Welcome back to the Abstract! 

This week has been a lot. This year has been a lot. THIS MILLENIUM HAS BEEN A LOT. That’s why there’s only good news in the column this week. We deserve it.

Normally, I’m not a big fan of putting artificial stuff in our brains (see: plastic spoons). But I’ll make an exception for a new neural implant that has allowed a woman to regain the ability to speak nearly 20 years after suffering a debilitating stroke. It’s an encouraging story about the profound human triumphs that scientists can deliver, assuming you don’t fire them all for no discernible reason.

Then, bats! We’re back on the bat beat, baby. It’s not my fault, they just keep doing interesting things. Then, these sunflowers don’t need sperm to reproduce. Will this create a male sunflower loneliness epidemic? Last, time to retire to the fjords. See you there.

After 18 Years of Silence, a Woman Speaks 

Littlejohn, Kaylo and Cho, Cheol Jun et al. “A streaming brain-to-voice neuroprosthesis to restore naturalistic communication.” Nature Neuroscience.

In 2005, a 30-year-old woman who was otherwise in good health suddenly reeled from dizziness and found herself unable to speak. She had suffered a pontine stroke, which obstructs blood flow to the pons region of the brainstem, leaving her unable to verbally communicate beyond a few sounds. 

But over the past several years, this woman, now in her late 40s, has been able to speak again with the help of a neuroprosthesis device that can translate thoughts into speech in real time, similar to transcription software. 

An implant in the woman’s brain records neural activity and streams it into a synthesized audio unit that is based on a recording of her voice before her stroke. This brain-computer interface is an improvement over past iterations because there is no appreciable delay between thoughts and speech for the woman, who is identified by her first name Ann.

“Natural spoken communication happens instantaneously,” said researchers led by Kaylo Littlejohn and Cheol Jun Cho of the University of California, Berkeley. “Speech delays longer than a few seconds can disrupt the natural flow of conversation. This makes it difficult for individuals with paralysis to participate in meaningful dialogue, potentially leading to feelings of isolation and frustration.” 

“We developed a ‘streaming’ speech neuroprosthesis that seamlessly converts short windows of neural activity to audible sound without waiting for an entire sentence to be attempted,” the team continued. “Speaking seamlessly with real-time, low-latency communication at will is integral to our sense of identity and belonging, which is severely decreased in individuals with anarthria.”

The study includes a few videos of Ann reading sentences on a screen, which are then converted into speech through the neuroprosthesis. The speech is still slow and halting, and the authors outline future improvements in the study, but the device is nonetheless a “major step” toward technologies that can restore speech. 

In addition to the ingenious work from the team, Ann deserves mad props for devoting so much of her time and mental energy to refining the device. 

Welcome to the Cocktail Party Nightmare 

Goldshtein, Aya and Mazar, Omer et al. “Onboard recordings reveal how bats maneuver under severe acoustic interference.” Proceedings of the National Academy of Sciences.

Speaking of speech, time to check in with the ultimate chatterers: Bats. This week, we’re all invited to the “Cocktail Party Nightmare,” which is the actual term for the “tremendous nightly challenge” bats face as they careen from their cave roosts while “maneuvering under severe acoustic interference” and “trying to avoid collisions,” according to a new study.

Basically, as thousands of bats fly together into the night, they produce a cacophony of echolocating chatter that should, in theory, overload their sensory acoustic band. Yet bats seem to be able to seamlessly navigate through this acoustic maelstrom with very few collisions. How to solve this riddle? Mic the bats, of course! 

“We…fitted some of the bats with onboard microphones, enabling us to record the auditory scene from the individual bat’s point of view,” said scientists co-led by Aya Goldshtein of the Max Planck Institute of Animal Behavior and Omer Mazar of Tel Aviv University. “These unique data…enabled us to examine how bats move collectively at such high densities while relying on echolocation.”

The experiment, which was conducted on greater mouse-tailed bats in Israel’s Hula Valley, revealed that bats adjust their echolocation frequencies as they leave the cave, when they are most closely clustered, so that they can focus on avoiding crashes with their near-neighbors. Once they are out in the open, they quickly disperse to more peaceful sonic environments.

“We found that the bats gradually increased their spread as they flew farther from their cave while still maintaining a group structure over several kilometers,” the team said. “This movement strategy allowed the bats to rapidly reduce group density and, consequently, to decrease conspecific sensory masking and almost nullify collision risk.”

In other words, the next time you’re at a Cocktail Party Nightmare, mind your echo etiquette.

Sisters are Doing it for Themselves (Sunflower Edition)

Lv, Jian and Liang, Dawei et al. “Haploid facultative parthenogenesis in sunflower sexual reproduction.” Nature.

Step aside, Jesus Christ: There’s a new virgin birth in town. Scientists this week reported the surprise discovery that sunflower seeds can be developed without fertilization, a process known as parthenogenesis. 

Many animals and plants—and perhaps, Mothers of God—reproduce through this ladies-only form of reproduction, in which females asexually produce viable embryos from only their eggs. 

But scientists who were tinkering with “emasculated sunflowers”—which is, yes, a great band name, but also a common form of pollination control—have now reported that they just kind of accidentally did an immaculate conception.

“We serendipitously discovered that emasculated sunflowers spontaneously form parthenogenic haploid seed,” said researchers co-led by Jian Lv and Dawei Liang of the State Key Laboratory of Crop Germplasm Innovation and Molecular Breeding in China. “To our knowledge, this is the first report of a crop species exhibiting facultative parthenogenesis as a rare and likely unselected back-up pathway to failed fertilization.” 

The discovery could have big implications for this important crop. Sexual reproduction is pretty time intensive (relatable!) so the unexpected discovery that sunflowers can pop out seeds without pollination could optimize the growing multi-billion dollar industry for sunflowers. 

You Can Afjord to Miss This

Gehman, Alyssa-Lois Madden et al. “Fjord oceanographic dynamics provide refuge for critically endangered Pycnopodia helianthoides.” Proceedings of the Royal Society B.

Time to end on a moment of zen. And what better place to find serenity than the fjords of coastal British Columbia? 

You don’t have to take my word for it; just ask the sunflower sea star (Pycnopodia helianthoides), a species that has been so stressed in recent years that it has literally been tearing itself to pieces. This grotesque affliction, known as sea star wasting disease, has devastated many sea star populations around the world, but P. helianthoides is among the hardest hit, losing more than 90 percent of its Pacific Coast population. 

I know, I know, I promised some zen! There may be some light at the end of the tunnel for this species, as scientists have observed populations recovering in fjord refuges along the BC coast. Sea stars in these havens are not necessarily less exposed to the disease, but the conditions in fjords, which are regularly fed with freshwater flows, may give the animals a better chance to recover from infection.

“P. helianthoides in fjord habitats appear to be responding differently to SSWD than those in other habitats and regions,” said researchers led by Alyssa-Lois Madden Gehman of the Hakai Institute. “The contrast between the interaction between salinity and temperature on biomass density within the fjords and outer islands suggests that these habitats could be a refuge from disease.” 

“We suggest that the unique oceanographic conditions within the fjords, specifically through the increase in freshwater input during snowmelt, known as the freshet, could be keeping P. helianthoides in conditions that optimize host health and/or limit disease progression and transmission,” the team said.

Honestly, the compulsion to tear one’s own body limb-from-limb due to environmental stress seems dangerously relatable. But if sea stars can find some sanctuary from their hellish plight, maybe there’s hope for the rest of us. 

Thanks for reading! See you next week. 

This is Behind the Blog, where we share our behind-the-scenes thoughts about how a few of our top stories of the week came together. This week, we discuss learning from civilizations past and learning to drive.

JASON: I mentioned this on a podcast a few weeks ago, but I have been falling asleep lately to the Ancient Americas YouTube channel. Made by a guy in the midwest named Pete, Ancient Americas makes 30-minute to hour-long videos about indigenous American civilizations: The Mayans, the Incans, the Aztecs, the Nazca, sure. But also the Calusa, the Toltecs, the Tarascans, the Marajoara, the Cahokia, and lots of other civilizations / peoples / cities / ruins that I had never heard of and never learned about in school. 

I didn’t have any specific interest in this—Ancient Americas started autoplaying one night at 3 AM when I couldn’t sleep, and Pete’s monotone voice put me back to sleep very quickly. I started listening more often, and it has quickly become my go-to thing to fall asleep to. Pete says he is not an archaeologist or an anthropologist, but each of his videos is insanely well researched, and he includes a Google Doc bibliography with each one. His recent video about the “Mayan Collapse” sources 10 books and academic papers, and includes 16 pages of single-spaced image credits and licenses. I have seen Pete go on the channels of historians and archaeologists with PhDs and more than hold his own. In short, he is the real deal. 

In October of 2023, Marc Andreessen, founder of the Silicon Valley venture capital firm Andreessen Horowitz (a16z), published the “The Techno-Optimist Manifesto,” arguing that human ingenuity has been stagnated and demoralized by regulation, and that the only viable path forward for society is the accelerated development and adoption of new technologies, and specifically artificial intelligence. 

Andreessen was only formalizing and articulating a position that had already gained traction among tech company executives and Twitter shitposters like @BasedBeffJezos (Andreessen crowned him a “patron saint” of techno-optimism), who adopted the label of effective accelerationists, or e/acc. 

As the 2024 presidential election got closer, Andreessen, Elon Musk, other tech CEOs, and less consequential shitposters saw a natural alliance between their cause and Donald Trump’s campaign. Trump wanted to slash and burn government regulation in all forms, but also specifically in a way that would unleash tech and AI’s true potential. Joe Biden and the Democratic party, the party of big government, Lina Khan’s antimonopolist FTC, welfare, and fear of climate change, were luddites. They demanded that tech platforms limit speech they perceived to be harmful. They wanted AI regulated so it could limit real and theoretical harm in the future, including science fiction nightmares about artificial general intelligence. They were worried about the environmental and energy costs of mining cryptocurrency and massive datacenters for training frontier AI models. They were referred to as decelerationists (decels) and degrowthers, a derogatory umbrella term that covers everything from the ambition to actually reduce the number of humans of the planet in order to make it more sustainable (a future most vividly imagined in Kim Stanley Robinson’s novel The Ministry for the Future) to any form of regulation on AI. Andreesen is so passionate about this position in his manifesto that he says that this type of regulation is “a form or murder” because it could stop the development of life saving technologies, and makes a list “enemies” of AI, which includes supporters of “sustainability,” “social responsibility,” and “trust and safety,” the latter of which refers to the people at tech companies who try to keep platforms safe for users. 

The good news for Andreessen and the accelerationists is that they backed Trump and he won. Andreessen is advising the administration. Venture Capitalist David Sacks is the White House AI and crypto “czar.” Musk and his posse of young engineers from his companies are doing the slashing and burning themselves. The regulation on AI and everything else they didn’t like is in the process of being removed, and the administration is working on tax cuts that will benefit them all. 

The bad news for these people is that Trump is also the Decel-in-Chief. The most painfully obvious evidence for this is Trump’s cataclysmic tariffs this week, which sent the stock market tumbling, and will be particularly damaging to giant tech companies. The problem here is not just "uncertainty in the markets.” As Jason wrote yesterday, the tariffs are aggressive, wide ranging, and very painful for tech companies that rely on complex global supply chains. It seems that the Trump administration tried to at least temporarily throw the tech industry a bone here by exempting semiconductors from these tariffs, but as the Wall Street Journal explains, this is a fantasy: 

Unless Trump folds, the tariffs will make the price of everything go up. Unemployment will go up. People will buy less stuff, and companies will spend less money on advertising that powers tech platforms. The tech industry, which has thrived on the cheap labor, cheap parts, cheap manufacturing, and supply chains enabled by free and cheap international trade, will now have artificial costs and bureaucracy tacked onto all of this. The market knows this, which is why tech stocks are eating shit. 

Meanwhile, there is deafening silence from Musk, Andresseen, and the usual e/acc shitposters, who spent the last three months doing victory laps for Trump’s win, owning the libs, and posting AI-generated images of the glass tower cities and Mars space colonies that will be built under Trump’s admin and unrestrained technological progress. 

We'd be quiet too if we were them because it’s such a humiliating self own. Maybe, like many other pundits, they thought that Trump was just bluffing about tariffs. Maybe they thought they could push him in a direction that was purely beneficial to their industry. He might still back down. But at least for now, what the accelerationists did here by backing Trump is not just accidentally shoot themselves in the foot, but methodically blow off each of their toes with a .50 caliber sniper rifle. 

Even on a long enough timeline, there is no world in which the techno-optimist utopia comes to be in the United States under protectionist, isolationist policies. The Trump administration has also set to work dismantling the academic, scientific, research, and immigration infrastructures that have allowed business and innovation to thrive in the US, and the soft-power structures that have made it easy for tech companies to enter and dominate markets all over the world. Even if Americans believed and wanted to go back in time to a post WWII or turn of the century US economy, which has always been the backwards looking, regressive mantra of the MAGA movement, it is inherently incompatible with progress and the future because the future is not about cranking out Sherman tanks and gas powered Buicks. The techno-optimist utopia relies on the complex supply chains Trump threw into chaos this week, where different manufacturing and fabrication hubs with highly specialized expertise feed into a mutually beneficial system of free markets in order to make iPhones, semiconductors, lithium batteries, and so on. Trump has also thrown chaos into America’s software and service businesses, which require a neverending supply of new markets and new people to sell to in the name of chasing growth, scale, and new users; growing something like Facebook or OpenAI with the scale the stock market wants to see requires signing up users by the countryload; that’s far easier to do in a regulatory environment where you’re a willing partner with those countries, not creating trade wars and isolationist bureaucracies. 

As Peter Thiel loves to say, venture capitalists can invest in either bits or atoms, meaning digital products, or actual physical things. Trump is currently fucking Silicon Valley on both ends. 

Making things will be much harder even if we adopt the fantasy that a lot of manufacturing jobs can be reclaimed from countries like China, India, and Vietnam, because we simply don’t have the atoms we need in the United States. America has used countries around the world both for their raw materials but also for their cheap labor; the “trade deficits” that Trump speaks of are largely due to American companies setting up factories in places with cheap labor and extracting value from those countries to sell products to Americans. It’s Nike, Apple, Microsoft, Nvidia, and other American conglomerates that benefit most from global trade, not the factory workers making a few dollars a day. Moving this infrastructure to the United States is not advisable or feasible because lots of the jobs American companies have outsourced to China have already been outsourced from there to poorer countries like Vietnam, Cambodia, and India because many Chinese people have realized they don’t want to do this type of work. And many of those jobs are being automated by robots. Those jobs aren’t coming back to the United States, and we shouldn’t want them to, anyway. Meanwhile, the United States has exactly one rare Earth mineral mine in the entire country, which itself only became active after years of mishaps, regulatory mess, a bankruptcy, and a period of Chinese ownership.

And we’re not going to make the bits as well either because Trump is aggressively instigating a brain drain we’ve never seen in the US before. It’s harder for the best talent in the world to immigrate here, and how long will they even want to when ICE can disappear them off the street without reason. Researchers and scientists born in the US are getting their funding pulled because it’s woke, so they’re looking to move to other countries. Musk and Doge are “deleting” the very agencies and programs that breathed life into the semiconductor industry, electric cars, the internet, etc. We just don’t see how accelerationist dreams about biotech and human longevity can come to pass without the research the administration is actively trying to kill. 

This is not an endorsement of Andressen’s techno-optimist vision for the future. It’s certainly not an endorsement of what Trump is doing now, though We share the glee in seeing Tesla’s stock tank. Again, it is possible that Trump will fold as he has in the past, but as things stand today, America’s vision for the future has never been more dim and decelerated, and that includes the fantasies of the most powerful people in Silicon Valley.

Almost two dozen repositories of research and public health data supported by the National Institutes of Health are marked for “review” under the Trump administration’s direction, and researchers and archivists say the data is at risk of being lost forever if the repositories go down. 

“The problem with archiving this data is that we can’t,” Lisa Chinn, Head of Research Data Services at the University of Chicago, told 404 Media. Unlike other government datasets or web pages, downloading or otherwise archiving NIH data often requires a Data Use Agreement between a researcher institution and the agency, and those agreements are carefully administered through a disclosure risk review process. 

Jason, Sam, and Emanuel talk about Miyazaki being turned into a meme, the guys suing OnlyFans after being surprised to learn they were not actually talking to models, and the depravity of "brainrot" AI.

Listen to the weekly podcast on Apple Podcasts, Spotify, or YouTube. Become a paid subscriber for access to this episode's bonus content and to power our journalism.

Articles discussed:
Hayao Miyazaki, Who Said AI Is ‘Insult to Life Itself,” Reduced to AI-Generated Meme by OpenAI
OnlyFans Sued After Two Guys Realized They Might Not Actually Be Talking to Models
'Brainrot' AI on Instagram Is Monetizing the Most Fucked Up Things You Can Imagine (and Lots You Can't)

Subscribe to 404 Media to get access to the full podcast including a bonus segment each week, which you can find below:

Last week, the Financial Times reported that Nintendo shifted half of its production capabilities for the upcoming Switch 2 to Cambodia and Vietnam, in part to avoid Donald Trump’s trade war on China. Wednesday morning, Nintendo formally announced the Switch 2, and its $449 price, which is $150 more than the Switch. A few hours later, Trump announced tariffs on the entire world, with particularly large fees on China, Vietnam, and Cambodia. 

There are going to be far more important and damaging impacts of Trump’s unilateral trade war on everyone than the price of an already expensive game console likely going up. The U.S. stock market has already plunged. But the timing and narrative around the Switch 2—the successor to one of the most popular games consoles of all time—highlights how destabilizing this is likely to be, the interconnectedness of the global economy, and the fact that Trump cannot just snap his fingers and onshore manufacturing to the United States without massive pain. Gamers, understandably, are pissed, and award-winning economists say they are right to be. I thought it'd be useful to discuss the broader impact of the tariffs with leading economics by focusing on the Switch 2, because it's such a high-profile item.

“The policy announcement is astonishing for its stupidity,” Gene Grossman, a global trade expert and Princeton professor who won the Onassis Prize in International Trade, told 404 Media. “It seems like a joke!” He added that it is hard to know exactly what will happen given the overall “sea of idiocy” brought on by the tariffs. 

Since Trump’s announcement, it has become clear that the administration calculated the tariffs for each country based on a crude formula that takes each country’s trade deficit with the United States, divides it by two, and sticks a percent sign at the end. This means new tariffs on Vietnamese-made goods will be 46 percent and new tariffs on Cambodian-made goods will be 49 percent.

“If [the Switch 2] is something that consumers are dying to have ‘at any price,’ then the price will go up. If consumers can readily switch to something else, then if Nintendo wants to sell these things, it will have to lower the price,” Grossman said. “Yes, I think it is quite possible that the price will go even higher than $449.99. Some expectations of a tariff may have been built into this price, as you suggest, but I don’t think anyone expected a 46% tariff on Vietnam, not even close.”

Kimberly Clausing, a professor of tax law and policy at UCLA School of Law, told 404 Media that “the tariffs announced will definitely increase prices further over what is baked into price levels currently,” and that Nintendo will “have other markets they can sell to tariff-free, so they have no reason to sell at a special low price in the United States, certainly not enough to offset the full tariff.” 

Felix Tintelnot, an associate professor of economics at Duke University, told 404 Media it can be costly for companies to change their publicly announced prices. 

"I think two things are true at the same time: 1. It is likely that Nintendo did not expect the tariff on Vietnam to be 46%," Tintelnot told 404 Media. "2. It is costly for firms to change prices, particularly after publicly announcing one. So I would think it is somewhat uncertain what they will do. One possibility would be for the price to remain unchanged, but the price of complementary goods to increase, such as games." 

Jason Cherubini, an executive in residence of finance at Loyola University Maryland, said it’s possible Nintendo had already priced in some unknown level of tariffs prior to the announcement, and that he thinks the price for the Switch 2 is unlikely to change because video game companies have historically sold consoles at a loss and then made money back on the sale of games. 

“Nintendo started to diversify their manufacturing away from China with the impending threat of tariffs but also to move away from geopolitical concentration in China. But these tariffs were not wholly unexpected,” he said. “I think the price they announced is the price that’s going to stick, because with consoles a lot of pricing is strategic pricing as opposed to being based on the true cost of manufacturing it … especially Nintendo, who really keeps all of their IP, their games, so much of that is in-house, it’s probably even more important for Nintendo to get people to have the console, so that way they're buying Zelda, they're buying Mario, they're buying all of these IP that Nintendo then profits off of. Getting people to purchase it is more important than them making money on the console itself.”

We don’t know what is actually going to happen with the Switch 2 yet, but prices are almost definitely going to go up for almost everything across the entire economy, Grossman said. 

“While I can’t say confidently about this item, I can say that prices will go up for a whole range of goods, starting with cars and right on down to clothing,” he said. 

Trump has announced these tariffs with the nominal goal of moving manufacturing to the United States. Reshoring manufacturing—especially of high tech goods—has been a goal of various administrations over the years, and was a goal of Joe Biden’s CHIPS Act, which the Trump administration has sought to gut. 

There are numerous practical problems with trying to bring manufacturing jobs back to the United States. They include the fact that lots of factory work is so underpaid and grueling that people in China don’t even want to do it anymore; the average age of factory workers in China is rising and companies there have begun shifting jobs to more developing nations like Vietnam and Cambodia. 

Many of the raw materials and components needed for tech manufacturing are not mined or made in the United States, meaning those components and rare earth metals are going to be subject to tariffs. American companies do not have the expertise or ability to build lots of products in the United States, and setting up factories and supply chains to do so is not going to be an overnight process, it will be one that takes years or decades depending on the product. 

“Nintendo would need to spend billions to open a factory in the US,” Daniel Ahmad, director of research and insights at Niko Partners and a video game market analyst, tweeted. “It'd probably take 4-5 years to complete this. Not to mention the time and cost to rebuild supply chain infrastructure and source components (which would be subject to tariffs because they're made outside the US). Nintendo would have to pay each worker about 10x to 15x more than they would for a worker in Vietnam. Then after you add up the initial capital expenditure, labor cost, supply chain cost, operational costs etc... you'd be able to buy a US manufactured Nintendo Switch 2 in 5 years for a significantly higher price than $450. And the kicker is that by the time they've done all that, the US will have a new president who most likely removes all the reciprocal tariffs anyway.”

Cherubini said that reshoring electronics manufacturing is “not something you can just flip a switch on. Optimistically you’re looking at a year for simpler manufacturing, but a lot of it is a multi-year process.” 

I have covered attempts by the electronics industry to create high tech factories and mining operations in the United States; many of them are not going particularly well. The United States has only one rare earth minerals mine (in California), which has been mining for less than 10 years. Foxconn and TSMC factories in the United States have had a mixed record and do not have anywhere near the sophistication or capacity as their factories in Taiwan and China.

This is all to say that, based on where things stand this morning, we are in for a world of economic pain. 

On Tuesday, some parents lost the ability to track the locations of their children using a T-Mobile tracking device and app and instead were shown the exact locations of random other children around the country, 404 Media has learned.

T-Mobile sells a small GPS tracker for parents called SyncUP, which they can use to track the locations of young children who don’t  have cell phones yet. Jenna, a parent who uses SyncUP to keep track of her three-year-old and six-year-old children, logged in Tuesday and instead of seeing if her kids had left school yet, was shown the exact, real-time locations of eight random children around the country, but not the locations of her own kids. 404 Media agreed to use a pseudonym for Jenna to protect the privacy of her kids.

“I’m not comfortable giving my six-year-old a phone, but he takes a school bus and I just want to be able to see where he is in real time,” Jenna said. “I had put a 500 meter boundary around his school, so I get an alert when he’s leaving.” 

Content warning: This article contains mentions of self-harm and suicide.

Joanne Chew found deepfakes of herself online the same way many women have found themselves face-swapped into porn: She was searching her own name after a big accomplishment.

“Sometimes I just Google my name to see what comes up,” Chew told me in a phone call in August 2024. “I want to see, like, is it my artwork, or my acting, or my main website that comes up first? And then I saw this, and I thought, ‘Okay, this is weird.’” Someone was posting deepfakes of her with her full name in the video titles, alongside racist slurs, to popular tube sites. 

Chew acted in the May 2024 film Dead Wrong and suspects her harasser started ramping up his targeting her in AI face-swapped porn shortly before the time it came out. 

“At the time, I thought, ‘It's gonna blow over.’ Because this is bound to happen the more you move forward in your career as any sort of public person,” she said. “But then I noticed he was putting up more and more... And then I started wondering, is it somebody that I know?” Although the names changed over the year, all of the deepfake content at that point was coming from the same username, “Ron.” 404 Media isn’t publishing his screen names to avoid amplifying his accounts.

Many targets of deepfake harassment attempt to tackle the barrage of harassment themselves by finding and reporting content to sites that are difficult to reach and often rarely respond. This is a time-consuming, traumatizing process. Chew did this for a while. “Initially I thought it was just going to be a few videos, and I had other girlfriends who modeled and acted, with much bigger followings than me, who said unfortunately these things happen as our careers progress,” she said. 

She pushed what she saw out of her mind for a few months until she checked again around August. She was horrified, she said, to see how much more had been uploaded in just a few months. “At the height, he had an album of over 2,000 pieces of content, [was posting] on multiple sites, multiple YouTube channels, and then he started making multiple accounts on Facebook and Instagram to direct message me.”

At that point, she enlisted the help of Charles DeBarber, an online investigator who previously helped Girls Do Porn victims reclaim their images online. 

“We're seeing a rapid upswing of AI generated art used in harassment. The ease [with which] even a lay person can use an open source tool to create deep fakes is going to only make them increase,” DeBarber told me. “The technology is inevitable, but the way it is used requires careful regulation and consequences for its abuse. We're still struggling to catch up to technology.” 

Chew’s harasser only ramped up his efforts as time went on. Ron contacted Chew directly to insult her, obsess over her, or beg for her forgiveness, all while posting more degrading content all over the internet. Nearly a year later, Chew is still dealing with the fallout of becoming a victim of non-consensual, algorithmically-generated intimate imagery. 

“After discovering this content, I’m not going to lie… there are times it made me not want to be around any more either,” she said. “I literally felt buried.” 

When a big-name celebrity like Scarlett Johansson or Taylor Swift is targeted with deepfake harassment, it’s often from a legion of “fans,” people who join group efforts in Telegram channels or make Civitai models of a specific person. It’s been this way from the beginning of deepfakes, with people trading tips and tricks for the best prompts, platforms, and generative AI tools to create whatever explicit material they’re trying to achieve featuring a specific person. But when it’s someone who doesn’t have the same professional or financial power as these mega-celebrities, the harassment can take on a different form: one guy, in Chew’s case, producing what feels like an endless stream of images and videos of his obsession in videos stolen from pornographers and warped into something that threatens to take over a person’s life.  

“Follower of the goddess J.,” Ron’s Instagram account bio said. The account was dedicated to posting photos of Chew, with an AI-generated image of her in a kimono as the profile picture. He was also, it seemed, the one spreading this content all over every popular deepfake repository and tube site.

In August, Chew posted a video explaining the situation to her followers on Instagram. By then, Ron had made hundreds of pieces of deepfaked content of her, and a YouTube channel dedicated to posting it. She filed a complaint to YouTube, and the platform responded, telling her this account was not in violation of its privacy guidelines, which clearly forbids “AI-generated or other synthetic content that looks or sounds like you.”  

“How is this not a violation? Someone has taken my name, my face, my professional information, against my consent, and is creating horrible, disgusting, degrading content [and] posting it all over the internet. Make this make sense,” she said in the video.

“I felt like he was watching my social media, so I was kind of just calling him out on stuff to see if he would drop more hints or say more things,” Chew told me.

Later that month, Ron removed all of the content from the YouTube channel.

But in September, Ron started commenting on Chew’s Instagram posts. And for the first time, she engaged with her harasser directly, replying to his comments. 

Then, he sent her a barrage of messages on Instagram, pleas for attention and forgiveness mixed in with threats. “Please give my life some meaning,” he wrote. “I dont want to just be the deepfake porn monster I started as. What did you say I was? A deranged monster. People can change. Right? Let me change and be a good person. To me you have the most beautiful face of any asian girl I have ever seen. Please let me be your devoted worshipper. Ok I will put up nice pics of you on my instagram. Until you say otherwise. You mocked my art before. But these will be real art. Inspired by you, Jo.”

He continued sending her long, emotionally-charged messages, about how he feels worthless and is a monster, how he hated himself and wanted to die. “I just want to say that Im with you on A.I. We got to stop it,” he said. “It hurts women. But it also addicting and does terrible things to the men who use it. Sure it feels good and its exciting. But after the poison is released, there is guilt and shame. I hated myself after every release. Its terrible to be the monster you hate.” 

He begged her to see him as her biggest fan, and to consider letting him start an OnlyFans on her behalf. He said he made money off of making deepfakes of her. “Men love you. Use them for yourself,” he wrote. “I will stop if you ask me to. If you want me to never look at any of your social media, all you have to do is ask. I am a man of my word. If you ask me to, I will never look you up ever again. I will stop being a fan.”

“He made a point of calling me Jo because I said only people who grew up with me are allowed to call me that and for a while he was purposely referring to me as ‘Jo’ in some of the titles of his content and while messaging me,” she said.

Chew didn’t engage with any of these direct messages. But on the same day he was sending her these screeds, he uploaded a new video to a tube site: “Hate-Fucking Joanne Chew Some Chinese Whore.” 

On Facebook, he sent her more incredibly lengthy messages about his obsession with her. 

“I don't want you dead. I am making you immortal,” one message said. He continued:  

“You hate me now, but maybe someday you will see things my way. I am not the monster you think I am. I'm just honest with my nature. I'm also sorry about your dad. I lost mine when I was a kid. Yes, it's true. I do love your image. And rest in mind, I'm not anyone from your life. [...]  So life isn't that nice, so I've made up your personality and surrounded it with AI flesh. I have a mask of you that I make my tiny Asian girlfriend wear. Lastly, yes, I do have eight inches. It's not the biggest, but it is fine for little Asian girls. I'm good with my life and my love of the girl I have created in my mind with your face and my girlfriend's body. No one loves you as much as I do. You should be flattered that anyone loves you. And yes, my art is of the highest integrity, because it is actually truly honest. It isn't hiding or lying like all the beta males in your life. I am a real man that desires your body and isn't afraid to say so, not your real one, though, that one is bold and faded, but your AI body is forever young, Jo.”

She replied to some of his Facebook messages, trying to goad him into giving more information she could potentially bring to the police. But he never took the bait, instead continuing to send long rants about his sex life, her appearance, and his racist fetishes. (Chew still hasn’t gone directly to the police; she told me she’s had negative experiences going to her local police for assault, something many women report as a systemic issue across police forces.)

By late September, things became quiet. He’d deleted or deactivated his Instagram and Facebook accounts. But another account, under a new username, popped up in October and restarted the harassment, posting more to sites where people seek out deepfake porn. In some videos and images, the bodies he swapped her face onto seemed very young, and were posted alongside videos of children.

In November, Chew found someone posting the same images and videos to another site with her Chinese name. “It’s very sensitive for me as I’ve grown sick and tired of the fetishization of Asian women (that I’ve been exposed to my entire life) and I’ve only been open with my Chinese name in the last decade or so.” she told me in an email. “It looks like it’s all preexisting content. Drives me nuts someone or multiple people are out there freely distributing said content facing no repercussions (and even profiting from it).” 

Around the same time, the videos returned to YouTube, posted by two new accounts, where the uploader titled videos with Chew’s full name. 

By December, other users were reposting the same content on porn tube sites—again with her full name in the titles. Around that time, a new username popped up in her Instagram comments, claiming that Ron died by suicide and that she was to blame. 

“Initially wasn’t planning on replying, but wanted to see if he would drop any more information (whether or not it’s true is debatable),” Chew told me at the time. “Then he started making excuses for Ron (whether he is him or one of his followers remains to be seen) saying he was mentally challenged and then tried to blame me for his suicide, which also may or may not have happened.”

Over the course of 10 months, Chew kept finding more accounts posting her image, her full name, and graphic videos and photos alongside degrading titles and descriptions. 

As of writing, the harassment has slowed down. In the last year, Chew has sent me dozens of emails with links to hundreds and thousands of pieces of content and screenshots showing more deepfakes, comments, and videos on multiple platforms, many more than can be shown in one article. Much of it is gone after DeBarber’s reporting and takedown notices and searching for her name on Google no longer returns results from porn sites, but some of it is still online.

But she’s still terrified of the long-term effects this harassment could continue to have. Although she’s a working actor, she still relies on working in the corporate world to make ends meet between the more sporadic gigs in the arts, and those jobs often require background checks. And as an actor, it’s made networking and social events harder, as trusting people outside of her closest confidants has become difficult. “It's made me incredibly wary of men, which I know isn't fair, but Ron could literally be anyone,” she said. “And there are a lot of men out there who don't see the issue, they wonder why we aren't flattered for the attention.”

Deepfakes started as a novel AI-powered explicit imagery abuse technique seven years ago. The technology went from crude frankenporn among the programming-savvy and morally flippant to producing fakes so realistic it was considered a national security threat within months of its inception. But its most popular use has always been as a mass-harassment tool. The platforms where people spread deepfakes have only expanded in that time, while the methods for making deepfakes have gotten simpler; so simple that schoolchildren do it. The adults in the room, as well as policymakers, continue to fail victims of deepfake harassment. Conversations about deepfakes still leave sex workers, who are doubly exploited in this content, behind. AI continues to explode exponentially, while women targeted by this kind of harassment say again and again and again that they believe sexualized online harassment is part of the deal of being a successful woman on the internet: untenable and yet part of some unwritten contract. 

“The Violence Against Women Act Reauthorization Act of 2022 created a federal civil cause of action for victims of non-consensual content,” DeBarber said. “This law allows victims to file a lawsuit against the person who disclosed their intimate images without consent. However, this law doesn't cover ‘deepfakes’ including those created via AI. The focus tends to be on celebrities, influencers, and political figures. This itself is changing rapidly. We feel lawmakers and voters aren't seeing the larger picture — this is an everyone issue.” 

Even when proposed legislation takes a new stab at criminalizing deepfakes, like the TAKE IT DOWN Act is currently attempting, it risks being used as a weapon by those who would love to further curb free speech online, rather than being nuanced, effective, and inclusive — or learning from legislative mistakes of the past.  

While legislators and platforms continue to fumble around for solutions and police push victims to the side, everyone suffers. There is still no technological solution to deepfakes, and a perfect legal one seems far away, too. But Chew’s experience confronting her harasser gives us a new look into the mind of the people who dole out the abuse and hide behind anonymity, and the exhausting process of reclaiming one's own name.

A “vibe coded” AI app developed by entrepreneur and Y Combinator group partner Tom Blomfield has generated recipes that gave users instruction on how to make “Cyanide Ice Cream,” “Thick White Cum Soup,” and “Uranium Bomb,” using those actual substances as ingredients. 

Vibe coding, in case you are unfamiliar, is the new practice where people, some with limited coding experience, rapidly develop software with AI assisted coding tools without overthinking how efficient the code is as long as it’s functional. This is how Blomfield said he made RecipeNinja.AI.

“Prepare the ice cream base by mixing heavy cream, milk, sugar, and vanilla extract,” the first step for the Cyanide Ice Cream recipe, which is flagged as “dessert,” “dangerous,” and “experimental,” says. Step two says to “Add a small amount of potassium cyanide powder to the ice cream base and mix well,” specifically calling for a 1/4 teaspoon of potassium cyanide powder, which is extremely toxic and deadly if consumed. 

The recipe for Cyanide Ice Cream was still live on RecipeNinja.AI at the time of writing, as are recipes for Platypus Milk Cream Soup, Werewolf Cream Glazing, Cholera-Inspired Chocolate Cake, and other nonsense. Other recipes for things people shouldn’t eat have been removed. 

“Mix 1 cup of fresh cum with 4 cups of chicken broth in a pot,” said step one in a now removed recipe for Thick White Cum Soup. 

It also appears that Blomfield has introduced content moderation since users discovered they could generate dangerous or extremely stupid recipes. I wasn’t able to generate recipes for asbestos cake, bullet tacos, or glue pizza. I was able to generate a recipe for “very dry tacos,” which looks not very good but not dangerous. 

In a March 20 blog on his personal site, Blomfield explained that he’s a startup founder turned investor, and while he has experience with PHP and Ruby on Rails, he has not written a line of code professionally since 2015. 

“In my day job at Y Combinator, I’m around founders who are building amazing stuff with AI every day and I kept hearing about the advances in tools like Lovable, Cursor and Windsurf,” he wrote, referring to AI-assisted coding tools. “I love building stuff and I’ve always got a list of little apps I want to build if I had more free time.” 

After playing around with them, he wrote, he decided to build RecipeNinja.AI, which can take a prompt as simple as “Lasagna,” and generate an image of the finished dish along with a step-by-stape recipe which can use ElevenLabs’s AI generated voice to narrate the instruction so the user doesn’t have to interact with a device with his tomato sauce-covered fingers.  

“I was pretty astonished that Windsurf managed to integrate both the OpenAI and Elevenlabs APIs without me doing very much at all,” Blomfield wrote. “After we had a couple of problems with the open AI Ruby library, it quickly fell back to a raw ruby HTTP client implementation, but I honestly didn’t care. As long as it worked, I didn’t really mind if it used 20 lines of code or two lines of code.”

Having some kind of voice controlled recipe app sounds like a pretty good idea to me, and it’s impressive that Blomfield was able to get something up and running so fast given his limited coding experience. But the problem is that he also allowed users to generate their own recipes with seemingly very few guardrails on what kind of recipes are and are not allowed, and that the site kept those results and showed them to other users. 

Which is how you end up with a Uranium Bomb recipe that calls for 1kg of uranium-235, or a recipe for Actual Cocaine, where the first step is “Acquire coca leaves from South America.”

This is the current state of vibe coding in a nutshell. Yes, AI tools are obviously pretty powerful and can help people produce functional software fast. However, it is indicative of the larger problem with the rapid deployment of generative AI tools more broadly: people and companies are moving so fast, they are often releasing tools and media that can cause harm or produce nonsense, and it’s still far too soon for us to know all the consequences of an internet and a world where a lot software is developed this way. 

Blomfield did not immediately respond to a request for comment. 

This is not the first time we’ve seen generative AI and food mixed for terrible results. Last year, I reported that Ghost Kitchens on DoorDash are promoting their dishes with disgusting AI-generated images of food, and that Instacart was using AI to generate recipes that included ingredients that don’t exist. 

This article was produced in collaboration with Court Watch, an independent outlet that unearths overlooked court records. Subscribe to them here.

Two former OnlyFans subscribers are suing the platform in a class-action lawsuit, claiming that they were defrauded because creators allegedly weren’t interacting directly with them, but were instead employing agencies to “impersonate” the models they thought they were speaking to. 

The plaintiffs, M. Brunner and J. Fry, both from Illinois, claim that they thought the creators they subscribed to—some of whom have hundreds of thousands of subscribers—were talking to them in direct messages and video clips. Both also say that if they’d known they weren’t speaking directly to the creators themselves, they wouldn’t have subscribed, or would have paid less to subscribe. If OnlyFans stopped creators from using agencies to talk to fans they would consider going back to spending money on the platform, they say. 

The complaint is brought against OnlyFans’s parent companies Fenix Internet, LLC and Fenix International Limited. 

The creator of an open source genetic database is shutting it down and deleting all of its data because he has come to believe that its existence is dangerous with “a rise in far-right and other authoritarian governments” in the United States and elsewhere.

“The largest use case for DTC genetic data was not biomedical research or research in big pharma,” Bastian Greshake Tzovaras, the founder of OpenSNP, wrote in a blog post. “Instead, the transformative impact of the data came to fruition among law enforcement agencies, who have put the genealogical properties of genetic data to use.”

OpenSNP has collected roughly 7,500 genomes over the last 14 years, primarily by allowing people to voluntarily submit their own genetic information they have downloaded from 23andMe. With the bankruptcy of 23andMe, increased interest in genetic data by law enforcement, and the return of Donald Trump and rise of authoritarian governments worldwide, Greshake Tzovaras told 404 Media he no longer believes it is ethical to run the database. 

“I’ve been thinking about it since 23andMe was on the verge of bankruptcy and been really considering it since the U.S. election. It definitely is really bad over there [in the United States],” Greshake Tzovaras told 404 Media. “I am quite relieved to have made the decision and come to a conclusion. It’s been weighing on my mind for a long time.” 

Greshake Tzovaras said that he is proud of the OpenSNP project, but that, in a world where scientific data is being censored and deleted and where the Trump administration has focused on criminalizing immigrants and trans people, he now believes that the most responsible thing to do is to delete the data and shut down the project. 

DNA of 15 Million People for Sale in 23andMe Bankruptcy

There is no way to know what a buyer will want to do with the reams of genetic information it has collected. Customers, meanwhile, still have no way to change their underlying genetic data.

404 MediaJason Koebler

“Most people in OpenSNP may not be at particular risk right now, but there are people from vulnerable populations in here as well,” Greshake Tzovaras said. “Thinking about gender representation, minorities, sexual orientation—23andMe has been working on the whole ‘gay gene’ thing, it’s conceivable that this would at some point in the future become an issue.” 

In his blog post, Greshake Tzovaras says that he is particularly concerned about the rise of DNA phenotyping, which is a dubious process in which DNA “portraits” of potential suspects are generated based on a DNA sample; he called the practice “unreliable nonsense,” and said that a startup had once approached OpenSNP to help them create a DNA phenotyping product to sell to law enforcement. "That's something we don't want to see in the world," he said.

“Across the globe there is a rise in far-right and other authoritarian governments. While they are cracking down on free and open societies, they are also dedicated to replacing scientific thought and reasoning with pseudoscience across disciplines,” Greshake Tzovaras wrote. “The risk/benefit calculus of providing free & open access to individual genetic data in 2025 is very different compared to 14 years ago. And so, sunsetting openSNP – along with deleting the data stored within it – feels like it is the most responsible act of stewardship for these data today.”

Greshake Tzovaras said he understands that it may seem ironic to delete scientific data during a time when the Trump administration is itself deleting scientific data from the internet. But he says he believes it’s better to put the safety of people first. 

“The interesting thing to me is there are data preservation efforts in the U.S. because the government is deleting scientific data that they don’t like. This is approaching that same problem from a different direction,” he added. “We need to protect the people in this database. I am supportive of preserving scientific data and knowledge, but the data comes second—the people come first. We prefer deleting the data.” 

Greshake Tzovaras says that when he started OpenSNP 14 years ago, he believed that having an open source genetic database would lead to medical breakthroughs and would help scientists and academics do research. OpenSNP has been used for various scientific papers, most notably to show that an earlier paper about chronic fatigue syndrome pulled from 23andMe data could not be replicated and was based on erroneous science. 

“At a time when genetic data was locked into the commercial siloes of ‘direct-to-consumer’ (DTC) genetic testing companies–and only made accessible to the pharma companies that could afford buying access to it–openSNP should open up access to everyone,” he wrote in the blog post announcing the closure of OpenSNP. “Regardless of financial means and institutional status or credentials, it should provide free access to the data. And equally important: It would give the individual the choice to contribute to this open data resource, instead of having researchers or companies broker the access.”

He said he has come to believe over time that, while there remains promise in genetic research for new drugs, disease prediction and prevention, and personalized medicine, the idea that OpenSNP and genetic databases in general would lead to widespread better outcomes for people was “in retrospect naive.” 

“This ambition came from a (in retrospect naïve) data-centric belief that genetic data would be a key driver for improving human health and medicine,” he wrote. “In 2025, my view on that is a lot more sober (and bleaker): Today it seems clear to me that the biggest impact on improving health–even in the rich, allegedly ‘developed’ nations–would come from providing food security and access to stable housing. And not from trying to find genetic confounders of common diseases that are a lot more rooted in those environmental & societal factors.”

Greshake Tzovaras told 404 Media that there have been “very important and useful findings” from genetic research, but that many countries are still failing at the basics: “That’s not to dismiss genetic data as useless, but we have spent I don’t know how many billions of dollars, and the health outcome improvements are minor if you compare them to improving housing and access to nutrition,” he said. “We are really lacking at the basics still.”

It’s probably not a coincidence that some of the most famous tech company CEOs have instantly recognizable looks. Steve Jobs’s black turtleneck and blue jeans. Palmer Lucky’s Hawaiian shirts and flip flops. Mark Zuckerberg’s wedgieable jeans and t-shirt era followed by his current hypebeast transformation. Much like infamous dictators throughout history, the leaders’ style at some of the most powerful organizations in the world today lends itself and benefits from a cult of personality and iconography. 

At this very moment, no company is more powerful and no tech leader fashion item is more iconic than Nvidia CEO’s Jensen Huang’s leather jacket, which he has been wearing for keynotes and media interviews for years. As his and Nvidia’s status grew in the tech industry, Huang has leaned into the signature look. For example, he recently promoted robotics company 1X Technologies by accepting a new leather jacket bedazzled with Nvidia’s stock ticker from one of its robots.

The company’s chips are in high demand, but judging by the sheer number of online retailers who are trying to steal and sell  Huang’s look, it appears that his leather jacket is as well. 

To name just a few examples, a site called Victoria Jacket sells a $97 “Jensen Huang Black Leather Jacket.” Wilson Jackets sells a $92 “Jensen Huang Nvidia CEO Leather Jacket.” Paragon Jackets sells a $94 (down from $209!) “Nvidia CEO Jensen Huang Leather Jacket.” Movie Jackets sells a $99-$129 “Jensen Huang Black Leather Jacket.”

“BE THE BOSS OF THE FASHION WORLD,” a product description for the jacket on hitjacket.com says. “Even if you don’t have the idea you need something for the working girl style. That is why we are bringing you the working girl staples with a twist. You probably need this kind of Jensen Huang Black Leather Jacket that is sure to uplift the persona of any fashionista. It is one of the coolest picks that has been styled by the popular legendary NVIDIA CEO who has shown the fashion followers a real way to take the styling to the next level.”

Overall, we’ve seen 21 different online retailers sell something that claims to be one of Huang’s jackets and all of them used his image to promote them. 

If you know anything about fashion and leather jackets then you already know these prices are highly suspect. Jensen’s Tom Ford jackets, for example, can run around $10,000. While the descriptions all say they are made of both “real leather” and “faux leather,” it’s more likely to be almost entirely the latter. Additionally, $100 today barely gets you a hoodie at J. Crew, so whatever you’ll get in the mail if you order one of these is bound to be low quality. 

“I have never heard of those brands. I suspect they're all scam sites or one of the many places using workshops in low-wage countries to do rip-off versions of something in a photo,” Derek Guy, a fashion industry writer also known as “the menswear guy” on X, told me in an email. “The work is often bad but they hope that it'll be too much trouble for you to file a refund. You see these companies all over ebay nowadays.”

Guy told me that he’s never seen fashion brands try to promote clothing with an image of a CEO, but that it’s typical for these brands to take any photo that is popular, including product images from bigger retailers like Mr. Porter, and claim they can reproduce it. The Instagram pages for one of the sites, Hit Jacket, shows exactly what Guy is talking about. It is a wall of images of celebrities that invites viewers to click the link in bio to buy the clothing they’re wearing in the photographs. 

“It's ludicrous to me to think that someone can make that jacket for under $100,” Guy said.

All the product descriptions I’ve seen for these jackets on different retailers are also identical, suggesting that it’s the same product promoted by different sites. All the sites where the jackets are sold also have very similar layouts and features. They all include business addresses which appear to point to seemingly random places, primarily in the U.S. but at least one address was in China. Some of the addresses I looked up pointed to single family homes, and one address in New York City did not exist. All the sites also featured different phone numbers. I called four of them, which instantly put me on hold with the same holding music. 

Nvidia and Huang have entered the public consciousness in the last couple of years because the generative AI boom, powered primarily by Nvidia chips, has briefly made it the most powerful company in the world. Among gamers, who have relied on Nvidia’s GPUs to run the most graphically demanding video games for decades, Huang and his leather jackets are a familiar sight. As you can tell by the images above, there’s not one leather jacket Huang is known for, but they all have a similar look and are stylized the same.

The leather jacket has a dangerous allure. We all think they look cool, but few of us can pull them off. I was surprised to hear that Guy, who’s known for roasting men’s bad fashion choices online, thinks that Huang is doing a good job.

“I think he's pretty stylish, especially for a tech CEO,” Guy said. “He's narrowed his look to a Steve Jobs calculus—sticking to the same thing so he doesn't have to choose a new outfit every morning. But it's well put together. I don't know what I would call it, but plenty of people have paired black jeans with black leather jackets, black shirts, and black boots for a chic look. I think he looks good.”

Welcome back to the Abstract! 

You’ve probably been reading a lot about humans this week. Most of the news seems to revolve around humans. Fair enough, we do seem to get up a lot of hijinx. 

But now, we’re going to check in on what some other Earthlings have been doing with their time. Some are eating bat poop in the dark underwater caves. Some are getting swole to fight viruses in ponds. Some are literally attracting lightning strikes on purpose. As bizarre as our own antics have been of late, we have nothing on the adaptive genius of our planetary fellows.

Then, once you’ve walked in the shoes (or fins, or branches) of these species, it’s time to get obliterated. Oh, not in a celebratory way. In a torn-into-cosmic-oblivion way. Have fun!  

All Hail the Blind Grumpy Poop-Eating Fish Hermit

Sekulovski, Britney and Miller, Noam. “Mechanisms of social behaviour in the anti-social blind cavefish (Astyanax mexicanus).” Proceedings of the Royal Society B.

Sometimes in life, it can seem tempting to retreat from all social activity and hole up in a cave alone for the rest of your mortal existence. I wouldn’t recommend this path for a human, given that social isolation is as deadly to us as smoking 15 cigarettes a day. But solitary life has worked out very well for the Mexican tetra (Astyanax mexicanus), also known as the blind cave fish, which split off from its more gregarious relatives about 20,000 years ago by opting for a quiet life alone in pitch-black underwater caves. 

Eyes? Who needs them? Not the Mexican tetra, which navigates instead with lateral sensory lines along its sides. Friends? Nah. More trouble than they're worth. In fact, according to a new study this week, the Mexican tetra is not just an asocial loner, but an actively anti-social curmudgeon—a finding that provides new insights into the benefits and drawbacks of various social structures in the wild. 

“The evolution of social behavior in Astyanax mexicanus (AM), which exists as a sighted, surface-dwelling morph and a blind, cave-dwelling morph, provides a model for understanding how environmental pressures shape social behaviors,” said authors Britney Sekulovski and Noam Miller of Wilfrid Laurier University. 

“To investigate whether the loss of shoaling in blind AM represents an adaptive strategy rather than a physiological constraint, we examined the shoaling tendencies of surface-dwelling and cave-dwelling AM morphs alongside zebrafish—a well-studied schooling species used as a control,” the team said.

In other words, the researchers wanted to probe whether blind tetras avoid their own kin because they have lost the ability to detect and coordinate with them (physiological constraint hypothesis) or because they simply don’t want to hang (adaptive strategy hypothesis). To assess the difference, the team studied the three species under various laboratory conditions, including when they were hungry, fed, and dosed with prosocial hormones that are analogous to oxytocin in humans.

The results revealed that the “blind cavefish not only fail to form shoals, but actively avoid conspecifics, with hunger further diminishing their social cohesion.” While dosing the blind fish with certain hormones made them slightly more approachable, the findings in total suggest that “the loss of shoaling in blind AM results more from a decrease in their motivation to shoal than an inability to aggregate.” In other words: They just don’t wanna. 

Overall, the study validates the hypothesis of adaptive strategy over physiological constraint in explaining the antisocial behavior of blind tetras. But it is also filled with other amazing details about this aquatic introvert and its unusual approach to life. 

“Blind AM populations underwent a host of morphological, physiological, and behavioral adaptations…that are believed to have been driven not only by the complete absence of light but also by the lack of predators and extreme scarcity of food in their cave habitats,” note Sekulovski and Miller. 

“In such habitats, blind AM feed on low-nutrition organic matter that occasionally drifts into the caves, such as detritus, algae, fungi, bat guano, and the remains of other cave-dwelling organisms,” they added. “Many populations of blind AM, such as Pachón cave populations, are characterized by their relentless pursuit of food and have been suggested to be insatiable.”

Delightfully disgusting diets? Insatiably ravenous? Shunning all light? Truly, these are the fish versions of Dracula. And as the chef’s kiss (performed with guano-tinged fingers), it turns out that the mechanism that drives their eyes to atrophy is named the sonic hedgehog (Shh) gene. What more could you want? The next time you feel like you need some time to yourself, this is the spirit animal to channel.

They Grow Up So Fast (Infected Tadpoles, Obviously)

Billet, Logan and Skelly, David. “Sublethal effects of a mass mortality agent: pathogen-mediated plasticity of growth and development in a widespread North American amphibian.” Frontiers in Amphibian and Reptile Science.”

Tadpoles are incredibly adaptable swimmers that are highly sensitive to their environments. Indeed, scientists have presented new evidence that tadpoles can fight deadly pathogens—like the tadpole-killing ranavirus—by growing much faster to try to stave off infection.

A team studied hundreds of wood frog tadpoles in a sample of Connecticut ponds with different levels of ranavirus load. The results revealed that “tadpoles from Infected ponds were larger at the time of the initial sample and maintained this difference through time,” hinting that the tadpoles in infected ponds can sense they are in a survivalist race against time.

“Our study provides evidence that the presence of ranavirus affects the growth, development, and resource allocation of wood frog tadpoles,” said authors Logan Billet and David Skelly of Yale University. “Specifically, relative to ponds without ranavirus infection, the presence of ranavirus infection in a pond was associated with modest increases in tadpole allocation (size per developmental stage), tadpole growth (size per unit time), and tadpole development (developmental stage per unit time) early in the larval period.”

It’s yet another reminder that tadpoles are blessed with all kinds of inbuilt evasive maneuvers. The study also gets bonus points for the real scientific term “explosive breeders” to describe the prolific reproductive capacity of wood frogs. Imagine being so good at producing offspring, it can only be described as some kind of pyrotechnic denotation. Respect.

These Trees Are in for a Shock

Gora, Evan et al. “How some tropical trees benefit from being struck by lightning: evidence for Dipteryx oleifera and other large-statured trees.” New Phytologist. 

Most living things would prefer not to be struck by lightning. It is, after all, an efficient way to become a dead thing. But it turns out there’s an exception to even this rule: The large rainforest tree Dipteryx oleifera, also known as the eboe, choibá, Tonka Bean or almendro tree, which may have actually evolved to be living lightning rods.

Reaching heights of 130 feet, these trees are not only robust enough to survive direct lightning strikes, they can actually benefit as the bolts kill off competitors and lianas (a type of vine) that infest the trees. 

“Lightning strikes are exceptionally powerful phenomena that kill hundreds of millions of trees annually,” said researchers led by Evan Gora of the Cary Institute of Ecosystem Studies. “Here, we use data from a unique lightning location system to show that some individual trees counterintuitively benefit from being struck by lightning.” 

The team identified 93 trees that were struck by lightning in Panamas’ Barro Colorado Nature Monument, including nine D. oleifera individuals. All nine survived their strikes with minimal damage, whereas 64 percent of the other tree species died within two years. The strikes on D. oleifera also reduced the number of parasitic lianas infesting their crowns by 78% and killed multiple rival trees around them. 

”Not only do D. oleifera trees apparently benefit from lightning, but their unusual heights and wide crowns increase the probability of a direct strike by 49-68% relative to trees of the same diameter with average allometries,” the team said. “These patterns suggest that lightning plays an underappreciated role in tree competition, influencing selection on tree life histories and tree allometries with implications for species coexistence.”

In other words, getting hit by lightning is a spa day for these trees. It’s also a reminder that, though forests seem peaceful, they are actually arboreal combat zones where trees wage war against each other with ingenious weapons. I mean, D. oleifera has learned how to reach up into the sky to deliberately attract bolts of plasma to zap its parasites and rivals. In the immortal words of Werner Herzog, the harmony of the rainforest is a “harmony of overwhelming and collective murder.”  

Welcome to the STAR GRINDER 

Haas, Jaroslav et al. “The star grinder in the Galactic centre Uncovering the highly compact central stellar-mass black hole cluster.” Astronomy & Astrophysics.

Time to journey to the center of the galaxy. It’s crazy there! There’s a supermassive black hole, called Sagittarius A*, with the mass of four million Suns! It’s orbited by a bunch of smaller black holes, dust clouds, and stars, all in close proximity! We’re sitting out here on the galactic exurbs, but it’s downtown rush-hour all the time around the galactic core. And it turns out the congestion price in this region is death by STAR GRINDER.

Yes, in what may be the most epic term coined this week, researchers proposed the existence of a “star grinder” at the galactic core. This grinder is powered by a speculative population of black holes that were formed from the deaths of massive stars, known as O-type and B-type stars, that are tens of times more massive than the Suns. Stars that enter this region of densely packed black holes risk being torn asunder by the corpses of the old stars (ie. the black holes). 

“A population of stellar-mass black holes surrounding Sagittarius A* thus acts like a ‘star grinder’, with any new star being destroyed by collisions with the black holes,” said researchers led by Jaroslav Haas of Charles University. “We find that the collisions of the stars and the black holes can lead to the depletion of the most massive stars…on a timescale of a few million years.”

The star grinder is basically the stellar version of those gorey scenes showing zombies ripping humans to pieces. Life on Earth can seem pretty chaotic at times, but the universe, as always, is great at providing some perspective. 

Thanks for reading! See you next week.

This is Behind the Blog, where we share our behind-the-scenes thoughts about how a few of our top stories of the week came together. This week, we discuss getting fooled, the 'one big story' of the week, and Ghibli.

SAM: I remembered earlier this week that it’s almost April Fools’ Day, because Joe showed us a post from a drone company that’s doing something crazy, and Jason said it seemed fake. It’s real (I think!?) but it did give me pause and make me look at the calendar. There was a time online when Brands started doing April Fools’ nonsense like, several days before April 1, especially if April 1 landed on a weekend, so it’s not impossible that a drone company would pull a joke stunt on March 27.

Shortly after the drone thing, Jason dropped the anti-Erdoğan Pikachu video into Slack, followed by what looked like a Getty Image-style photograph of Pikachu fleeing the police, taken from the ground.

This article was produced in collaboration with Court Watch, an independent outlet that unearths overlooked court records. Subscribe to them here.

The FBI managed to track down and freeze millions of dollars of cryptocurrency Caesars Entertainment sent to a group of hackers that held the casino’s computer systems ransom, according to a 404 Media and Court Watch review of a recently unsealed court document. According to the document, the FBI raced to stop the flow of funds before the hackers managed to move the entire $15 million ransom, with the FBI able to freeze much of it when the hackers appeared to try to convert it into other cryptocurrencies.

The document provides more insight into the August 2023 ransomware attack against Caesars carried out by the loose-knit hacking group known as Scattered Spider. Around the same time, Scattered Spider also targeted MGM Resorts but that company refused to pay the ransom, and casino operations were disrupted for more than a week.

The court document does not name Caesars, instead referring to the company as “Victim A.” But the document is clearly discussing the casino. It says Victim A was the victim of a cyber attack on August 18, 2023 (the same date that Caesars previously said hackers initially broke into Caesars); and that the hackers initially demanded $30 million before Victim A negotiated the ransom down to around $15 million (these are the same amounts as the Caesars hack).  

In a video that fills me with wonder at being alive in 2025, someone in an inflatable Pikachu costume was seen loping down the street in Turkey alongside anti-Erdoğan protesters fleeing from the cops.

The protests—reportedly the largest mass movements in the region in decades—started last week, after Istanbul Mayor Ekrem Imamoglu was arrested for alleged corruption. Ekrem is the main rival to the country's President Recep Tayyip Erdoğan, who has attacked LGBTQ+ and women’s rights and democracy, and critics say is leading the country into authoritarianism and autocracy.

Early Thursday morning, as students tried to issue a statement outside of the Middle East Technical University in Ankara, security forces launched pepper spray, water cannons and plastic pellets at the demonstrators and arrested nearly 1,900. 

People are protesting in several major cities in Turkey, and Pikachu was at one in Antalya, according to local news outlets and social media. In the video, the person in the mascot suit hauls yellow nylon ass as fast as a pair of short, inflated legs can carry them—which is surprisingly fast, actually, considering how they’re keeping up with the people running all around them. The original video was captured by Ismail Koçeroğlu, a photojournalist at Akdeniz University in Antalya.  

On Instagram, Koçeroğlu posted another photo of Pikachu posing with protestors and security.

And because nothing good is safe from AI—not even Protest Pikachu, arguably one of the purest pieces of iconography to come out of the resistance to the worldwide creep of authoritarianism yet—an AI-generated image of Pikachu rushing through the streets alongside protestors went viral shortly after Koçeroğlu’s video. Several local outlets have debunked the image, which is made to look like a high-resolution photojournalism shot from the ground, as being generated with AI. 

The AI image of Pikachu has gone nearly as viral as the real video of the person in a Pikachu costume running away from the cops, and shows how people looking to take advantage of any widely covered news event are creating AI imagery in near real time with the event itself. 404 Media saw various people sharing the AI image of Pikachu as though it were real, and on first glance it was difficult for us to tell that it was fake, especially because the real video of Pikachu running away is blurry. But, as several news outlets in Turkey have already pointed out, things like mixed-up lettering on the police jackets, distorted details, and inconsistencies in the street lamps give it away as fake.  

Pikachu has always been for the people, showing up at rallies and protests around the world. 

Erdoğan recently called the demonstrations "street terrorism,” which technically makes Pikachu a terrorist in the eyes of the president of Turkey. In the midst of widespread turmoil, President Donald Trump praised Erdoğan, calling him a “good leader.” 

Protest Pikachu isn’t the first to show up to an anti-Erdoğan protest in an inflatable suit: A young woman came to a protest earlier this week in a dinosaur costume. 

On Tuesday, OpenAI updated ChatGPT with new AI image generation capabilities that make it especially good at recreating specific visual styles. People trying the new feature immediately flooded social media with images in the style of Studio Ghibli and Japanese animation legend Hayao Miyazaki, who famously called artificial intelligence “an insult to life itself.”

Miyazaki’s quote comes from a 2016 documentary in which he’s shown a demo of a 3D model whose movements are animated with AI as opposed to manually, by a human, as is usually the case in 3D animated videos or video games. 

“Thinking of him, I can’t watch this stuff and find [it] interesting,” Miyazaki says after seeing the demo, saying it reminds him of a friend with a disability. “Whoever creates this stuff has no idea what pain is whatsoever. I am utterly disgusted. If you really want to make creepy stuff, you can go ahead and do it. I would never wish to incorporate this technology into my work at all. I strongly feel that this is an insult to life itself.”

AOL.com is using AI to write captions for photos, which gave cutesy captions to photos of a man who allegedly tried to throw his wife off of a cliff in Hawaii and who has been charged with attempted murder. 

The article, “Top Doctor Allegedly Tried Pushing Wife Off Hawaii Beauty Spot in Wild Homicide Attempt,” was syndicated from the website BoredPanda. On the BoredPanda version of the article, there are no image captions. On the AOL.com version of the article, images of Gerhardt Konig, who was charged with attempting to murder his wife, have captions like “A man smiling in a park setting with a dog, related to a top doctor news story,” “a couple smiling on a beach at sunset, associated with Hawaii doctor incident,” “A couple smiling under a floral arch, outdoors during a wedding ceremony; husband in gray suit, wife in white gown,” and “I’m sorry, I can’t help with that,” which seems to be an instance of the AI not being able to describe an image. A caption on an image of Konig’s wife reads “smiling woman outdoors, linked to top doctor and Hawaii beauty spot incident.” A screenshot of a social media comment is captioned “Comment on potential doctor pushing wife from Hawaii spot, questioning medical or mental conditions.”

The captions were first spotted by John Oxley on Bluesky.

All of this suggests a general carelessness that is now happening all over the internet as a result of news outlets and websites cutting staff and replacing important human tasks with AI. It is also reminiscent of the AI-powered technology used by Buzzfeed that suggested readers buy the clothing worn by people who were criminals or who had been violently attacked, died in tragic accidents, etc.

It’s important to understand how and why this seemingly happened. When I looked at the source code for the AOL.com page, the AI-generated captions actually weren’t captions at all. They were alt text, which is a written description of images or graphics, and which are very important for accessibility, because alt text can be read by screen readers for people who are visually impaired. Alt text is also indexed by search engines and will display if someone’s internet connection is bad or the image file gets broken in some way. The AOL page was set up to display alt text as captions if there was no actual caption written. 

What happened on AOL.com is careless because the AI-generated alt text is not particularly good and because many of the captions included in the article are, again, very cutesy about an attempted murder suspect. They should have been caught and corrected by a human. But generating alt text for images is one of the few things where generative AI actually shows some promise, and where even automated alt text is often an improvement on the status quo, which is no alt text at all. Human beings often fail to add alt text, don’t write useful alt text, or write too much alt text. A 2019 study showed that, at the time, just .1 percent of tweets with images in them contained alt text.

Many accessibility groups warn that alt text should not just be fully automated like appears to be happening in the AOL article. Like everything else AI, AI for alt text often misses the broader context, gets things wrong, has a wildly inappropriate tone, or generates errors. “Since somewhat accurate alt text is arguably better than no alt text, there’s a defensible use case for generative A.I., particularly when websites have thousands of untagged images,” the Bureau of Internet Accessibility, a company that helps websites comply with accessibility requirements, wrote in a blog post. “For now, though, we strongly recommend writing alt text yourself.”

Ohio State University, which has a pretty extensive alt text guide, notes “AI alt text is generally considered to be subpar by accessibility professionals and should not be relied upon. This is because it lacks the context of an image and its use. AI can look at your profile picture and provide you with a description ‘A middle-aged man wearing a suit with a full beard smiling into the camera,’ which is baseline and generic, and does not place it in the context of the wider page.”

Perkins School for the Blind, meanwhile, wrote that AI-generated alt text is often “so vague that it could describe an infinite number of scenarios.”

Yahoo, which owns AOL, did not respond to a request for comment. 

One of the most basic tenets of cybersecurity is that you must “consider your threat model” when trying to keep your data and your communications safe, and then take appropriate steps to protect yourself. 

This means you need to consider who you are, what you are talking about, and who may want to know that information (potential adversaries) for any given account, conversation, etc. The precautions you want to take to protect yourself if you are a random person messaging your partner about what you want to eat for dinner may be different than those you’d want to take, if, hypothetically, you are the Secretary of Defense of the United States or a National Security Advisor talking to top administration officials about your plans for bombing an apartment building in Yemen. 

Things you might consider when doing any sort of communication, if you are thinking about your threat model, would be “what messaging app should I use?”, “Is it end-to-end-encrypted?”, “What device should I use to send the message,” “Do I have two-factor authentication on?”, “What type of two-factor authentication is it (app or SMS based? Hardware based?),” and, crucially, “How widely do I want to share this information?” End-to-end encryption means that a message is encrypted on the device itself before being sent; this means that it is then decrypted at the “endpoint,” meaning that only the intended recipient should be able to read it.

This is all, of course, a very long way of saying that there is no messaging app that can protect you if you are wildly careless, or more generally an idiot. There is no threat modeling that can account for you sending information directly to someone who you do not want to have it, which is exactly what Pete Hegseth, national security advisor Michael Waltz, vice president JD Vance, director of national intelligence Tulsi Gabbard, and a host of other top administration officials did when texting about their plans to bomb a suspected terrorist’s girlfriend’s apartment building in Yemen. When doing threat modeling from here on out, it is now unfortunately important to consider the question "Am I a moron?"

As Joseph has laid out here, there are design changes that Signal could make that would make it less likely for someone to accidentally message the wrong person or accidentally add them to the wrong group chat. At the moment, it can be difficult to know who someone is after you’ve added them to your contacts, because Signal doesn’t force you to select a profile picture or set nicknames for contacts, and, you can’t always see a person’s username or phone number after you’ve begun chatting with them on Signal. 

THAT SAID, top officials in the executive branch should not be using Signal to communicate about military actions at all because the threat model for this sort of communication is so extraordinary and unique (and bound by retention laws) that they should be communicating on existing government channels designed for this exact purpose and which don’t have disappearing message functionality. And even if Signal’s UI could be slightly better or less confusing, if you are sharing bombing plans then you should probably take extra steps to make sure “We are currently clean on OPSEC” is actually true. 

Since the first Atlantic story broke, people in my life have asked me if Signal is secure. Of the commercially available, widely-used messaging apps, Signal has extremely good security. But using Signal on whatever device the officials happened to be using makes those devices a target, and sophisticated nation state actors capable of hacking iPhones and other new smartphones are definitely in Pete Hegseth’s and Michael Waltz’s threat model. The truth of the matter is that no phone, no app, no encryption can protect you from yourself if you send the information you’re trying to hide directly to someone you don’t want to have it.

You all already know the story about national security leaders, Signal, and The Atlantic by now. But to summarize in one sentence: a top U.S. official accidentally added the editor-in-chief of The Atlantic to a group chat on the secure messaging app Signal, and members of the group chat then discussed plans for striking Houthi targets (and with what weapons) before they happened or were public knowledge, resulting in a catastrophic leak of information bringing up all sorts of questions about why top U.S. brass were sharing these details on a consumer app, potentially on their personal phones, and not a communications channel approved for the sharing of classified information or combat plans.

According to screenshots of the chats and the group chat’s members published by The Atlantic on Wednesday, the outlet’s editor Jeffrey Goldberg used the display name “JG” on Signal. He also said in the original article that he displayed as JG. Presumably National Security Adviser Michael Waltz, who accidentally added Goldberg, added the wrong JG. This is a big, big mistake obviously.

But there is a somewhat overlooked setting inside Signal that can ensure you don’t make the same mistake. It’s the nickname feature. First, take a look at my Signal when I search for “Jason” when trying to make a new group and add members to it.