This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. How I learned to stop worrying and love AI slop —Caiwei Chen If I were to locate the moment AI slop broke through into popular consciousness, I’d pick the video of rabbits bouncing on a trampoline that went viral last summer. For many savvy internet users, myself included, it was the first time we were fooled by an AI video, and it ended up spawning a wave of almost identical generated clips.
My first reaction was that, broadly speaking, all of this sucked. That’s become a familiar refrain, in think pieces and at dinner parties. Everything online is slop now—the internet “enshittified,” with AI taking much of the blame. Initially, I largely agreed. But then friends started sharing AI clips in group chats that were compellingly weird, or funny. Some even had a grain of brilliance.  I had to admit I didn’t fully understand what I was rejecting—what I found so objectionable. To try to get to the bottom of how I felt (and why), I spoke to the people making the videos, a company creating bespoke tools for creators, and experts who study how new media becomes culture. What I found convinced me that maybe generative AI will not end up ruining everything after all. Read the full story.
A new CRISPR startup is betting regulators will ease up on gene-editingHere at MIT Technology Review we’ve been writing about the gene-editing technology CRISPR since 2013, calling it the biggest biotech breakthrough of the century. Yet so far, there’s been only one gene-editing drug approved, and it’s been used commercially on only about 40 patients, all with sickle-cell disease.It’s becoming clear that the impact of CRISPR isn’t as big as we all hoped. In fact, there’s a pall of discouragement over the entire field—with some journalists saying the gene-editing revolution has “lost its mojo.”So what will it take for CRISPR to help more people? A new startup says the answer could be an “umbrella approach” to testing and commercializing treatments which could avoid costly new trials or approvals for every new version. Read the full story. —Antonio Regalado America’s new dietary guidelines ignore decades of scientific research The first days of 2026 have brought big news for health. On Wednesday, health secretary Robert F. Kennedy Jr. and his colleagues at the Departments of Health and Human Services and Agriculture unveiled new dietary guidelines for Americans. And they are causing a bit of a stir.That’s partly because they recommend products like red meat, butter, and beef tallow—foods that have been linked to cardiovascular disease, and that nutrition experts have been recommending people limit in their diets.These guidelines are a big deal—they influence food assistance programs and school lunches, for example. Let’s take a look at the good, the bad, and the ugly advice being dished up to Americans by their government. —Jessica Hamzelou This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here.

The must-reads I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology. 1 Grok has switched off its image-generating function for most usersFollowing a global backlash to its sexualized pictures of women and children. (The Guardian)+ Elon Musk has previously lamented the “guardrails” around the chatbot. (CNN)+ XAI has been burning through cash lately. (Bloomberg $) 2 Online sleuths tried to use AI to unmask the ICE agent who killed a womanThe problem is, its results are far from reliable. (WP $)+ The Trump administration is pushing videos of the incident filmed from a specific angle. (The Verge)+ Minneapolis is struggling to make sense of the shooting of Renee Nicole Good. (WSJ $)3 Smartphones and PCs are about to get more expensiveYou can thank the memory chip shortage sparked by the AI data center boom. (FT $)+ Expect delays alongside those price rises, too. (Economist $)4 NASA is bringing four of the seven ISS crew members back to EarthIt’s not clear exactly why, but it said one of them experienced a “medical situation” earlier this week. (Ars Technica) 5 The vast majority of humanoid robots shipped last year were from ChinaThe country is dominating early supply for the bipedal machines. (Bloomberg $)+ Why a Chinese robot vacuum firm is moving into EVs. (Wired $)+ China’s EV giants are betting big on humanoid robots. (MIT Technology Review) 6 New Jersey has banned students’ phones in schoolsIt’s the latest in a long line of states to restrict devices during school hours. (NYT $) 7 Are AI coding assistants getting worse?This data scientist certainly seems to think so. (IEEE Spectrum)+ AI coding is now everywhere. But not everyone is convinced. (MIT Technology Review) 8 How to save wine from wildfires 🍇Smoke leaves the alcohol with an ashy taste, but a group of scientists are working on a solution. (New Yorker $)
9 Celebrity Letterboxd accounts are good funUnsurprisingly, a subset of web users have chosen to hound them. (NY Mag $)10 Craigslist refuses to dieThe old-school classifieds corner of the web still has a legion of diehard fans. (Wired $)
Quote of the day “Tools like Grok now risk bringing sexual AI imagery of children into the mainstream. The harms are rippling out.” —Ngaire Alexander, head of the Internet Watch Foundation’s reporting hotline, explains the dangers around low-moderation AI tools like Grok to the Wall Street Journal. One more thing How to measure the returns on R&D spendingGiven the draconian cuts to US federal funding for science, it’s worth asking some hard-nosed money questions: How much should we be spending on R&D? How much value do we get out of such investments, anyway?To answer that, in several recent papers, economists have approached this issue in clever new ways.  And, though they ask slightly different questions, their conclusions share a bottom line: R&D is, in fact, one of the better long-term investments that the government can make. Read the full story.
—David Rotman We can still have nice things A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.) + Bruno Mars is back, baby!+ Hmm, interesting: Apple’s new Widow’s Bay show is inspired by both Stephen King and Donald Glover, which is an intriguing combination.+ Give this man control of the new Lego AI bricks!+ An iron age war trumpet recently uncovered in Britain is the most complete example discovered anywhere in the world.

Here at MIT Technology Review we’ve been writing about the gene-editing technology CRISPR since 2013, calling it the biggest biotech breakthrough of the century. Yet so far, there’s been only one gene-editing drug approved. It’s been used commercially on only about 40 patients, all with sickle-cell disease. It’s becoming clear that the impact of CRISPR isn’t as big as we all hoped. In fact, there’s a pall of discouragement over the entire field—with some journalists saying the gene-editing revolution has “lost its mojo.” So what will it take for CRISPR to help more people? A new startup says the answer could be an “umbrella approach” to testing and commercializing treatments. Aurora Therapeutics, which has $16 million from Menlo Ventures and counts CRISPR co-inventor Jennifer Doudna as an advisor, essentially hopes to win approval for gene-editing drugs that can be slightly adjusted, or personalized, without requiring costly new trials or approvals for every new version. The need to change regulations around gene-editing treatments was endorsed in November by the head of the US Food and Drug Administration, Martin Makary, who said the agency would open a “new” regulatory pathway for “bespoke, personalized therapies” that can’t easily be tested in conventional ways. 
Aurora’s first target, the rare inherited disease phenylketonuria, also known as PKU, is a case in point. People with PKU lack a working version of an enzyme needed to use up the amino acid phenylalanine, a component of pretty much all meat and protein. If the amino acid builds up, it causes brain damage. So patients usually go on an onerous “diet for life” of special formula drinks and vegetables. In theory, gene editing can fix PKU. In mice, scientists have already restored the gene for the enzyme by rewriting DNA in liver cells, which both make the enzyme and are some of the easiest to reach with a gene-editing drug. The problem is that in human patients, many different mutations can affect the critical gene. According to Cory Harding, a researcher at Oregon Health Sciences University, scientists know about 1,600 different DNA mutations that cause PKU.
There’s no way anyone will develop 1,600 different gene-editing drugs. Instead, Aurora’s goal is to eventually win approval for a single gene editor that, with minor adjustments, could be used to correct several of the most common mutations, including one that’s responsible for about 10% of the estimated 20,000 PKU cases in the US. “We can’t have a separate [clinical trial] for each mutation,” says Edward Kaye, the CEO of Aurora. “The way the FDA approves gene editing has to change, and I think they’ve been very understanding that is the case.” A gene editor is a special protein that can zero in on a specific location in the genome and change it. To prepare one, Aurora will put genetic code for the editor into a nanoparticle along with a targeting molecule. In total, it will involve about 5,000 gene letters. But only 20 of them need to change in order to redirect the treatment to repair a different mutation. “Over 99% of the drug stays the same,” says Johnny Hu, a partner at Menlo Ventures, which put up the funding for the startup. The new company came together after Hu met over pizza with Fyodor Urnov, an outspoken gene-editing scientist at the University of California, Berkeley, who is Aurora’s cofounder and sits on its board. In 2022, Urnov had written a New York Times editorial bemoaning the “chasm” between what editing technology can do and the “legal, financial, and organizational” realities preventing researchers from curing people. “I went to Fyodor and said, ‘Hey, we’re getting all these great results in the clinic with CRISPR, but why hasn’t it scaled?” says Hu. Part of the reason is that most gene-editing companies are chasing the same few conditions, such as sickle-cell, where (as luck would have it) a single edit works for all patients. But that leaves around 400 million people who have 7,000 other inherited conditions without much hope to get their DNA fixed, Urnov estimated in his editorial. Then, last May, came the dramatic demonstration of the first fully “personalized” gene-editing treatment. A team in Philadelphia, assisted by Urnov and others, succeeded in correcting the DNA of a baby, named KJ Muldoon, who had an entirely unique mutation that caused a metabolic disease. Though it didn’t target PKU, the project showed that gene editing could theoretically fix some inherited diseases “on demand.” 

It also underscored a big problem. Treating a single child required a large team and cost millions in time, effort, and materials—all to create a drug that would never be used again.  That’s exactly the sort of situation the new “umbrella” trials are supposed to address. Kiran Musunuru, who co-led the team at the University of Pennsylvania, says he’s been in discussions with the FDA to open a study of bespoke gene editors this year focusing on diseases of the type Baby KJ had, called urea cycle disorders. Each time a new patient appears, he says, they’ll try to quickly put together a variant of their gene-editing drug that’s tuned to fix that child’s particular genetic problem. Musunuru, who isn’t involved with Aurora, does not think the company’s plans for PKU count as fully personalized editors. “These corporate PKU efforts have nothing whatsoever to do with Baby KJ,” he says. He says his center continues to focus on mutations “so ultra-rare that we don’t see any scenario where a for-profit gene-editing company would find that indication to be commercially viable.” Instead, what’s occurring in PKU, says Musunuru, is that researchers have realized they can assemble “a bunch” of the most frequent mutations “into a large enough group of patients to make a platform PKU therapy commercially viable.”  While that would still leave out many patients with extra-rare gene errors, Musunuru says any gene-editing treatment at all would still be “a big improvement over the status quo, which  is zero genetic therapies for PKU.”

The new year has barely begun, but the first days of 2026 have brought big news for health. On Monday, the US’s federal health agency upended its recommendations for routine childhood vaccinations—a move that health associations worry puts children at unnecessary risk of preventable disease. There was more news from the federal government on Wednesday, when health secretary Robert F. Kennedy Jr. and his colleagues at the Departments of Health and Human Services and Agriculture unveiled new dietary guidelines for Americans. And they are causing a bit of a stir. That’s partly because they recommend products like red meat, butter, and beef tallow—foods that have been linked to cardiovascular disease, and that nutrition experts have been recommending people limit in their diets. These guidelines are a big deal—they influence food assistance programs and school lunches, for example. So this week let’s look at the good, the bad, and the ugly advice being dished up to Americans by their government.
The government dietary guidelines have been around since the 1980s. They are updated every five years, in a process that typically involves a team of nutrition scientists who have combed over scientific research for years. That team will first publish its findings in a scientific report, and, around a year later, the finalized Dietary Guidelines for Americans are published. The last guidelines covered the period 2020 to 2025, and new guidelines were expected in the summer of 2025. Work had already been underway for years; the scientific report intended to inform them was published back in 2024. But the publication of the guidelines was delayed by last year’s government shutdown, Kennedy said last year. They were finally published yesterday.
Nutrition experts had been waiting with bated breath. Nutrition science has evolved slightly over the last five years, and some were expecting to see new recommendations. Research now suggests, for example, that there is no “safe” level of alcohol consumption. We are also beginning to learn more about health risks associated with some ultraprocessed foods (although we still don’t have a good understanding of what they might be, or what even counts as “ultraprocessed”.) And some scientists were expecting to see the new guidelines factor in environmental sustainability, says Gabby Headrick, the associate director of food and nutrition policy at George Washington University’s Institute for Food Safety & Nutrition Security in Washington DC. They didn’t. Many of the recommendations are sensible. The guidelines recommend a diet rich in whole foods, particularly fresh fruits and vegetables. They recommend avoiding highly processed foods and added sugars. They also highlight the importance of dietary protein, whole grains, and “healthy” fats. But not all of them are, says Headrick. The guidelines open with a “new pyramid” of foods. This inverted triangle is topped with “protein, dairy, and healthy fats” on one side and “vegetables and fruits” on the other. There are a few problems with this image. For starters, its shape—nutrition scientists have long moved on from the food pyramids of the 1990s, says Headrick. They’re confusing and make it difficult for people to understand what the contents of their plate should look like. That’s why scientists now use an image of a plate to depict a healthy diet. “We’ve been using MyPlate to describe the dietary guidelines in a very consumer-friendly, nutrition-education-friendly way for over the last decade now,” says Headrick. (The UK’s National Health Service takes a similar approach.) And then there’s the content of that food pyramid. It puts a significant focus on meat and whole-fat dairy produce. The top left image—the one most viewers will probably see first—is of a steak. Smack in the middle of the pyramid is a stick of butter. That’s new. And it’s not a good thing.

While both red meat and whole-fat dairy can certainly form part of a healthy diet, nutrition scientists have long been recommending that most people try to limit their consumption of these foods. Both can be high in saturated fat, which can increase the risk of cardiovascular disease—the leading cause of death in the US. In 2015, on the basis of limited evidence, the World Health Organization classified red meat as “probably carcinogenic to humans.”  Also concerning is the document’s definition of “healthy fats,” which includes butter and beef tallow (a MAHA favorite). Neither food is generally considered to be as healthy as olive oil, for example. While olive oil contains around two grams of saturated fat per tablespoon, a tablespoon of beef tallow has around six grams of saturated fat, and the same amount of butter contains around seven grams of saturated fat, says Headrick. “I think these are pretty harmful dietary recommendations to be making when we have established that those specific foods likely do not have health-promoting benefits,” she adds. Red meat is not exactly a sustainable food, and neither are dairy products. And the advice on alcohol is relatively vague, recommending that people “consume less alcohol for better overall health” (which might leave you wondering: Less than what?). There are other questionable recommendations in the guidelines. Americans are advised to include more protein in their diets—at levels between 1.2 and 1.6 grams daily per kilo of body weight, 50% to 100% more than recommended in previous guidelines. There’s a risk that increasing protein consumption to such levels could raise a person’s intake of both calories and saturated fats to unhealthy levels, says José Ordovás, a senior nutrition scientist at Tufts University. “I would err on the low side,” he says. Some nutrition scientists are questioning why these changes have been made. It’s not as though the new recommendations were in the 2024 scientific report. And the evidence on red meat and saturated fat hasn’t changed, says Headrick. In reporting this piece, I contacted many contributors to the previous guidelines, and some who had led research for 2024’s scientific report. None of them agreed to comment on the new guidelines on the record. Some seemed disgruntled. One merely told me that the process by which the new guidelines had been created was “opaque.” “These people invested a lot of their time, and they did a thorough job [over] a couple of years, identifying [relevant scientific studies],” says Ordovás. “I’m not surprised that when they see that [their] work was ignored and replaced with something [put together] quickly, that they feel a little bit disappointed,” he says. This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here.

In partnership withInvisible Enterprises are sitting on vast quantities of unstructured data, from call records and video footage to customer complaint histories and supply chain signals. Yet this invaluable business intelligence, estimated to make up as much as 90% of the data generated by organizations, historically remained dormant because its unstructured nature makes analysis extremely difficult. But if managed and centralized effectively, this messy and often voluminous data is not only a precious asset for training and optimizing next-generation AI systems, enhancing their accuracy, context, and adaptability, it can also deliver profound insights that drive real business outcomes. A compelling example of this can be seen in the US NBA basketball team the Charlotte Hornets who successfully leveraged untapped video footage of gameplay—previously too copious to watch and too unstructured to analyze—to identify a new competition-winning recruit. However, before that data could deliver results, analysts working for the team first had to overcome the critical challenge of preparing the raw, unstructured footage for interpretation. The challenges of organizing and contextualizing unstructured data Unstructured data presents inherent difficulties due to its widely varying format, quality, and reliability, requiring specialized tools like natural language processing and AI to make sense of it.
Every organization’s pool of unstructured data also contains domain-specific characteristics and terminology that generic AI models may not automatically understand. A financial services firm, for example, cannot simply use a general language model for fraud detection. Instead, it needs to adapt the model to understand regulatory language, transaction patterns, industry-specific risk indicators, and unique company context like data policies. The challenge intensifies when integrating multiple data sources with varying structures and quality standards, as teams may struggle to distinguish valuable data from noise.
How computer vision gave the Charlotte Hornets an edge  When the Charlotte Hornets set out to identify a new draft pick for their team, they turned to AI tools including computer vision to analyze raw game footage from smaller leagues, which exist outside the tiers of the game normally visible to NBA scouts and, therefore, are not as readily available for analysis. “Computer vision is a tool that has existed for some time, but I think the applicability in this age of AI is increasing rapidly,” says Jordan Cealey, senior vice president at AI company Invisible Technologies, which worked with the Charlotte Hornets on this project. “You can now take data sources that you’ve never been able to consume, and provide an analytical layer that’s never existed before.” By deploying a variety of computer vision techniques, including object and player tracking, movement pattern analysis, and geometrically mapping points on the court, the team was able to extract kinematic data, such as the coordinates of players during movement, and generate metrics like speed and explosiveness to acceleration.  This provided the team with rich, data-driven insights about individual players, helping them to identify and select a new draft whose skill and techniques filled a hole in the Charlotte Hornets’ own capabilities. The chosen athlete went on to be named the most valuable player at the 2025 NBA Summer League and helped the team win their first summer championship title. Annotation of a basketball match Before data from game footage can be used, it needs to be labeled so the model can interpret it. The x and y coordinates of the individual players, seen here in bounding boxes, as well as other features in the scene, are annotated so the model can identify individuals and track their movements through time.

[embedded content]

Taking AI pilot programs into production  From this successful example, several lessons can be learned. First, unstructured data must be prepared for AI models through intuitive forms of collection, and the right data pipelines and management records. “You can only utilize unstructured data once your structured data is consumable and ready for AI,” says Cealey. “You cannot just throw AI at a problem without doing the prep work.”  For many organizations, this might mean they need to find partners that offer the technical support to fine-tune models to the context of the business. The traditional technology consulting approach, in which an external vendor leads a digital transformation plan over a lengthy timeframe, is not fit for purpose here as AI is moving too fast and solutions need to be configured to a company’s current business reality.  Forward-deployed engineers (FDEs) are an emerging partnership model better suited to the AI era. Initially popularized by Palantir, the FDE model connects product and engineering capabilities directly to the customer’s operational environment. FDEs work closely with customers on-site to understand the context behind a technology initiative before a solution is built. 

“We couldn’t do what we do without our FDEs,” says Cealey. “They go out and fine-tune the models, working with our human annotation team to generate a ground truth dataset that can be used to validate or improve the performance of the model in production.” Second, data needs to be understood within its own context, which requires models to be carefully calibrated to the use case. “You can’t assume that an out-of-the-box computer vision model is going to give you better inventory management, for example, by taking that open source model and applying it to whatever your unstructured data feeds are,” says Cealey. “You need to fine-tune it so it gives you the data exports in the format you want and helps your aims. That’s where you start to see high-performative models that can then actually generate useful data insights.”  For the Hornets, Invisible used five foundation models, which the team fine-tuned to context-specific data. This included teaching the models to understand that they were “looking at” a basketball court as opposed to, say, a football field; to understand how a game of basketball works differently from any other sport the model might have knowledge of (including how many players are on each team); and to understand how to spot rules like “out of bounds.” Once fine-tuned, the models were able to capture subtle and complex visual scenarios, including highly accurate object detection, tracking, postures, and spatial mapping. Lastly, while the AI technology mix available to companies changes by the day, they cannot eschew old-fashioned commercial metrics: clear goals. Without clarity on the business purpose, AI pilot programs can easily turn into open-ended, meandering research projects that prove expensive in terms of compute, data costs, and staffing.  “The best engagements we have seen are when people know what they want,” Cealey observes. “The worst is when people say ‘we want AI’ but have no direction. In these situations, they are on an endless pursuit without a map.” This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff. It was researched, designed, and written by human writers, editors, analysts, and illustrators. This includes the writing of surveys and collection of data for surveys. AI tools that may have been used were limited to secondary production processes that passed thorough human review.

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Researchers are getting organoids pregnant with human embryos At first glance, it looks like the start of a human pregnancy: A ball-shaped embryo presses into the lining of the uterus then grips tight, burrowing in as the first tendrils of a future placenta appear. This is implantation—the moment that pregnancy officially begins. Only none of it is happening inside a body. These images were captured in a Beijing laboratory, inside a microfluidic chip, as scientists watched the scene unfold.
In three recent papers published by Cell Press, scientists report what they call the most accurate efforts yet to mimic the first moments of pregnancy in the lab. They’ve taken human embryos from IVF centers and let these merge with “organoids” made of endometrial cells, which form the lining of the uterus. Read our story about their work, and what might come next. —Antonio Regalado
LLMs contain a LOT of parameters. But what’s a parameter? A large language model’s parameters are often said to be the dials and levers that control how it behaves. Think of a planet-size pinball machine that sends its balls pinging from one end to the other via billions of paddles and bumpers set just so. Tweak those settings and the balls will behave in a different way.   OpenAI’s GPT-3, released in 2020, had 175 billion parameters. Google DeepMind’s latest LLM, Gemini 3, may have at least a trillion—some think it’s probably more like 7 trillion—but the company isn’t saying. (With competition now fierce, AI firms no longer share information about how their models are built.) But the basics of what parameters are and how they make LLMs do the remarkable things that they do are the same across different models. Ever wondered what makes an LLM really tick—what’s behind the colorful pinball-machine metaphors? Let’s dive in.  —Will Douglas Heaven What new legal challenges mean for the future of US offshore wind

For offshore wind power in the US, the new year is bringing new legal battles. On December 22, the Trump administration announced it would pause the leases of five wind farms currently under construction off the US East Coast. Developers were ordered to stop work immediately. The cited reason? Concerns that turbines can cause radar interference. But that’s a known issue, and developers have worked with the government to deal with it for years. Companies have been quick to file lawsuits, and the court battles could begin as soon as this week. Here’s what the latest kerfuffle might mean for the US’s struggling offshore wind industry. —Casey Crownhart This story is from The Spark, our weekly newsletter that explains the tech that could combat the climate crisis. Sign up to receive it in your inbox every Wednesday. The must-reads
I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology. 1 Google and Character.AI have agreed to settle a lawsuit over a teenager’s deathIt’s one of five lawsuits the companies have settled linked to young people’s deaths this week. (NYT $)+ AI companions are the final stage of digital addiction, and lawmakers are taking aim. (MIT Technology Review)2 The Trump administration’s chief output is online trollingWitness the Maduro memes. (The Atlantic $)
3 OpenAI has created a new ChatGPT Health feature It’s dedicated to analyzing medical results and answering health queries. (Axios)+ AI chatbots fail to give adequate advice for most questions relating to women’s health. (New Scientist $)+ AI companies have stopped warning you that their chatbots aren’t doctors. (MIT Technology Review) 4 Meta’s acquisition of Manus is being probed by ChinaHolding up the purchase gives it another bargaining chip in its dealings with the US. (CNBC)+ What happened when we put Manus to the test. (MIT Technology Review)5 China is building humanoid robot training centersTo address a major shortage of the data needed to make them more competent. (Rest of World)+ The robot race is fueling a fight for training data. (MIT Technology Review) 6 AI still isn’t close to automating our jobsThe technology just fundamentally isn’t good enough yet—for now. (WP $) 7 Weight regain seems to happen within two years of quitting the jabsThat’s the conclusion of a review of more than 40 studies. But dig into the details, and it’s not all bad news. (New Scientist $)8 This Silicon Valley community is betting on algorithms to find loveWhich feels like a bit of a fool’s errand. (NYT $)9 Hearing aids are about to get really goodYou can—of course—thank advances in AI. (IEEE Spectrum) 10 The first 100% AI-generated movie will hit our screen within three yearsThat’s according to Roku’s founder Anthony Wood. (Variety $)+ How do AI models generate videos? (MIT Technology Review)
Quote of the day “I’ve seen the video. Don’t believe this propaganda machine. ”  —Minnesota’s governor Tim Walz responds on X to Homeland Security’s claim that ICE’s shooting of a woman in Minneapolis was justified.
One more thing Inside the strange limbo facing millions of IVF embryosMillions of embryos created through IVF sit frozen in time, stored in cryopreservation tanks around the world. The number is only growing thanks to advances in technology, the rising popularity of IVF, and improvements in its success rates.At a basic level, an embryo is simply a tiny ball of a hundred or so cells. But unlike other types of body tissue, it holds the potential for life. Many argue that this endows embryos with a special moral status, one that requires special protections.The problem is that no one can really agree on what that status is. So while these embryos persist in suspended animation, patients, clinicians, embryologists, and legislators must grapple with the essential question of what we should do with them. What do these embryos mean to us? Who should be responsible for them? Read the full story. —Jessica Hamzelou We can still have nice things A place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.) + I love hearing about musicians’ favorite songs 🎶+ Here are some top tips for making the most of travelling on your own.+ Check out just some of the excellent-sounding new books due for publication this year.+ I could play this spherical version of Snake forever (thanks Rachel!)

For offshore wind power in the US, the new year is bringing new legal battles. On December 22, the Trump administration announced it would pause the leases of five wind farms currently under construction off the US East Coast. Developers were ordered to stop work immediately. The cited reason? National security, specifically concerns that turbines can cause radar interference. But that’s a known issue, and developers have worked with the government to deal with it for years. Companies have been quick to file lawsuits, and the court battles could begin as soon as this week. Here’s what the latest kerfuffle might mean for the struggling offshore wind industry in the US.
This pause affects $25 billion in investment in five wind farms: Vineyard Wind 1 off Massachusetts, Revolution Wind off Rhode Island, Sunrise Wind and Empire Wind off New York, and Coastal Virginia Offshore Wind off Virginia. Together, those projects had been expected to create 10,000 jobs and power more than 2.5 million homes and businesses. In a statement announcing the move, the Department of the Interior said that “recently completed classified reports” revealed national security risks, and that the pause would give the government time to work through concerns with developers. The statement specifically says that turbines can create radar interference (more on the technical details here in a moment).
Three of the companies involved have already filed lawsuits, and they’re seeking preliminary injunctions that would allow construction to continue. Orsted and Equinor (the developers for Revolution Wind and Empire Wind, respectively) told the New York Times that their projects went through lengthy federal reviews, which did address concerns about national security. This is just the latest salvo from the Trump administration against offshore wind. On Trump’s first day in office, he signed an executive order stopping all new lease approvals for offshore wind farms. (That order was struck down by a judge in December.) The administration previously ordered Revolution Wind to stop work last year, also citing national security concerns. A federal judge lifted the stop-work order weeks later, after the developer showed that the financial stakes were high, and that government agencies had previously found no national security issues with the project. There are real challenges that wind farms introduce for radar systems, which are used in everything from air traffic control to weather forecasting to national defense operations. A wind turbine’s spinning can create complex signatures on radar, resulting in so-called clutter. Previous government reports, including one 2024 report from the Department of Energy and a 2025 report from the Government Accountability Office (an independent government watchdog), have pointed out this issue in the past. “To date, no mitigation technology has been able to fully restore the technical performance of impacted radars,” as the DOE report puts it. However, there are techniques that can help, including software that acts to remove the signatures of wind turbines. (Think of this as similar to how noise-canceling headphones work, but more complicated, as one expert told TechCrunch.) But the most widespread and helpful tactic, according to the DOE report, is collaboration between developers and the government. By working together to site and design wind farms strategically, the groups can ensure that the projects don’t interfere with government or military operations. The 2025 GAO report found that government officials, researchers, and offshore wind companies were collaborating effectively, and any concerns could be raised and addressed in the permitting process. This and other challenges threaten an industry that could be a major boon for the grid. On the East Coast where these projects are located, and in New England specifically, winter can bring tight supplies of fossil fuels and spiking prices because of high demand. It just so happens that offshore winds blow strongest in the winter, so new projects, including the five wrapped up in this fight, could be a major help during the grid’s greatest time of need.

One 2025 study found that if 3.5 gigawatts’ worth of offshore wind had been operational during the 2024-2025 winter, it would have lowered energy prices by 11%. (That’s the combined capacity of Revolution Wind and Vineyard Wind, two of the paused projects, plus two future projects in the pipeline.) Ratepayers would have saved $400 million. Before Donald Trump was elected, the energy consultancy BloombergNEF projected that the US would build 39 gigawatts of offshore wind by 2035. Today, that expectation has dropped to just 6 gigawatts. These legal battles could push it lower still. What’s hardest to wrap my head around is that some of the projects being challenged are nearly finished. The developers of Revolution Wind have installed all the foundations and 58 of 65 turbines, and they say the project is over 87% complete. Empire Wind is over 60% done and is slated to deliver electricity to the grid next year. To hit the pause button so close to the finish line is chilling, not just for current projects but for future offshore wind efforts in the US. Even if these legal battles clear up and more developers can technically enter the queue, why would they want to? Billions of dollars are at stake, and if there’s one word to describe the current state of the offshore wind industry in the US, it’s “unpredictable.” This article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.