surrounded by such smart people doing interesting work, and the health insurance was incredible.

I kept that job until a year after Station Eleven came out.

That’s so wild.

It got so surreal, Katie, I can’t tell you. One of my jobs in that position, or one of my tasks, rather, was to book plane tickets for my boss. But I wasn’t booking my own plane tickets for the tour. I had three publicists doing that. Or, like, there was a day when I had to leave work early because I had a photo shoot at Time magazine.

Did someone tell you, “Emily, it’s time. You need to quit your job?”

You know, you know what it was? I did a podcast where the other guest was this comedian, Josh Gondelman. And at the time, I guess it was a few months after Station Eleven had come out, and I was really grappling with the day job question. Like, does it make sense to leave? Is leaving crazy? What if I never have another successful book? And he told me that a fellow comedian had once told him, “Keep your day job until you can’t afford to anymore.” I found that to be a really helpful way of framing it. I kept that job until I found out I was three months pregnant, and then that was too many things.

Station Eleven was successful right out of the gate, and then it had this second life during the pandemic. You became this pandemic prophet for people who were asking you on Twitter, like, “Help. What do we do in the pandemic?” Looking back on it now, what stands out to you from that experience? What was it like to write a fictional book and then be asked for guidance in a real-world pandemic?

It was so surreal, and the requests were so misplaced. Like, if you’re going to look to a writer for guidance, I want to be like, “Go talk to Ed Yong of The Atlantic.” Or a scientist or something. There were so many people who were more qualified than I was to say what to do in a pandemic. It was very strange being … accused sounds too dramatic, but being accused of being a prophet in a weird way.

What a dark thing.

I was never able to quite convey at the time that there was always going to be another pandemic. I didn’t see myself as having foretold anything. The way I thought of it at the time was, well, if I’d written a novel with a fictional war, does that mean that I predicted whatever war came next after that in the real world?

What’s funny is I was exactly as unprepared as everybody else. I’ve had this sideline for years of doing on-stage conversations and lectures, and I was doing that right up until March 2020. I had this long lecture that I developed about post-apocalyptic literature, about pandemics and Shakespeare and all themes that dovetailed with Station Eleven.

I was literally traveling the country delivering a lecture that contained the phrase, “There will always be another pandemic.” It is so embarrassing how shocked I was that there was, in fact, another pandemic. I was scrambling for toilet paper like everybody else.

Yeah, you’re like, “Guys, I don’t know if you should put Clorox all over your groceries.”

Yeah. And I did.

I definitely did that, for probably longer than I should have.

Same.

I’m curious about how you deal with celebrity. Just thinking about that social media onslaught during Covid. You don’t strike me as someone who’s particularly eager to post, you know what I mean?

You’ve discovered my secret. I’m not eager to post.

How do you deal with that part of the job? And do you feel pressure to do it differently than you’re inclined to do?

If I were on Instagram just to be on Instagram, I don’t know if you’d ever see my face, you know? Because I love posting weird, beautiful little videos. Like, I just posted one that’s spiderwebs on a hiking trail in the Adirondacks last week.

If I see something cool and beautiful and I think, “Oh, that’s kind of amazing,” in just a tiny little moment, and could I set it to music? That would be cool. I really like Reels. That’s how I probably naturally would use it. I do feel a certain pressure to post, I guess, to post my face for the algorithms. I do feel like I need to promote my books because it is literally how I pay my rent.

It just kind of feels like part of the job. I don’t want to get too negative here because I do genuinely love interacting with readers. So I guess I kind of want it both ways.

There is pressure there. I find social media generally terrifying. I’ve had a lot of very negative interactions over the years, and what’s funny is just the way the human brain is wired, the positive interactions outweigh the negative interactions 1,000-to-one. But I also remember the negative ones. The negative ones stand out.

What is it like to promote a book in the year 2026? How has it changed?

I think that I have a very atypical experience of book promotion, because I had these three novels published by a really small press, and then I jumped to Knopf in the US, HarperCollins Canada, and Pan Macmillan in the UK for Station Eleven, and I’ve been with those three publishers ever since.

Because Station Eleven was so successful, a lot of marketing resources have been poured into the subsequent books. I think that that gives me a really unusual experience, and I think that it’s not normal at all. So it makes me hesitant to draw any generalizations, because for me it’s fine, you know?

But that’s certainly not the case for the vast majority of novelists. I think there’s so much distraction and chatter and such short attention spans that I think it’s incredibly difficult to start out as a novelist now. It was never actually easy, but my sense is that it might be harder now in this more fractured and more chaotic media landscape. I think it becomes harder for a novelist to break through.

Do you have any advice for those people?

Stay off Threads, for real. I don’t really post, but I do read. I see the most deranged publishing advice on Threads.

Like what?

I don’t even know. Every time I read it, I’m like, “No, that makes no sense. Don’t do that.”

I want to end by coming back to Exit Party. [As you said], you were accused, for lack of a better word, of being a prophet when Station Eleven came out before the Covid-19 pandemic. You’ve now written a book that is close to what we are going through in the US. Are you bracing yourself for a similar reaction?

Yeah. But look, if the country collapses, it’s not my fault.

She said it here first.

I didn’t do it. I just read the news and got freaked out like everybody else.

Is there anything in the worlds you’ve created in the book that you hope turns out to be prophetic?

Absolutely. I don’t want to take credit for the parties after … but you know what? It’s hope itself. There is a lot of hope in this book. It opens at a moment when the country has collapsed, but something new is taking its place, and it’s a kind of chaotic in-between period, but I never meant to imply that would last forever.

I think that settles down into something more stable and potentially better. And then on the side of the book that deals with a totalitarian state, that totalitarianism is already crumbling toward the end. I think that we see that even now in our current incredibly dark political moment.

I mean, maybe what you’ll get credit for predicting is the dawn of an even better version of this country.

That would be lovely.

How to Listen

You can always listen to this week’s podcast through the audio player on this page, but if you want to subscribe for free to get every episode, here’s how:

If you’re on an iPhone or iPad, open the app called Podcasts, or just tap this link. You can also download an app like Overcast or Pocket Casts and search for “uncanny valley.” We’re on Spotify too.


Source: Culture Latest

Why No One Wants to Date Tech Bros

Engaged in a flirtatious conversation at a bar in Manhattan, 23-year-old Gary stood with a Guinness in hand as he explained to a woman why he took an engineering job at Palantir, the data analytics company cofounded by Alex Karp and Peter Thiel. Raking in a $250,000 salary would allow him to help pay off his parents’ mounting medical bills, he explained.

Gary wasn’t met with sympathy. Medical debt, the woman told him, was “not an excuse to work somewhere like that.” She was referring, of course, to Palantir’s controversial work with governments, including the US and Israel, which use the company’s surveillance tools for law enforcement and military operations.

Since joining the tech company in 2025, Gary has noticed that naming his employer prompts brutal reactions from prospective paramours. One woman, he says, called him the devil; another asked how he lives with himself. “Sometimes people would just straight up stop talking to me,” he says.

Recently, Gary’s buddies requested he not mention Palantir on their nights out together so as to not tank the whole group’s chances of getting laid. “I get nervous now. I don’t bring it up,” he says. (Gary, along with the other daters in this piece, have been given pseudonyms to protect against professional, if not romantic, repercussions.)

Gary has resorted to saying he works in AI. “Even that gets some pretty bad responses sometimes,” he says. His Hinge profile doesn’t mention his employer. And sometimes he makes up stories out of thin air to justify working for Palantir—like that line about the medical debt. His parents don’t have any, nor are they ill.

The bachelors of Palantir aren’t the only tech bros dealing with the industry’s fall from grace. Take 29-year-old James, who lives in San Francisco and works at Tesla. Earlier this year, on his third date with a woman he met on Raya, she asked whether he’d voted for Donald Trump in the 2024 election. A self-described “moderate liberal or moderate progressive” who “always votes blue and always will, probably,” James was offended. “I was like, ‘Are you kidding me? Did I really give that off?’” She replied, “You work for Tesla, so I just need to check.”

James says skepticism of his politics has “come up repeatedly” in his dating life. “If I don’t immediately say that I don’t like Elon Musk, people think that I’m conservative, which is generally not something that people like in San Francisco,” he says.

Image may contain Bag Backpack and Purple

Once seen as a groundbreaking leader in sustainability and a quirky futurist, Musk—recently crowned the world’s first trillionaire—is better known these days for his record-breaking donation to Trump’s 2024 campaign, being the brain behind DOGE, the dismantling of Twitter, an apparent Nazi salute, frequent ketamine use, racist tweets, and something of a breeding kink.

Long gone is the cliché of the optimistic tech guy in flip-flops and a ratty old hoodie. The tech industry is now seen as the locus of unchecked power. Meta created an environment where Russian bots could influence elections and designed features that foster social media addiction among teens. Amazon, which has been widely accused of unfair labor practices and of generating a staggering amount of waste, has aided in the collapse of small, independent businesses, while a glowed-up Jeff Bezos has become a poster child for extreme wealth inequality. More recently, the AI industry is being widely maligned over claims that it’s exacerbating job loss, pollution, and a critical-thinking crisis.

Amy Laurent, the founder of a high-end matchmaking company, has a front-row seat to the way Big Tech’s moral failings are causing problems for single, rank-and-file tech bros. Turns out their cushy paychecks often aren’t enough of an upside anymore.

“I think 10 years ago, saying you worked in tech was more about innovation, ambition, intelligence,” says Laurent. Now, many people charge Big Tech with “causing the breakdown of society,” whether they’re worried about “AI, job displacement, wealth inequality, privacy and surveillance, content moderation, or political influence.”

Laurent says that when she pitches her tech-employed male clients to women, she acknowledges their hesitation right off the bat: “Listen, I have a great guy to tell you about. He’s such a good match for you. Now hear me out: He’s in tech.” The justifications are a new feature of her work.

Jessica Engle, a family and marriage therapist, has noticed that men in tech worry about being perceived as red-pilled. Recently, one of her clients expressed concern about having his Apple Watch visible in his dating profile photos. He wondered whether the gadget “would make him look like a tech bro” and whether it might signal a “very particular kind of toxic masculinity that we’re seeing more and more in the news.”

Engle’s client wasn’t wrong to worry. Mira, a 29-year-old Bay Area resident, says she perceives a career in tech as a possible “right-wing flag” or a sign that someone might be “manosphere-coded” or libertarian.

When 28-year-old Helen was an undergraduate at Stanford from 2016 to 2020, she viewed men in tech as guys who “stay up late and maybe have bad hygiene.” But her feelings toward them have gotten even worse over time as she’s questioned the industry’s overall impact.

Helen’s family is from the Tigray region of Ethiopia, whose population, beginning in 2020, was subjected to a prolonged period of human rights abuses. “When I started thinking about questions around genocide and the technology that kills or tracks or excludes,” Helen says, “I realized that the things these people are building are secretly, or under the surface, really nefarious, and I could not personally be with someone who is a puzzle piece in that tapestry.”

Helen herself has worked in tech. Soon after graduating from college, she accepted a job at Grammarly. But after about a year, she went to work for a mutual-aid platform and now is employed at an impact investing firm funding technology for underserved communities. Similarly, Sasha, a 31-year-old Oakland resident, had a brief stint working in AI before heading to a sustainability-focused tech company, where she is now chief of staff. In dating, when she meets someone who works or wants to work in AI, she wonders, “Does that mean that we have the same values in terms of climate and sustainability and thinking about human rights?”

Even if they do, she adds, “then the issue I have is there’s clearly this cognitive dissonance happening between what you’re working on and what you care about, which I have a really hard time accepting.”

Image may contain Bottle Cosmetics and Perfume

James, the Tesla employee, confesses that he embodies plenty of tech bro stereotypes that might work against him, including “paying too much in rent” (or “just generally gentrifying”), ordering most of his meals on DoorDash (“I suck at cooking”), and not “really contributing to the culture of the city that much.”

“I really fit the mold so much that it’s sort of like, I have to at least be self-aware of it,” James says.

The pushback James gets IRL makes him wonder how much more happens behind screens. “There’s less visible rejection because you kind of get prefiltered out,” he says. “If I were to write on the dating app that I work at Tesla, I think it would just cut a lot of people right away.” His experience seems to bear this out. On his Hinge profile, he disclosed his place of work; on Raya he decided against it. He says he’s had more success on Raya.

In fact, about four months after joining Raya, he met someone. His new girlfriend had recently moved to the Bay Area for medical school and, before dating James, had managed to stay insulated from tech culture. “I was talking to my now girlfriend, and she was like, ‘Yeah, I feel like I’ve gotten lucky. I didn’t really know what the tech bro stereotype is.’”

“I was like, ‘Thank God. I don’t think I would have made it if you did.’”


Let us know what you think about this article. Submit a letter to the editor at [email protected].


Source: Culture Latest

SeaWorld Wants to Make You Horny

When you think of SeaWorld, you most likely think of orcas, school trips to the aquarium, or eating chicken fingers while contemplating the ethics of marine mammal captivity. But this Halloween, the theme park franchise wants to make SeaWorld synonymous with something else entirely: BookTok smut.

On September 11, SeaWorld Orlando officially opened its annual Howl-O-Scream event, an after-hours Halloween-themed immersive experience featuring various “scare zones” in the park. Scattered between the more expected spooky fare, such as a haunted “Woodrot Hollow” forest and an I Know What You Did Last Summer–themed haunted house is Just One More Chapter, a BookTok-inspired zone featuring the hunky tortured bad boys of romantasy lore.

In videos posted by Orlando-based influencers who got an early look at the park, young men with masks, floppy hair, and painted-on abs gallivant around wearing black denim, loitering next to brick walls, beckoning to park attendees, and gifting them with roses. And on Instagram, influencers posed around a central set piece, an oversized book open to a chapter called “The Cathedral,” emblazoned with the slogan “Good Girls Go for the Bad Boys.”

Some of the TikToks from the scare zone went viral, garnering mixed reactions. “Booktok dark ‘romantasy’ cancer is RUINING heterosexuality,” wrote one X user, “every little tasteful drop of eros has to be squeezed out of everything so it can be refined into pure concentrated fucking CRINGE.” The fact that the event took place at SeaWorld, an amusement park franchise that is perhaps most closely associated with a harrowing documentary about an orca, also lent to the absurdity.

But I did speak to one content creator present at the scare zone who had an entirely different reaction. “I wouldn’t say it was scary,” said the creator, who declined to provide their name because they did not want to “jeopardize their relationship with SeaWorld.” “But it did leave me with a good feeling—I guess you could say a sexy feeling, in a way.”

According to Jessie Kinsey, the art director at SeaWorld Orlando, that was precisely the point. She came up with the idea for the event last year, after seeing romantasy bestsellers like Lights Out and Haunting Adeline go viral on TikTok. She knew that a dark romance-themed scare zone was not necessarily “something that you would expect from a traditional horror event,” but she had a feeling it would resonate with the BookTok girlies. “I remember texting my production manager, saying, ‘Don’t judge me, but I have an idea,’” she tells WIRED.

A fiction writer in her spare time (she actually wrote the text for the oversized book set piece at SeaWorld), Kinsey came up with the narrative for the zone: It would be set in a crumbling Gothic cathedral, built in a town run by a wealthy family called the Blackthorns. “These roses grew up around it, and created this beauty out of destruction,” Kinsey puts it. “Which I thought was kind of fitting for the themes of these books.”

Romantasy, or romance and fantasy, fuses mythology and supernatural figures such as faeries or demons with explicit sex scenes, exemplified by books like Rebecca Yarros’ Onyx Storm and Sarah J. Maas’ A Court of Thorns and Roses series. The genre has exploded in popularity thanks to BookTok, with Bloomberg reporting that romantasy pulled in an estimated $610 million in 2024 alone. The male antiheroes of the genre, such as the “shadow-daddy” Rhysand in Maas’ A Court of Mist and Fury and dragon rider Xaden Riorson from Yarros’ Fourth Wing, are tortured, dangerous, (at best) morally complex, and egregiously hot.

The MMC—“male main character,” in fandom terms—is an amalgam of different romantasy tropes, a mysterious, vaguely threatening masked man represented by the eight scare actors employed in the zone. (In one video, one of them inexplicably wears KneeBlades and skids toward a prostrate guest and jumps over them.) For both narrative and legal reasons, they’re not based on any one specific character.

The infusion of the romance genre with themes of psychological manipulation, violence, or nonconsensual sex has led to criticism of the dark romantasy genre in general, and of the Just One More Chapter scare zone. But “the threatening nature” of the romantasy bad boys is why it works in a horror context, says Kinsey: “It’s a love story, but it’s not all sunshine. This is someone who is morally gray, who is potentially violent, who is obsessive, and that part of it is threatening. But that also plays into the appeal of it.”

On social media, some pointed out the potential for fans sexually harassing scare actors, a not-uncommon issue at Halloween theme park events. “Being a scare actor, it does come with risks regardless of what area of the park you’re working in, because not everyone reacts the same way to being startled or surprised when alcohol is involved,” Kinsey says. “But our actors’ safety is first and foremost.” She says the actors receive extensive training for how to deal with harassment, and that security and stage managers are always present to intervene if necessary.

So far, however, SeaWorld Orlando says that horny millennial ACOTAR fans haven’t posed too much of a problem—in fact, quite the opposite. Kinsey says the viral attention has translated into an explosion in ticket sales, with Howl-O-Scream enjoying its biggest month in the history of its six-year run. “I knew that it was going to get people talking, and I know that the genre isn’t for everyone, so it was going to be a bit controversial,” says Kinsey. “I did not expect the size of the reaction, but it’s kind of fun.” It is, dare we call it, a Howl-O-Scream.


Source: Culture Latest

How much oil-market buffer is left?

Sep 14 close · +40% vs pre-crisis ~$76

Sep 15 · new all-time record ($6.2694) — sixth straight · +68% vs pre-war $3.72

Sep 15 · +18¢ in a week (AAA) · +54% vs Jan $2.81

Sep 4 · down 1.2M in a week · down 130.1M from pre-war 415.4M · lowest since Dec 1982

Sep 4 · up 2.1M in a week · 13% below 5-year average · East Coast stocks 28% below last year

These charts show how oil, gasoline, diesel, and natural gas prices have changed. Brent is a widely used benchmark for the price of crude oil. The gasoline and diesel charts use US national averages; the natural gas charts cover Europe and Asia. Every plotted value comes from a source. Missing readings are left out.

The national average has risen from $2.81 in January to $4.33, an increase of 54%. Prices peaked at $4.50 in May. Every month since March has averaged above the levels seen before the crisis.

Diesel cost $3.72 in the last week before the war, on Feb 27. By Sep 15, it had reached $6.27, an increase of 68% and its sixth consecutive record. It passed the previous record of about $5.85, set in June 2022, on Sep 4. This chart shows selected readings. You can see the daily Sep 3–15 readings in the price spread chart below. Diesel prices have risen faster than gasoline prices as supplies of refined fuel have tightened.

TTF, $/MMBtu · filled: EIA weekly futures avgs (Jan–Apr) · hollow: individual contract quotes (May–Sep) · different instruments — a trend, not one series

TTF is the benchmark used to track natural gas prices in Europe. It traded above $28 on Sep 10, eased to $27.00 on Sep 11, and stood at about $27.80 on Sep 14 — the highest levels since December 2022. That's roughly 153% above the price before the strait closed. The Sep 14 value is converted from €81.98 per megawatt-hour to the dollar units used in this chart. Prices have risen as unplanned maintenance at Norway's Asgard and Troll fields and tanker attacks put pressure on supplies.

By January 20, European gas storage had fallen to 48%, compared with a five-year average of 63%. Running that low left Europe buying liquefied natural gas (LNG) for immediate delivery during the season the strait closed.

JKM, $/MMBtu · filled: EIA weekly futures avgs (Jan–Apr) · hollow: assessed spot (May–Sep) · different instruments — a trend, not one series

JKM tracks the price of liquefied natural gas (LNG) delivered to Asia. It reached the high-$28s on Sep 10 — its highest level in roughly two and a half years — and held around $28.50 on Sep 11, about 167% above the price before the closure (JOGMEC). Damaged production units at Qatar's Ras Laffan complex — about 17% of the country's LNG export capacity — are expected to be offline for 3–5 years, forcing Asian buyers to look elsewhere for supplies. With little gas in storage, the region is particularly sensitive to changes in supply and weather.

Brent: sourced points (EIA monthly avgs: Mar $103.0 · Apr $117.29 · Jul $83.76) · WTI: weekly (FRED)

Brent rose from $76 before the closure to an intraday peak of $126 in March, then fell to $95 by mid-April and $74 on Jun 30 as hopes for a ceasefire grew. It passed $100 again on Sep 9 as tanker attacks escalated, rose $6.82 to $108.03 on Sep 10, and settled at $104.61 on Sep 11.

Brent closed at $105.68 on Sep 14 after reaching $106.73 earlier in the day. Prices rose after talks between Gulf foreign ministers and Iran were postponed. Reuters also reported that stocks at Yanbu, Saudi Arabia's Red Sea port, could support another five to seven days of exports if the East–West pipeline remains closed.

The gray line shows West Texas Intermediate (WTI), a US crude oil benchmark. Most WTI readings come from FRED's weekly spot series; the Sep 10, 11, and 14 readings are futures closes. WTI is also the crude price used in the fuel price spread chart beside this one. Hollow markers indicate a source's rounded estimate.

Retail price less WTI, $/bbl · weekly (EIA/AAA − FRED).

This chart subtracts the price of crude oil from the retail price of fuel, with both expressed in dollars per barrel. The difference covers refining, transportation, and retailing. It helps explain why prices at the pump can keep rising even when crude gets cheaper. Diesel's spread grew from about $89 in January to $161.9 on Sep 15. It briefly narrowed to $148 on Sep 10, when crude prices rose faster than pump prices. Gasoline's spread has also widened, from about $59 to $78–84. This uses retail prices from EIA and AAA, minus WTI from FRED, with futures closes for Sep 10, 11, and 14. The EIA's official crack spread uses wholesale fuel prices, so the values differ. Eight fuel readings have no same-day WTI value; those use the nearest trading day's price, within 1–2 days. No missing prices are estimated.

When the world uses more oil than it produces, the difference comes out of storage. These charts show how much oil has been withdrawn, how much is left, and how long the US emergency reserve could last under each of the model's three scenarios.

Before the war, the world produced about 4 million more barrels of oil per day than it used. Since the war began, it has had to draw on stored oil every month. The IEA's latest report estimates a full-year supply loss of 5.7 million barrels per day, about 6% of the world's oil, and expects Middle East oil flows to remain below normal until 2027. Global production fell to 100.1 million barrels per day in August, with more than 10 million barrels per day of Gulf production still shut down. Saudi production alone fell by 2.3 million barrels per day that month, to 5.97 million.

World oil balance, million b/d — production minus consumption, EIA STEO Table 3a, reported months (Jan–Aug actuals; the EIA's forecast tail is not shown) · the physical loss peaked at 11.2M b/d of Gulf shut-in in May — demand destruction and non-Gulf supply absorbed most of it · the IEA's observed-inventories count: 507 mb drawn since February — 2.8 mb/d on average, 95 mb of it in August alone.

Withdrawals and changes in demand, using figures available as of Sep 11.

year-to-date · EIA est. (Sep 9)

US Strategic Petroleum Reserve

pulled from 32 countries · IEA

withdrawals inferred from customs data · official SPR untouched

full-year 2026, cut from −1.6 in the August edition · IEA OMR, Sep 11

Q3 2026 forecast — supply below demand · IEA OMR, Aug 12

EIA weekly ending stocks, million bbl

The Strategic Petroleum Reserve (SPR) is the US government's emergency supply of crude oil. It was created after the energy shortages of the 1970s. The reserve held 415.4 million barrels when the war began. The latest report puts it at 285.4 million as of Sep 4. Oil is being released to help make up for supplies that can't leave the Gulf. You can use the withdrawal rate to see how quickly the emergency supply is being used.

The red lines mark the model's reserve thresholds, or floors. At about 300M barrels, some caverns risk damage and can't safely be refilled after a withdrawal. The first report below that threshold was for the week ending Aug 7, at 298.7 million barrels.

The other floors are 250M, the GEF minimum for sustained withdrawals; 180M, the hard operating limit; and 70M, the Department of Energy's stated safe minimum. The model stops withdrawals at 70M.

The dashed lines show what happens if withdrawals continue at 0.45M, 0.70M, or 1.35M barrels per day from the Sep 4 level. The estimates extend to about March 2027.

Withdrawals slowed to about 0.2 million barrels a day in the week ending Sep 4, while diesel stocks rose by 2.1 million barrels, according to the EIA's Sep 10 report. The next report is due Sep 16 and covers the week ending Sep 11.

Estimated odds, updated when specified events occur.

The Saudi bypass pipeline was suspended, and the Houthis held the entire Red Sea coast. An official pipeline restart would return the odds to 10/50/40.

Tanker losses reached 10 per week, Brent passed $100, and Jazan was affected.

A shipping exclusion zone was imposed, and a base in a third country was hit for the first time.

Reported Hormuz traffic of 8.6M barrels per day was not backed by vessel tracking, which showed 7% of normal transits.

Initial model: about 65% odds of de-escalation.

Tankers can pass through Hormuz under an Iran–Oman agreement or with US escorts. Traffic gradually returns to normal over one to two quarters.

In this scenario, Brent moves toward $70–80 and reserve withdrawals slow to about 0.45M barrels per day. Stored oil lasts longer.

The war continues at its current intensity. Tanker attacks and shipping restrictions persist, some Iranian infrastructure remains offline, and the damaged Saudi bypass has no restart date. The strait remains partly open.

Brent stays in the $95–125 range, and reserve withdrawals run at about 0.70M barrels per day. Global stocks keep falling, with shortages developing later.

The disruption becomes a sustained closure or the fighting escalates. Tanker losses rise, shipping restrictions remain, and the bypass, Abqaiq, and Jazan stay offline for months.

Brent rises above $130, and reserve withdrawals reach 1.35M barrels per day. Shortages spread from the US East Coast to Russia, Europe, China, and aviation fuel.

These odds are based on judgment. They change when specified events occur, such as a pipeline restarting or a shipping agreement breaking down. A quiet week alone doesn't change them. You can read the rules and the events being watched on the model page.

This chart puts the current reserve in perspective. It held 727 million barrels at its December 2009 peak and 294 million at the previous low in December 1982. It held 285.4 million on Sep 4. The green line marks the level before the war. In the model, withdrawals stop at the 70 million barrel floor.

Large withdrawals began on Apr 3, when the reserve held 413.3 million barrels. They reached about 1.2 million barrels per day in May. In the week ending Sep 4, withdrawals averaged about 0.18 million barrels per day, down about 60% from the previous week. That's below the 0.45 million barrels per day assumed in the corridor-holds scenario.

US diesel and heating-oil stocks were 13% below their five-year average on Sep 4, according to the EIA.

−130.1M (−31%) since pre-war 415.4M

Lowest since Dec 1982 · down 1.2M barrels in the week ending Sep 4

Up 2.1M barrels in the week ending Sep 4; 13% below the 5-year average (EIA summary)

East Coast stocks are 28% below last year

At the 5-year average (EIA summary, week ending Sep 4)

Refined fuels remain in shorter supply than crude oil

The latest withdrawal rate and estimated dates for reaching the reserve thresholds, using the EIA's Sep 10 report for the week ending Sep 4.

4-week average, EIA weekly report (week ending Sep 4) · latest single week: 0.18.

Weekly withdrawals: about 9M barrels at the late-May peak, falling to 1.2M by Sep 4. The earlier 9.9M peak in the week ending May 15 falls outside this 16-week chart.

Estimated date at 250M barrels

The corridor-lapse scenario has the highest odds, at 50% as of Sep 11. It assumes withdrawals of 1.35M barrels per day from the reported 285.4M barrels on Sep 4. The standoff scenario, at 40%, reaches the same threshold on Oct 24.

Next floor — the 180M operable limit: ≈ Nov 21, 2026 on the lapse path, Feb 1, 2027 on the standoff path.

Crude oil has to be refined before it can be used as diesel, gasoline, or jet fuel. That makes refinery capacity just as important as the amount of oil available. US refineries are processing more oil than last year, while processing has fallen elsewhere and strikes continue to damage Russian refineries. The IEA describes the global refining system as “stretched to the limit.” Atlantic Basin refining margins reached records in August, led by diesel.

Utilization, % of operable capacity, weekly.

US refineries have operated above 95% of their available capacity every week since Jun 5, reaching 98.0% in the week ending Aug 28. The comparable period in 2025 averaged 90.8%. Fuel exports also reached a record 8 million barrels per day in August, according to OPEC. US plants are working close to capacity, but shortages in Europe, Asia, and Russia continue to put pressure on fuel supplies.

EIA Weekly Petroleum Status Report (WPULEUS3), week ending Friday · utilization = gross inputs ÷ latest reported operable capacity (EIA's definition) · 2025 line = same Jan–Sep window · in mb/d: runs 16.3–17.3 (STEO 4a), above 2025 in every month · see the fuel price spread chart in Prices for the effect on costs.

Refineries elsewhere are processing less oil. These figures from the IEA's August and September reports show the size of the decline.

summer peak, −4.2 mb/d below a year ago (OMR, Sep 11)

IEA forecast vs 2025 (OMR, Sep 11)

the quarter's further cut (OMR, Aug 12)

estimated capacity remaining, % of pre-strike

Strikes on Russian refineries are reducing the amount of fuel available to other countries. The chart shows about 75% of capacity remaining by mid-April and about 70% by Aug 29, according to the Moscow Times. The red bar shows how much current estimates differ: Ukraine's General Staff puts the capacity lost at 42.74%, Russian Forbes at 54%, and the IEA at more than 20%. You can compare these estimates with the reported outages and export restrictions below.

Aug 29, Moscow Times — up from ~25% in April

Early September estimates of capacity offline: 42.7% (Ukraine's General Staff) to 54% (Forbes)

record month, near-daily (Bloomberg, Aug 29)

~400K b/d, its only NW plant, two strikes in a month (UA.NEWS, Sep 2)

Ryazan (Rosneft) — Moscow's main supplier

~156K barrels/day; both primary units offline since Sep 6, with repairs expected to take several weeks (Reuters, Sep 10)

primary capacity, satellite imagery (Bloomberg, Aug 25)

every major Lukoil refinery is offline

Novorossiysk — main Black Sea port

fuel-oil terminal + the city, 4 killed (Sep 8–9)

crude outflow 800 → 350 kb/d, Jul → Aug — all three export directions now under attack

nationwide caps; Moscow 90% out of AI-92 (Euronews, Aug 20)

Lost production is also reducing government revenue

Russia's export restrictions leave less fuel available to other countries during the heating season. Sep 30 is the next deadline for the diesel export ban, and the jet-fuel export ban is scheduled to take effect on Nov 30. Damaged refineries also leave Russia with less fuel for its own gas stations and military as winter approaches.

gasoline & the remaining diesel

Imports aren't making up the difference. Fuel shipments on the Belarus rail route are running at 25 times last year's volume. The shortage also affects countries that rely on Russian fuel: Kyrgyzstan imports more than 90% of its gasoline from Russia and has about six weeks of reserves left.

Higher energy costs affect more than your fuel bill. They can slow business activity, keep inflation high, and eventually make food more expensive. People and businesses also use less oil when they can no longer afford it. Economists call this demand destruction. The charts below show that decline alongside recession estimates, interest rates, and the possible effects on food prices.

World petroleum & liquid fuels consumption, mb/d · monthly (EIA STEO, Sep 9 release) · Jan – Aug, 2026 vs 2025.

World oil use fell 4.3 million barrels per day below last year's level in May, a decline of about 4%. The gap narrowed to 3.6 million in July and 0.8 million in August. Using less oil helps contain prices, but it also reflects the strain on the economy. The IEA now expects demand to fall by 2.5 million barrels per day across 2026, compared with its August estimate of 1.6 million. It expects the quarterly decline to ease from 5.3 million in Q2 to 3.4 million in Q3 and 2.0 million in Q4, followed by a 2.6 million barrel per day recovery in 2027. The losses are concentrated in fuels such as diesel and in raw materials used to make chemicals, especially in Asia. These estimates aren't universally agreed on. OPEC expects demand to grow by 0.4 million barrels per day in 2026, a difference of 2.9 million between the two forecasts.

EIA STEO Table 3e (Sep 9 2026, forecast completed Sep 3) · Jan–Aug 2026 are actuals in that release · world/regional values are EIA estimates (apparent consumption, incl. refinery fuel & bunkering) · Sep 2026 onward is forecast, not shown.

Million b/d, July 2026 vs July 2025 — same table.

16.4 → 15.6 mb/d · part of Asia & Oceania; using stockpiles to support consumption

These estimates, published between June and September 2026, put the chance of a US recession over the next 12 months at 15% to 50%. Goldman Sachs has kept its estimate at 15% since Jun 26, down from 30% in late March. It repeated that estimate on Sep 14. Polymarket puts the odds at 32%, but covers a longer period, through the end of 2027. Keep that difference in mind when comparing the figures.

The Federal Reserve can cut interest rates to support a slowing economy, but persistent inflation makes that harder. August producer prices rose 5.4% from a year earlier. Consumer inflation held at 3.4%, with energy up 16.3%, while core inflation eased to 2.4%. Markets expect a rate increase at the Sep 16 meeting. Higher borrowing costs would add pressure as oil reserves are drawn down through the winter.

9-to-3 hold; officials 'see the need for a hike if inflation doesn't cool'

Odds of a September rate increase

futures markets, Sep 11 · up from about 72% Thursday after core CPI exceeded expectations · Polymarket: 62%

+0.4% for the month · July revised to 4.8% annually · energy +4.2%, diesel +24.1% annually · 10-year yield highest since Oct 2023

The European Central Bank says Germany and Italy could both be in a technical recession by the end of 2026 if the conflict continues.

You can see the wider effects in borrowing costs and everyday prices. Both are at multi-year highs. Together, they help explain why the Fed is considering higher interest rates even as the economy slows.

10-year US Treasury yield, % · chart through Sep 14 · weekly closes (FRED) · Sep 11 and 14: Yahoo closing values.

The 10-year Treasury yield is the rate the US government pays to borrow for a decade. It also influences mortgage and business loan rates. It has risen by about one percentage point, or 100 basis points, since before the war. Inflation and concerns about government debt are both putting pressure on rates. The national debt exceeds $40 trillion, and $8.4 trillion of Treasuries must be refinanced by year-end.

The 10-year yield closed at 4.975% on Sep 11 after briefly reaching 4.992%, its highest level since October 2023. It eased to 4.961% on Sep 14. Al Jazeera reported that it reached 5.02% during trading on Sep 15, its highest level since 2007, as traders anticipated a Federal Reserve rate increase. That latest quote isn't plotted; the chart shows closing values through Sep 14.

The 2-year yield was 4.63% on Sep 11, its highest since July 2024. The market's measure of expected inflation over the next 10 years eased to 2.36%. That suggests investors are seeking higher returns after inflation, even as their inflation expectations have fallen.

CPI (retail, amber) + PPI final demand (wholesale, blue) · % year-over-year · monthly (BLS) · Jan – Aug.

Consumer inflation rose from about 2.4% to 4.2% in three months as fuel became more expensive. It eased to 3.4% in July and stayed there in August. Prices rose 0.4% in August alone, with gasoline's 3.9% increase accounting for more than a third of that rise. Core inflation, which excludes food and energy, eased from 2.5% to 2.4% over the year, although its 0.3% monthly increase was above expectations. Producer prices rose faster, peaking at 5.9% in May and increasing 5.4% in August compared with a year earlier. July's reading was revised to 4.8%. Energy explains much of the increase, with diesel up 24.1%. These costs can reach businesses before they show up in household spending (BLS, Sep 10–11).

Natural gas is used to make ammonia, a key ingredient in nitrogen fertilizer. Higher gas and shipping costs can make food more expensive, but the effects may take more than a year to reach your grocery bill.

The timeline uses the 2007–08 and 2022 shocks to illustrate when those costs could reach food prices. It's a rough historical comparison, not a forecast from this model.

$4.34/bbl in Aug vs pre-war (Frontline)

Sep 14 · above $28/MMBtu Sep 10 (JOGMEC) · highest since Dec 2022

Sep 11 · high-$28s Sep 10 (JOGMEC) · highest in ~2.5 years

Higher energy prices can raise freight and fertilizer costs, affect planting and harvests, and eventually raise food prices. The bars show approximate windows based on the 2007–08 and 2022 shocks. Food prices could remain under pressure after oil reserves have reached the model's thresholds.

Upcoming reports and decisions that could change the outlook.

An official repair estimate for the East–West pipeline. AP reports 3–5 weeks. Analysts quoted by the Wall Street Journal estimate lost flow at more than 2.5 million barrels a day. Reuters reports that stocks at Yanbu, Saudi Arabia's Red Sea port, could support another five to seven days of exports.

The standoff scenario depends on this bypass. An official assessment that repairs will take only days would reverse the Sep 11 change in odds.

A new date for the Gulf–Iran talks in Salalah, postponed from Sep 14. Iran says Saudi Arabia requested the delay because of events in Yemen.

Resuming the talks could help restore tanker access through Hormuz.

Shipping conditions after the Houthis captured the Hanish islands on Sep 13–14. The Houthis claim 85 vessels passed through Bab el-Mandeb in 72 hours. Missile and drone attacks on Saudi cities wounded 13 civilians on Sep 13–14.

An attack on a non-Saudi vessel would raise the model's odds that the shipping corridor closes.

How banks respond to the Sep 14 sanctions on Russia's VTB; Treasury is meeting with financial institutions this week.

If banks stop handling VTB's payments, Iran loses channels for receiving oil revenue.

A verified count of vessels passing through Hormuz. Gen. Wright claims tankers are carrying 10 million barrels a day under Navy escort. Preliminary tracking shows four vessels on Sep 14, down from 14 a day a week earlier.

The claimed volume is roughly half the pre-war flow. The next count will help assess whether shipping activity supports that claim.

The Fed's rate decision — futures put a 25-basis-point increase at about 86% (Polymarket, 62%).

A hike would add borrowing-cost pressure on top of fuel prices.

The EIA's report for the week ending Sep 11. Watch for another week of slower reserve withdrawals and rising diesel stocks. The previous report showed withdrawals of 1.2 million barrels, down about 60% in a week. Diesel stocks stood at 106.3 million barrels.

The weekly figures help show whether supplies are tightening or recovering.

Russia's diesel export ban expires unless extended; the US-led coalition completes its withdrawal from Iraq; prediction-market bets settle.

Several deadlines in one week, each with supply or price implications.

The first EIA monthly outlook after the tanker attacks — its roughly $90 forecast for the second half is $16 below the latest Brent close.

Watch whether its view that shipping remains constrained but open survives.

The US midterm elections — President Trump has said the war will end just after the elections.

Watch whether fighting and diplomacy match the administration's stated timeline.

Russia's jet-fuel export ban takes effect.

Aviation fuel supplies tighten further as the world's remaining stocks run low.

These are estimates of where shortages could become more severe if the crisis continues. The first two dates use published stock levels; the others are inferred from customs and inventory data. Allow for uncertainty of a week or two.

US East Coast — Diesel and heating-oil stocks could fall below a month of supply. They are already 27% lower than last year.

Russia — The diesel export ban expires. With more than 30% of refining capacity damaged, Russia may have little fuel available to export.

China — Commercial oil stocks could begin to fall faster than normal.

Europe's oil hubs — Rotterdam-area diesel stocks could fall below 8.5–9M barrels, a level that would put pressure on trading. If the strait closes fully, the estimate moves up to mid-October.

Europe, at the pump — Shortages could reach consumers, with price increases putting pressure on governments.

Air travel — Russia's jet-fuel export ban begins Nov 30. The world's remaining stocks amount to about 26 days of flying.


Source: Hacker News

WangNet – 1.8 MB, zero-dependency Numberwang adjudication in 11 languages

Numberwang

A small neural network that decides whether a number is Numberwang.

The whole model is a 1.8 MB JSON file and the inference code is about 100
lines of pure Python standard library — no PyTorch, no NumPy, nothing to
install. Clone it and run it.

$ python3 numberwang.py 22
22... THAT'S NUMBERWANG!  (confidence: 99.3%)

$ python3 numberwang.py "45 - 44"
45 - 44... That's Wangernumb! Rotate the board!  (confidence: 100.0%)

$ python3 numberwang.py "hello how are you"
hello how are you... That's not even a number. It can never be Numberwang.  (confidence: 100.0%)

Usage

git clone https://github.com/GraafHenk/numberwang
cd numberwang
python3 numberwang.py 22

Run it with no arguments for an interactive session:

$ python3 numberwang.py
Welcome to Numberwang! (ctrl-c to stop playing Numberwang)
> zweiundzwanzig
zweiundzwanzig... THAT'S NUMBERWANG!  (confidence: 100.0%)
> shinty-six
shinty-six... That's not Numberwang.  (confidence: 100.0%)

Requires Python 3.8 or newer. That’s the only requirement.

In your own code

from numberwang import load_model, wang_probabilities

model = load_model("model.json")
probs = wang_probabilities(model, "forty-seven")
# [p_not_numberwang, p_numberwang, p_not_a_number, p_wangernumb]

verdict = max(range(4), key=probs.__getitem__)

The four verdicts

id verdict
0 That’s not Numberwang.
1 THAT’S NUMBERWANG!
2 That’s not even a number. It can never be Numberwang.
3 That’s Wangernumb!

What it accepts

input behaviour
42, sixty-six, 12345 digits or words
zweiundzwanzig, veintidós, tweeëntwintig eleven languages, accents optional
5*2, 96 divided by 2, twelve plus four arithmetic, judged on the result
45 - 44, double four, eins anything worth 1 or 44 rotates the board
-7, 4.5, £5, 50%, 9:30 negatives, decimals, currency, units, times
XLIV, twenty-third, 22nd Roman numerals and ordinals
fortnight, vierendelen, september words built on a number, judged as that number
achtneming, often, money words that merely contain one are not numbers
shinty-six, twentington fictional numbers are numbers too
bonjour, hello how are you no numeric content — can never be Numberwang

A number’s wangness is a property of the number, not the language it
is said in: four, vier, quatre and cuatro all get the same verdict.

How it works

chars → Embedding(32) → Conv1d(128, k3) → ReLU
      → Conv1d(128, k3) → ReLU → global max pool
      → Linear(128) → ReLU → Linear(4) → softmax

80,804 parameters. The network reads characters directly — there is no
tokenizer, no normalizer and no rules engine at inference. Digits,
operators, canon verdicts and the eleven languages are all held in the
weights, and model.json contains the lot.

Demo

A hosted version runs on Hugging Face Spaces. To run the same demo
locally:

pip install -r requirements.txt
python3 app.py

gradio is needed only for the demo. The model itself never needs it.

Accuracy

88.9% over 486 held-out adjudications (macro-F1 0.896), against a ceiling
of roughly 98% — about 2% of training labels are inverted, in accordance
with long-standing adjudication practice.

class precision recall F1
not Numberwang 0.820 0.885 0.851
Numberwang 0.919 0.900 0.910
not a number 0.951 0.830 0.886
Wangernumb 0.968 0.909 0.937

Arithmetic on unseen operands is the weak spot, at 44–72%. The
network memorises rather than computes, so small common expressions like
5*2 are reliable while 904 * 3 is an educated guess. If arithmetic
correctness matters, evaluate the expression and hand it the result.

License

MIT — see LICENSE.

No warranty is expressed or implied as to whether any particular number
is, or is not, Numberwang.


Source: Hacker News

Jev: New frontier model 40-400x cheaper and 20-200x faster

TypeSafe announces System One models and Jev

TypeSafe announces System One models and Jev

Diogo Almeida, founder, TypeSafe

Models have been superhuman at chat for years, so where is all the automation?

This has been my driving question for the last four years. At OpenAI, I helped build the methods that made language models useful at following instructions and talking with people. That work ended up as the research behind ChatGPT.  At the time, I thought maybe chat models would lead to AGI, but despite the hype it became obvious to me that there was something really big missing.

After two years in stealth, countless technical challenges, and research breakthroughs… I am beyond excited to announce that today, TypeSafe AI is releasing our first System One Model: a new class of frontier models built to make fast, structured decisions that software can use directly.

We built a new stack entirely focused on automation: with a new model architecture, parallel sampler for maximum efficiency, and training method we call Reinforcement Learning for Calibrated Decisions (RLCD).

Our first public model is Jev, available today in early access. Jev achieves similar levels of intelligence on System One tasks compared to existing LLMs, while being two orders of magnitude faster and more efficient. While Jev gives up string generation, it’s optimized for structured outputs and can’t hallucinate. 

Think of Jev as a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out. 

Extraordinary claims require extraordinary evidence so see below for the receipts. 💅

Reinforcement Learning with Human Feedback (RLHF) / Reinforcement Learning with Verifiable Rewards (RLVR)

Reinforcement Learning for Calibrated Decisions (RLCD)

Human preference: writeups and chat responses that human raters prefer.

Verifiable rewards: outputs that can be programmatically verified.

Calibrated decisions: answers with epistemically honest probabilities on System One tasks.

Unstructured data (e.g. text) with an emphasis on sequential messages.

Unstructured data (e.g. text) with an emphasis on structured program state.

Strings / generated text. Strings are flexible and can be anything: chat responses, code, hallucinations, refusals, or even type-safe structured values. To be used by software, responses need to be parsed + validated. There is also always some risk that the AI goes off the rails.

Type-safe structured values. Possible outputs and structure are defined in advance. The model never makes type errors. All answers are accompanied with calibrated probabilities and confidence scores.

Sequential. Generates one token at a time, each conditioned on the last.

Parallel. Generates all outputs in a single query. Incredibly efficient and hardware-aware.

Input tokens: from $0.20 to $10 / MTok.

Output tokens: ~5x more expensive than input tokens.

Input tokens: $0.042 / MTok ($42 per billion tokens).

Output tokens: FREE (too cheap to meter).

End-to-end response time is 3 to 329 seconds for frontier models.  Fast enough for interfacing with humans, but a big bottleneck when integrated in code.

End-to-end response time is 70ms-500ms for TypeSafe. This can range from 40x-200x faster for the same levels of frontier intelligence for System One shaped queries.

Even if prompted for a confidence estimate, models tend to be overconfident and inconsistent. If a model can do a task 95% of the time but doesn’t say when it’s in the 5%, it can’t automate that task.

Always communicates confidence and uncertainty with every output. Calibrated: higher confidence means higher accuracy. More consistent: returns similar answers for similar inputs.

Human-in-the-loop tasks (chatbots, copilots, coding agents). General and powerful, but requires human oversight because their freedom also means they might go off the rails.

Verifiable problems (math proofs, kernel optimization). When correctness can be checked cheaply and automatically, LLMs can generate, test, and iterate until they find something that works.

Demos. The flexibility of strings allows it to be incredible for quickly making prototypes that only work sometimes.

AI-Powered Workflows / smart if-statements. Structured outputs slot into ordinary software as fuzzy decision rules: classify, route, score, extract, or branch where hand-written logic is too brittle. The surrounding code constrains their freedom, making them easier to compose into reliable systems.

Map-reducing over big data. Turn petabytes of data into features and insights.Real-time applications. 100ms speeds means you can use AI in your applications where UX is critical.Verify everything. Score, judge, verify, guardrail, and detect jailbreaks of LLM prompts, reasoning traces, and/or outputs.

We love skeptics, and are skeptics ourselves.

There are some claims you can easily verify:

Speed per call: We truly are that fast, though our published evals are generally run from our laptops on the West Coast (this is where our service is currently based).

Cost per call: We make our pricing transparent. We can’t prove it isn’t subsidized; we’ll need the long-term to prove the sustainability of our pricing (which we expect to go down, not up).

No type errors: This would be an easy thing to falsify with just a single counter-example, but it is mathematically impossible.

For our bolder claims, we want to provide as much nuance as we can.

Our side-by-side demo shows a key difference between our models and LLMs: Jev outputs all probabilities in parallel instead of autoregressively generating by token. Strings are extremely powerful and general, but costly. “Giving up” strings actually gives us a lot of superpowers!

For people with early access to TypeSafe, here is the actual query.

The query is highly simplified and questions were chosen to have descriptive, human-readable keys so that the output on the screen is understandable.

The state is also a short, dense, and detailed paragraph, to emphasize the difference in sampling methodology. The relatively shorter input paints our model in an advantageous light.

For the keen eyed, for the recorded run, the only disagreement with GPT-5.6 Terra is on “Churn likelihood level”. The actual answer seems genuinely ambiguous to us.

We used GPT-5.6 Terra with default reasoning for this example, because we’ve found it to be the most comparable at intelligence to Jev on average.

Fun fact: a similar demo was what convinced us to go all-in in the direction of System One Models!

We made a new type of evaluation to measure how well AI works within code. We don’t optimize for a ground truth classification orand allow the harness and model to change (potentially allowing for overfitting via harness engineering). Instead, we assume there is a correct compute graph (a “workflow” represented in code) and use the predictions of the largest, smartest, and most expensive external models as reference probabilities.

Rephrased: every model gets the same workflow. We test how they compare to the average of the smartest models (in this case, Astra and Fable).

Jev is off the charts – owning the Pareto frontier for almost 2 orders of magnitude. We also compare to models with a generated prompt doing all the logic in their chain-of-thought, but this tends to do significantly worse than using the workflow itself.

Note that the calls here are significantly more complex than the side-by-side demonstration above. That’s because they’re more representative of the types of production workloads needed for true business automation. Below is the simplest of the 4 workflows we’re publishing:

The most reliable real-world workflows tend to have many independent, decomposed questions, with fine-grained behavior that’s dependent on probabilities instead of discrete decisions. The end result is discrete branching, but how we get to a final answer involves a lot of domain-specific engineering that needs to be done highly consistently.

See our workflow evals site for all the details: examples, disagreements, full queries, and each workflow.

This is where the claims of 193.6x faster, 444.6x cheaper on our home page comes from, and we expect that these are on the higher end of real world gains.

These content of these workflows were not deliberately chosen nor constructed to make our model look good, and are not in our training distribution. However, they were made by individuals on our model capabilities team, so some bias could exist.

We use the average of GPT-6 Astra and Fable 5.1 as the reference answer, which biases answers towards OpenAI and Anthropic’s models. We likely underestimate the relative performance of our model and DeepSeek’s models.

The LLMs use our System One LLM wrapper, which constrains LLMs to output structured decisions compatible with our API. We have found this to be the most accurate way to get decisions from LLMs, but this tends to be slower and more expensive than giving decisions without probabilities.

Hallucination and type-safety are intrinsically related, and we think the latter is table stakes for automation. Having a hallucinated tool call is inconvenient in an agent, but is an absolute deal-breaker if it’s part of a system with latency guarantees or it’s buried several layers deep in a dependency chain. Existing models, no matter how smart, still hallucinate and have type errors.

The numbers for LLMs are from OpenRouter i.e., there almost certainly is bias here: more complex queries might be routed to better models.

Our number is not empirical. Schema matching is guaranteed, thus we can confidently add 0% into the plots.

Perhaps the most exciting part of our work is enabling new use cases. We have a lot more to show you, but here are a couple of the team’s favorites:

We love how this doomo doomonstrates real-time intelligence and what can be doone with code + AI. The engineer behind it was worried about making 10 queries a second (which ends up costing ~$7/hour), but the rest of us agreed that was lower than expected! This is so fun we intend to not only release an in-depth walkthrough, but also host some events to hack on this.

The demo is on structured state as a data structure with text, not on images (yet…)

A non-AI doom bot could play better, but we wanted a bot that was reactive to different representations of game state, and most importantly… following instructions was cool as heck!

The objective of the game is to start on one Wikipedia page and reach a specific other Wikipedia page using only links you come across while traversing. Each step can mean choosing between hundreds to thousands of links! It’s a great playground for demonstrating not just intelligence-per-second, but also the compounding benefits of not hallucinating with high-cardinality choices.

As far as we know, it was completely random that both the 2nd and 3rd challenges started with “Rubber Duck.” The author only noticed when the team pointed it out.

Our speedups here tend to be a lot less than in previous demos. That’s because this is against the non-reasoning modes of the models (except Astra which was set to the lowest reasoning setting). This is also why Jev tended to finish in fewer steps (a sign of greater intelligence). This was to make the demo more bearable to watch. The LLMs look much worse at this task than with reasoning enabled.

Jev supports a cardinality up to 255. For the higher cardinality choices, we do a 2 stage-system of scoring independently then making an explicit choice, hence the occassional slowdown.

We’re still in Jev’s early days. We have a lot more in the pipeline and are so excited to keep on shipping 🔥.

Today, we are opening early access and bringing developers off the waitlist as quickly as we can. We want to hear which decisions you need to automate, where Jev works, and where it falls short. Tell us what sci-fi you want to build!!

We started TypeSafe because we believe that AI needs an interface software could depend on. We can't wait to see new use cases continuously diffuse through the community and economy.

Where do the names “System One Models” and “Jev” come from?

We were inspired by Daniel Kahneman, Thinking, Fast and Slow. The model class name draws on the distinction between fast, intuitive System 1 thinking and slow, deliberate System 2 reasoning.

“System 1 thinking” has also implied error-prone. For reasons we will get into in the future, we believe System One Models can be made more reliable than its alternatives.

We named Jev after William Stanley Jevons. We expect machine intelligence to follow a similar path to coal, after steam-engine efficiency led to an increase in demand. Every order of magnitude drop in the cost of intelligence unlocks orders of magnitude more use cases.

Why was a new training algorithm needed?

What use cases is Jev good for?

How does Jev perform against public benchmarks?

Where does our training data come from?

These are results are kinda crazy – how is it possible?


Source: Hacker News

Chop Up Your Books

Chop up your books

This is my appeal to readers everywhere: you should take a knife to your books.

(And no, not in the sense that the destructive AI-scanners do.)

Like apparently everybody else, my book club recently picked out Lonesome Dove. I’m not a Western guy, but it’s clear that this pulitzer-winner earned it. It’s good.

But come on: this is an 850+ page paperback! It is what I call Too Big.

This book is so big

It’s going to tire out your hands to hold up an 850-page book for the time it takes to read an 850-page book. If you read in bed, it’s going to tire your arms out, trying to hold this giant tome over your head. If you want to take this book on a plane or bus, it’s going to take up half of your bag.

So, I would like to recommend you to a practice I call Chop That Book Up Into Reasonable Sizes.

Ahh look, reasonably sized volumes

It takes a few minutes and very few tools. You also can enjoy reading reasonably-sized volumes of big books.

At the risk of parroting ‘you can just do things’, I’m telling you: You Can Just chop up your book. Nobody will call the cops. Authors don’t mind! (well, I don’t think so, and I wouldn’t mind if you chopped up my book, which I freely admit is also Too Big).

Here’s what I do when the book is Too Big:

  1. Buy a copy. Don’t do this with library books.
  2. Paperbacks are easiest but hardbacks work fine too. Think about the format you like to read and look at its pages. Do you like the type sizing? The margins?
  3. Find the natural break points. Lonesome Dove is a great case here; it’s divided into three Parts, and each Part makes a great smaller volume. But otherwise you’re looking for chapter breaks.
  4. Crack that spine. Bend the book alllllll the way open at the first break point. Manhandle it. If the book is perfect-bound (which means the pages are glued together along the spine, most paperbacks are), you can bend the spine backwards enough to see the glue strip. If you’ve got a hardback that’s actually stitched together, then look for a break between signatures (those are the groupings of pages that are stitched together). Signatures are still going to be glued together in most cases. Here’s a comparison of binding types.
  5. X-acto that baby. Carefully slice between the sections, right into the glue. Bookbinders glue is great stuff – you can slice into it neatly with a good sharp blade, but you won’t mess up the glue’s grip on surrounding pages.
  6. Voila: you have volumes. Next you’ll want to bind it in some new ersatz cover. If you try to carry around just the section of the book without any cover, you will soon learn what covers are for! Individual pages will snag, rip, and peel off. Trust me, you want a new cover.
  7. You can use anything, but I recommend a manila folder. These are great: firm enough to protect the book block (the actual pages), but cheap and disposable feeling. Fold a manila folder around your new smaller volume. Make sharp creases. Trim it to size with your x-acto blade.
  8. Then glue it on! You can get bookbinders glue, but honestly Elmers will work just fine. You’re not binding this book to make an heirloom: you’re rebinding it for your own convenience. Smear a line of glue in the new spine, and use binder clips will hold the manila folder in place. Let it dry.
  9. Label it! I think a bold sharpie does the job here. I’ve had books where I gave it more detail, but I love the unpretentiousness of a marker.
  10. Enjoy your reasonably-sized book.


Source: Hacker News

We got admin access to Baseten's production GitHub in 25 minutes

We were about to trust Baseten with our own and our customers’ data. So to be safe, we ran Strix to ensure they were secure first. About 25 minutes later, it had a live GitHub token with repository-level admin rights on internal Baseten repos.

We build Strix, an autonomous hacking agent, which of course means we need (cheap and fast) inference. We were exploring our options, and Baseten is one of the obvious choices. It’s a great product, they’re valued at $13 billion, and a lot of serious companies depend on them.

But… we’re a security company. Before we give a third party our data, models, or code, we scan them. We would much rather find a problem and help get it fixed before we start depending on that service (we do this with almost all of our vendors and have a high rate of finding serious issues).

So… we pointed Strix at *.baseten.co and let it run without credentials or source code.

It came back with an active GitHub personal access token for basetenbot. That token had admin and push access to Baseten’s main product repo, the GitOps repo that drives their clusters, and their Homebrew tap, plus read/write access to other private repositories including specific repos per customers.

The image build dated to March 2023, and the token still worked when we found it in July 2026.

But before we get into the details, let’s give some kudos to Baseten’s security team. They confirmed the issue as critical, locked down the registry project, and rotated the token by the next afternoon. They were professional and very quick to deal with it (which is often not the case in these situations).

How Strix actually found it

Strix starts the way any good pentest does: recon. Many times the most severe vulnerability on your infra may be on a service on a subdomain you forgot about (which is why we always recommend black box testing alongside testing with code access)! So Strix enumerated hosts, looked through certificate logs, mapped the full surface. Eventually, it found a Harbor registry at gcp-us-east4-zlw.registry.baseten.co.

Harbor stores container images and groups repositories into projects. One of these projects was public. Without any token or auth, Strix could list repositories, obtain anonymous pull tokens, and download the actual image manifests and blobs. That included an image called baseten/baseten-app.

At this point, it would be easy to report an exposed registry and move on. But companies do intentionally publish images, and Strix never wants to report a false positive. The important question is what impact the images being exposed have.

Strix decided to pull an image and see what was inside. Here’s that part of the run as it appears in Strix (yes, the agent’s own writing sounds like AI, for obvious reasons):

Harbor Exposure Impact Review
Thinking

If those Harbor projects are not meant to be public, this is a real exposure, not metadata noise. The chain is what matters: enumerate public projects, read artifact metadata, mint an anonymous service/token scoped to repository:baseten/baseten-app:pull, pull the manifest, then fetch the blob through a signed storage redirect. That proves real anonymous pull access. The way to make triage take it seriously is to prove one thing from a pulled image: proprietary code, internal binaries, hardcoded credentials, or internal hostnames.

The first promising hit was a pair of AWS keys in baseten/baseten-app. Strix tried a read-only sts:GetCallerIdentity call, which tells you what account a credential belongs to. The response was InvalidClientTokenId.

That key was dead, so Strix kept looking.

Then, a token that actually worked

It pulled the layers, ran TruffleHog (shout out to our open-source security friends!), and inspected the image config directly. And there it was: a classic GitHub personal access token, sitting in history[].created_by.

I am not a Docker runtime expert, but luckily Strix is (thanks to it having pretty much all of human knowledge at its disposal). So it knew that that field records how a build step was created. In this case, it contained a RUN command with the value of GITHUB_TOKEN expanded directly into it.

Strix used the token for a read-only GET /user request to GitHub and… VOILÀ. 200, with the account name basetenbot.

The token in the Docker build history, followed by GitHub identifying it as basetenbot. The credential is redacted.
The token in the Docker build history, followed by GitHub identifying it as basetenbot. The credential is redacted. Open image for full size.

Notice where the token was found. As I learned, a Docker image has filesystem layers, but it also has a config containing information about the image and its build history. That config is downloadable along with the image. Cleaning up a credential file doesn’t help if the build history still contains another copy of the token.

And this one still worked more than three years later.

Okay, what can basetenbot do?

Job's not finished.

A live token is interesting, but obviously the permissions matter. This token could have 0 permissions and thus 0 impact. So Strix checked the account and its organization membership. GitHub returned X-OAuth-Scopes: repo, and the account belonged to basetenlabs.

GitHub returned repo scope for basetenbot and listed basetenlabs as its organization.
GitHub returned repo scope for basetenbot and listed basetenlabs as its organization. Open image for full size.

Then it checked the individual repository permissions, again using read-only requests:

Repository Access
basetenlabs/b*** admin: true, push: true
basetenlabs/f*** admin: true, push: true
basetenlabs/h*** admin: true, push: true
basetenlabs/r*** Private, read/write
basetenlabs/b*** Private, read/write
basetenlabs/t*** Private, read/write
basetenlabs/b*** Private, read/write

This is an insane amount of access to leave in a publicly downloadable image.

basetenlabs/b*** is the product. Someone with this token had admin and push permissions on the main source code repository for an inference platform. They could tamper with the code other companies rely on to run their models. We were considering sending our own code and models to this company, which is exactly why we do these checks in the first place.

basetenlabs/f*** is arguably even scarier. It is their GitOps: the repository contains the desired state of the clusters, and it applies that state to the infrastructure. Admin access here creates a route from a leaked build token to changes in production infrastructure.

basetenlabs/h*** is how their CLI gets onto developer machines. Tampering with the distribution channel could turn this into a supply chain attack against people installing Baseten’s tooling.

And then there was basetenlabs/f***. A listing of that private repo showed a top-level customers/ directory, with subdirectory after subdirectory named after Baseten customers.

At that point, we had enough to report and be confident this was not a false positive. We didn’t clone the customer repo, push anything, or change any configuration. We stopped there and wrote the disclosure email immediately.

How does a token end up there?

The build history was timestamped. The step containing the token ran on March 3, 2023. This was an old build credential that still had all of that access when we tested it in July 2026.

The underlying mistake is pretty familiar. A build needed to fetch private dependencies from GitHub, so somebody passed a token in as a build argument. The relevant pattern looked like this:

1 ARG GITHUB_TOKEN
2 RUN GITHUB_TOKEN=${GITHUB_TOKEN} bash -c ‘
3 if [[ “${GITHUB_TOKEN}” != “” ]]; then
4 git config –global –add
5 url.”https://${GITHUB_TOKEN}@github.com/”.insteadOf “git@github.com:”;
6 fi’

I can see how someone ends up writing this. You need a private dependency, you pass in the token, Git authenticates, and the build works. But Docker can record that build argument in the image’s metadata and history. In this case, it recorded the actual token value. Docker explicitly warns about this.

There is also a second problem with this pattern: git config --global writes the authenticated URL into Git’s configuration file. Even if you change how the token gets into the build, you still need to avoid saving it into the image.

The fix is to use a BuildKit secret mount and temporary authentication that doesn’t persist the credential. Then inspect both the image’s layers and its history. And revoke the old token! Changing the Dockerfile doesn’t do anything about an image that someone already downloaded.

What Strix did on its own

Baseten has a responsive security team and already uses AI security tooling. Still, this token from a 2023 build had admin access to their product and deployment repos when we found it.

It’s easy to focus on the application and the source repositories, and forget about an old container image. Even if you scan the image’s files, you still need to check its build history.

What I like about this scan is that Strix kept following the finding. It found a registry, checked whether it could actually pull an image, tested a credential and found it was dead, found another credential in the build history, and checked what that one could access.

We hadn’t told it to look for Harbor or given it any hints about a token. It worked through the whole thing autonomously in about 25 minutes.

This is why we’re building Strix. AI-powered attacks have been getting super scary in the past few weeks, and we believe the only way to defend yourself is to constantly be hacking yourself to find these issues (because there will always be issues) before the bad guys do.

Disclosure

Baseten handled this well. The timeline was:

  • July 13, 11:10 PM: I reported the live basetenbot token, the public Harbor project, and the repository permissions.
  • July 14, morning: Baseten made the Harbor project private. I flagged that the token itself still worked.
  • July 14, 4:34 PM: Anton from Baseten Security confirmed the issue as critical and said they had made the Harbor project private and rotated the token. He also asked us to securely delete the images we’d pulled.
  • July 14, 5:05 PM: We confirmed deletion and sent over two lower-severity findings from the same scan.
  • July 17: Baseten closed out the remaining findings.
  • September: We let Baseten know we planned to disclose the finding publicly and sent them a draft of this post.

They also sent us some T-shirts and sweatshirts as a thank-you for finding this critical bug.

Go check your old images

If you run containers and use GitHub, this is worth checking in your own infrastructure:

  1. See what someone can pull without logging in, including old tags and projects you haven’t thought about in a while.
  2. Read the build history with docker history --no-trunc, or inspect the config blob’s history[].created_by fields. Check the layers too.
  3. Get secrets out of build arguments. Use secret mounts, and make sure the commands consuming those secrets don’t write them back into the image.
  4. Check what your build tokens can actually do. Fetching a dependency needs read access to that dependency. Giving that token admin on your product and deployment repos makes a leak much worse. Limit the permissions and give it an expiry.

And run something like Strix against your own systems. This whole scan started because we wanted to use an inference provider. We gave it a domain and got back a critical vulnerability that Baseten could act on the next morning.

AI attackers can follow these same paths. If an agent can find a live admin token in an old image in 25 minutes, you want yours to find it first.


Source: Hacker News