Buried beneath every American city is a mysterious jumble of cables, pipes, and gas lines, entangled like a pair of wired headphones the moment you pull them out of your pocket. Every time a contractor sticks a shovel in the asphalt, they embark on a high-risk excavate-by-dots obstacle course, dodging color-coded clusters of spaghetti-like power lines, gas mains, sewers, and fiber-optic cables. A line of blue spray paint means water. Red means power. Yellow is gas. Orange is telecom. 

About 700,000 times a year — that’s once every 47 seconds — something goes wrong. 

A backhoe slices into a cable and cuts off a 911 call center, or a woman comes home to find her dryer and hot water heater on fire after a contractor pierced her electrical line. Just last week, a contractor accidentally severed a single telecommunication cable that proved critical enough to ground flights at some of the nation’s largest airports on the eve of the UN General Assembly, ensnaring even world leaders like Canadian Prime Minister Mark Carney in delays. 

The ground beneath our city streets has never been so crowded, driven by a furious race to lay fiber optic cables carrying the data that fills your social media feeds and transmits your ChatGPT queries. Nor have the systems we use to find out where it’s safe to dig — namely, a hotline and all that colorful spray paint — ever felt so catastrophically insufficient. 

“If the locator mismarked the line, if the line wasn’t marked, if there was no map that even said the line was there,” said Sarah Magruder Lyle, president of the Common Ground Alliance, a trade association focused on underground damage prevention, “there are a variety of issues that could happen before the excavator even puts a shovel or a backhoe in the ground.”

According to Common Ground, such incidents cost the US about $83.2 billion per year, roughly as much as we invest in public transit annually nationwide. But the actual financial toll may be even higher, because every time a crew stumbles across a mislabeled power line or a gas pipe everyone else forgot about, it can delay construction for months on end. As Magruder Lyle put it, “if they get out there and they find something that they’re not supposed to, you have to go back and reengineer the whole project.” 

Such surprises can also lead to tragedy. A 5-year-old named Alistair Lamb was killed in Missouri last spring after a contractor drilled into a gas line near his home, spewing fumes that leveled his house when his unsuspecting father flipped a light switch. And just this May, another unmarked gas line killed three people when contractors struck it, leading an apartment complex in Dallas to explode.

Yet the messy underbelly of our cities remains largely invisible, with accountability for issues foisted upon and at times evaded by a hodgepodge of finger-pointing private utility companies, agencies, and contractors. And that’s a shame because there is, in fact, a better way to bury the cables and pipelines that our modern lives depend on, and a multitude of better ways that we can find the ones we’ve already buried. 

What Disney World gets right about the stuff we bury

To understand one utopian vision for how all this could work, look no further than the happiest place on Earth. That’s a moniker Disney World has earned, in my mind, not for its Mickey Mouse waffles (though my niece may disagree), but for its world-class utility corridors, also known as utilidors.

Directly beneath the Magic Kingdom is a sprawling network of tunnels. Initially designed to shepherd around Mickey and his friends, the system now also serves as one of the largest utility corridors ever created. Unlike the rest of America, where cables are typically dumped in ditches before being covered with dirt, the electrical lines that power Cinderella’s castle quite literally get the princess treatment. 

Alongside water pipes, sewers, and animatronic control systems, these power lines lace the walls and ceilings of the utilidors, meaning they’re protected from weather and changes in the soil. They are also easy to monitor, catalog, and access if something needs fixing. While the cost of building a controlled system like this is much higher upfront than burying directly in the ground, the practice virtually eliminates the risk of accidentally rupturing a gas main or slicing a fiber optic cable during routine maintenance. Once they’re built, they can also drastically reduce costs in the long run because you no longer need to tear up a street every time a utility needs fixing. 

“Obviously, it’s expensive,” said Debra Laefer, a professor of civil and urban engineering at New York University, “but in the long term, it would be hugely cheaper,” especially “in areas where we are digging all the time, there’s a direct business case for moving towards these utilidors.”

As fantastical as Disney World’s imagineered system may seem — and it was, to be clear, predicated on a not-easily-reproducible master plan — the idea of storing utility lines in easy-to-access tunnels is not just for amusement parks. It is possible even in places where most of those lines are currently lying in the dirt, as Prague proved back in the 1970s, when the city decided it was fed up with destroying its precious cobblestone streets every time a buried pipe or cable needed fixing. The city now boasts a sprawling utility tunnel network, occasionally patrolled, evidently, by a couple of robot dogs tasked with mapping out every cable. 

Singapore has also spent years building out walkable utility corridors that operate much like the Magic Kingdom’s, transmitting water, data, and energy from the safety of well-organized underground tunnels. Some of Tokyo’s largest districts store their cables and gas lines in huge concrete utility ducts, and many Parisian water lines and fiber optics are kept across the city’s centuries-old sewer system, a catacombs of waste made easily navigable by blue-and-white street signs that correspond to the avenues above. Many university campuses across the US also boast their own sprawling utilidors.

In most cases, cities chose not to relocate utilities all at once, a feat that would probably be financially prohibitive and deeply disruptive. Rather, they took advantage of all the times they’ve needed to relocate utility lines anyway, a standard (and expensive) practice for big building projects like the Wallis Annenberg Wildlife Crossing or the Second Avenue Subway, a major transportation project in New York City that’s been consistently delayed by unmapped utility lines. A group of graduate students found that building such a tunnel in parts of NYC would likely cost at least twice as much as dropping a utility line in the ground, but could recoup the costs within 47 years. 

“There are streets in New York City that have been excavated at least once a year, every year for the past 30 years,” Laefer said. If you just prioritize those streets, gathering the piping in a “big concrete box” underground, you could allow workers to complete most repairs without needing to disturb the streets or sidewalks above at all.

But building new utility tunnels would require that we first come to better understand the wires and piping already entangled under our cities, which if you recall, we have only a surprisingly flimsy picture of in most cities today. But if every surprise utility line is an opportunity to move it somewhere better, it is also an opportunity, at the very least, to map out where it is.

To that end, a few years ago, Laefer led a $1.2 million pilot funded by the National Science Foundation to create a test case for 3D-mapping New York City’s underground utilities, a project that’s since been adopted into a citywide initiative. The city says it plans to roll out the spruced up utilities locating system in 2028. The UK debuted a similar digital map of over 3 million kilometers of underground pipes and power, owned by hundreds of utilities and agencies a few years ago. The nation projects that it could help save over $500 million per year by minimizing delays and accidental damage. In places where Singapore has not yet stuffed its utilities into walkable tunnels, the city’s own mapping and utilities coordination has already saved it an estimated $230 million in avoided clashes since 2024. Not bad for an intervention that doesn’t even require any physical rebuilding.

So why hasn’t every American city copied London or Singapore?

Well, US cities tend to be plagued by even more private utilities and patchwork local and state norms and regulations than a nation like the United Kingdom. Getting them to pool together what they often consider proprietary underground data is challenging, and few local officials have opted to prioritize such efforts. But this may well change as more and more fiber optics go into the ground, more flights get canceled, more apartment complexes explode, more power lines flicker off, more sinkholes form, and more Americans come to be all too acquainted with the perils lurking just beneath our feet. 

“I think at this point, it’s more about political will and money than any technical obstacle,” Laefer said. “Technically, we’re there. It’s just a matter of doing it.”

The Soviet Union famously churned out tens of thousands of new apartment buildings in the 1960s, ’70s, and ’80s — some assembled in as little as 12 days — together comprising tens of millions of individual apartments. Such speed is not unique to communist command economies, or even to grim concrete mega-blocks. In Sweden and Japan today, modern, high-quality new homes can be factory-built and assembled in a matter of weeks using the same basic method — prefabrication — as the Soviets.

The US, alas, evinces no such urgency to build housing. 

In some of our highest-cost cities — San Francisco, New York, Miami — it takes, somewhat astonishingly, nearly five years to erect a modestly sized apartment building, from the time an application is first submitted to building completion. That was one of the top-line findings of a remarkable new economics working paper on housing construction timelines across 60 US cities. The gaps between the fastest and slowest cities were just as striking: In Raleigh, North Carolina, a builder can complete a 20-unit apartment building in about as much time as it takes just to approve the permit for one in San Francisco, where applications languish for more than a year. In Indianapolis, a single-family home can be built in six months from proposal to completion, less than one-eighth the time it takes in San Francisco.

Such delays have costs — in delayed housing supply at a time when America is millions of homes in the hole, in lost staff time that developers ultimately bake into the cost of the new homes, and in homes that simply never get built because the process is discouraging to anyone but the best-resourced developers. The authors, Leonardo D’Amico, Evan Soltas, and Audrey Wang, found that longer permitting lengths are correlated with higher housing prices and lower rates of housing production, though they don’t claim to have shown that these are causal relationships. The research (which has not yet been through peer review and, like any working paper, may be revised before final publication) represents the first such comprehensive dataset comparing homebuilding timelines across the US. To build it, the researchers assembled 1.1 million permits covering 2.7 million housing units between 2000 and 2025 — about one in 13 new homes permitted nationally over that period, they write. 

It probably won’t come as a shock to you that housing can be built faster in Sioux Falls, South Dakota, than in Los Angeles or pretty much anywhere in California. But perhaps the most notable takeaway (to me, anyway) from the paper was how massively home permitting times have lengthened across the board over the past 25 years. Across the cities in their sample, the researchers estimate that permitting time for a standardized multifamily building rose from 0.6 years in 2000 to 1.5 years in 2023, and, for a single-family home, from 1.6 months in 2000 to 4.9 months in 2023. 

Is there really any good reason why we should take nearly a year longer to approve an apartment building than, Soltas joked in a recent interview, “the barbarians of 2006”? Shouldn’t we be getting faster and more efficient at building housing over time, not slower — especially at a time of acute housing scarcity? 

You would think. So what is going on here?

One building, very different wait times

Permitting refers to everything that happens after a developer submits an application to build a house or multifamily building, before they’re allowed to break ground. The proposal is then reviewed for compliance with a range of different sets of rules, from building safety and energy codes to plumbing to zoning and, in some places, aesthetic design. Initial applications will often be found noncompliant and then sent back to the builder for revisions, Soltas, an assistant professor of economics at Princeton, told me.  

The gulf between housing starts in California and Texas has become something of a meme in housing abundance circles. But, for all that, we have until now lacked rigorous apples-to-apples comparisons of permitting and building times in different metro areas, Soltas said. That gap in seemingly basic knowledge, what should be “step zero” facts for the national housing reform debate, became apparent to him when he was teaching an urban economics class, which became one of the impetuses for the new paper. “I realized if I needed that information to teach my students, that it might be useful for other people,” he said. 

Rather than merely pulling permits from different cities and comparing how long whatever happened to be built in each city took to approve, the researchers used the data to estimate how long an identical, standardized project would take in each place between 2015 and 2019, keeping constant its size and the density of the surrounding neighborhood. Their uniform archetypes were a 2,800-square-foot detached single-family home on a 0.15-acre lot and a 20-unit multifamily building with 1,200 square feet per apartment.

Cities with longer permitting times also tended to have longer construction periods — the longer it takes to get an application approved, the longer it then takes to do the building itself, which compounds the gap between the fastest- and slowest-building cities (it’s possible that “cities that make the permitting process difficult also put in lots of burdens that slow the construction process,” Soltas said). One advantage of the researchers’ method is that the differences across cities can’t merely be chalked up to something like the complexity of building within already-dense environments like Manhattan — the analysis already controls for that. So the enormous gaps between different places become harder to dismiss as inevitable features of development in big cities.

The study wasn’t designed to answer the “why” of it all — those will be questions for future research. “Our paper doesn’t directly identify any smoking guns,” Soltas said. Still, some of the culprits are likely obvious — California, for example, has long subjected some housing projects to extensive environmental review requirements under the California Environmental Quality Act (CEQA). Apartment buildings, especially larger ones, are more likely to have to go through CEQA review, though they’re more environmentally friendly than single-family homes (recognizing this as an obstacle to housing affordability, California has been moving to exempt many housing projects from such reviews). 

Why are we taking so much longer to permit housing?

But why, as the study finds, is permitting becoming slower across the country, not just in the most regulated cities? “We estimate that, in 2000, our standardized multi-family project had a 71-percent probability of approval within six months of applying for a permit,” the researchers write. “In 2025, that project had a 44-percent approval probability within six months.” Permitting times have come down a bit since 2023, but it’s unclear whether that just represents the reversal of pandemic-era highs, or if it reflects real progress from the pro-housing policies that have been passed in many cities and states in recent years. 

That discovery dovetails with other economics research finding that zoning codes, which were first widely adopted in the US in the early-to-mid 20th century, have continued to become more restrictive even in recent years, making it harder to build housing. 

Without knowing why exactly permitting lengths have ballooned, it’s hard to say whether those increases have occurred for good reasons. At least some of it may have to do with the aforementioned intensification of zoning and other land-use restrictions, which, I can say from having written a lot about zoning, tend to be pretty ill-conceived and hard to justify. 

The slowdown may also be connected to expanding building code requirements. I’ve written at length about that issue, too; although ever-more stringent rules might seem like a great idea — who could be opposed to more safety? — they tend to be biased against multi-family housing and likely inflate the cost of building, all without having to show that the benefits are worth the costs. Particularly the cost in housing supply that never gets built, which has the effect of keeping more Americans living in older housing stock. And do we really think a home built to 2005 standards is dangerously subpar today?  

We might ultimately decide that many of today’s building requirements — particularly ones like stronger energy efficiency standards — are worth it, but first we’d need to determine what their full costs are, which we remain far from doing. 

At a minimum, it seems unlikely that an apartment building that’s okay to build in Grand Rapids today would be unsafe for New York. So what, exactly, is all that extra permitting time buying the city? Now, at least, Mayor Zohran Mamdani can put an exact number on the delay. 

My favorite restaurant, here in Madison, Wisconsin, closed this week. 

It was not my favorite because it had the greatest food I’d ever had, nor the most memorable ambiance. But I loved the Mad Rabbit Cafe because it had become my anchor in the neighborhood — it felt like home to me and a core group of friends as surely as the century-old houses lining the surrounding city blocks. Most importantly, it was entirely vegan — a rarity in this city, which, despite its ultra-progressive politics, embraces Wisconsin’s identity as “America’s Dairyland.” It was a refuge where my most closely held values found expression.

That closure is part of a national trend. US plant-based restaurants appear to be going through a contraction, amid the tumultuous aftermath of the Covid-19 pandemic and, perhaps, an anti-plant-based cultural vibe shift.

VegNews magazine reported that data from HappyCow, an app that lists vegan and vegetarian restaurants, finds the number of vegan restaurants in three of the country’s top destinations for vegan dining — New York City, Los Angeles, and Portland, Oregon — decreased by between 21 and 25 percent from 2022 to late 2024. Those cities alone make up a significant share of America’s vegan restaurants. Then, in 2025 and 2026, some of the country’s most renowned, long-running plant-based restaurants also announced that they’d shut down: Modern Love and Candle in New York City, the Vegan Joint in LA, Millennium in Oakland, a more than three-decade-old institution (though it later reopened as a leaner operation), among many others. 

According to a Vegan Society report, in October 2025, HappyCow (which is like Yelp but for plant-based dining) listed 1,717 all-vegan restaurants in the US. As of this writing, HappyCow lists 1,509.

Madison has never had anything like the vegan riches of New York or San Francisco, but even we are having our own vegan recession. When Mad Rabbit opened in early 2024, the city had four fully vegan restaurants; now, depending on how you count, one or two remain (and even the one that’s unambiguously vegan is a food-court-style stall). I visited Mad Rabbit at its last Sunday brunch, where it served what many people have described as the best tofu scramble they’ve ever had, and a crowd-pleasing rotation of crepe specials (this month, it was mango lassi). It’s a culinary niche that, once gone, is hard to replace in a mid-tier city like this one. 

Mad Rabbit’s owner, Nate Anger, told me the business hadn’t been in particularly bad financial shape; it was profitable for each of the three years in which it operated, but not profitable enough to justify the punishing time commitment required to run it. Managing it had taken over his life, as it’s wont to do for many people in the food service industry. “I live here. I spend far more time here than anywhere else,” he told me in a recent interview at the restaurant. “I’ve missed my kids’ baseball game and tournaments and school band recitals.” I looked around at the oddball, DIY space he’d conjured — walls covered in houndstooth print, an Alice in Wonderland-like vortex, and art of rabbits in mischievous-looking moods, one brandishing a pair of large knives (the sort of room you wouldn’t want to encounter while stoned). 

The animal rights movement has two very different emotional registers: the grief and terror of looking squarely at factory farming, and the joy of discovering how much creativity, abundance, and pleasure can exist outside it. All-vegan restaurants represent perhaps the fullest embodiment of that second feeling — like “a little heaven,” as Krupa Panchal, co-owner of Bombay Food Junkies, a vegan Indian spot in suburban St. Louis, put it to me. 

Even as vegan businesses struggle to survive, plant-based options at non-vegan restaurants have become far more mainstream than they were a decade or two ago. Set against the stagnating (if not declining) share of Americans who identify as vegan or vegetarian, that raises difficult and even paradoxical questions for the animal movement.

For that reason, the mass die-off of plant-based restaurants has been much dissected and lamented in the vegan community. As I faced losing my own — a possibility I’d always feared and tried to forestall by supporting it as much as possible — the problem became much more visceral for me. I found myself thinking about why it mattered to me so much: What is lost when these places disappear, and what does that tell us about whether the plant-based movement can grow and endure?

What vegan restaurants are for

The vegan restaurant predates the word “vegan” itself — and so, apparently, do questions about whether they can be financially viable. More than a century ago, French anarchist Georges Butaud co-founded the Foyer Végétalien in Paris, writing in a 1923 letter that he was unsure whether it would attract a sufficient “clientèle végétalienne” to sustain itself. Why start such a business at all, when the audience for plant-based food is so vanishingly small? “I already have some positive results. A few people have become vegan since we opened,” Butaud wrote. The restaurant “is a school where we reciprocally teach one another…I have believed in the usefulness of creating it, and I will be happy if, over the months to come, the number of vegans increases.” 

Many vegan restaurant proprietors today still share that ambition — to be good enough to help the movement grow. 

Long before it became routine to find plant-based options on normie restaurant menus (iceberg lettuce leaves don’t count), plant-based restaurants compensated for vegans’ exclusion from mainstream cuisine. But they have also aimed to provide proof of concept of the diverse, surprising possibilities of plant-based food for a larger audience, showing that plant-based food need not be confined to marginalia. 

Today, of course, restaurants are far more central to our daily lives than they were in Butaud’s time. Americans consume about a third of their calories and the majority of their food budgets outside the home, and food-serving businesses are perhaps our most important third spaces. Meanwhile, the problems associated with animal agriculture and overconsumption of animal products grow unabated in America and around the world. The US alone slaughters 10 billion land animals for meat every year, the overwhelming share of them raised in appalling conditions, and those farmed for milk and eggs endure no better. Reciting these grim facts tends not to be a very fruitful path toward making friends or convincing people that veganism can be a positive and enjoyable lifestyle. Plant-based restaurants, on the other hand, make veganism feel less like a prohibition than an invitation; hospitality rather than horror. 

Anyone who has ever gone vegan knows what it’s like to incessantly be asked the well-meaning but bewildering question: “What do you eat?” (And, I get it! It’s confusing for those of us who grew up viewing food free of animal products as side salad or overcooked green beans.) Plant-based businesses have long been laboratories for answering it, developing techniques that draw richness and variety from ingredients not often treated as the main event in Western cuisines.

The more than 50-year-old Moosewood Restaurant (although mostly vegetarian, not fully vegan) was an early conduit between restaurant experimentation and home cooking, its recipes spreading far beyond Ithaca, New York, through its enormously popular cookbooks. Soul Vegetarian in Atlanta, opened in 1979, developed a plant-based soul food repertoire that helped cut a path for dozens of other vegan soul restaurants across the US; the Chicago Diner, in the city long identified with stockyards and slaughterhouses, helped popularize plant-based versions of American diner and deli staples. And chef Miyoko Schinner, at her erstwhile San Francisco restaurant Now & Zen, experimented with plant-based cheeses long before launching Miyoko’s Creamery, one of the most acclaimed and influential vegan dairy companies of the past decade. 

The 21st century in particular has seen a flowering of vegan restaurants serving every palate and national cuisine, in every possible aesthetic — punk, hippie, psychedelic, fussy and upscale. They still expand my mind today, showing that a lifetime wouldn’t be enough to discover all the ways we can eat without the exploitation of animals.

The longtime Grasshopper restaurant in Boston, whose pan-Asian fare draws partly from Vietnamese Buddhist temple cuisine, makes mock meats that are worlds away from the rubbery substitutes that many Americans associate with the category (it closed its beloved, homey restaurant space in 2024, later reopening a stall inside a food court). Mad Rabbit changed even this longtime vegan’s idea of what tofu could be. Chicago-based Penelope’s Vegan Taqueria makes, as far as I’m concerned, the most exceptional Mexican food — vegan or not — this side of the Rio Grande. V-Grits, a Southern comfort spot in Louisville, Kentucky (RIP, 2024), provided evidence that some of America’s most exciting plant-based food could be found far from the big coastal cities best known for it. 

So why are so many of these places shuttering now? 

The vegan restaurant death spiral? 

Theories abound about the great dying of vegan restaurants.

Some of the casualties, it must be said, just weren’t very good in the first place. The raw vegan boom of the 2000s and 2010s, for example, produced numerous establishments built around that especially unforgiving constraint — no cooking. The result was often less-than-thrilling food (which isn’t to say it’s never good!), all in service of a restriction with scant evidence of health benefits to support it. Lots of these restaurants have closed in the 2020s, reflecting diminishing demand for that particular variant of vegan food. 

But the 2010s weren’t merely a golden age for zucchini noodles and date-walnut brownies — vegan restaurants of all stripes were multiplying, as was interest in the new generation of plant-based meats (Beyond and Impossible). It really did appear, for a minute there, that the future might be plant-based, vindicating decades of animal advocates’ work to push veganism into the mainstream. 

“I felt like veganism had really hit a big stride” in that period, Anger, the Mad Rabbit founder, said. Everyone was streaming documentaries advocating for plant-based diets — Forks Over Knives, Cowspiracy, What the Health, among others. Some polls suggested the share of Americans who called themselves vegan was growing. Anger said his market research led him to believe that the overall trend line for veganism would continue going up, even a little. Instead, “it just rolled right back over.” His restaurant had a core customer base of committed vegans, but its heavy reliance on regulars also showed how limited its reach was; stagnating interest in veganism may have been putting a ceiling on its growth. Keith Uruski, owner of Nooch, a vegan restaurant in Windsor, Ontario, open since 2018, told me that the business had thrived in its first few years, but has since slowed considerably, and he hasn’t paid himself in three years.  

In the years leading up to the pandemic, the assumption that the audience for vegan food would keep growing shaped much bigger bets than Mad Rabbit. Vegan chain restaurants including Veggie Grill, Native Foods, and Planta had pursued aggressive growth plans in the 2010s and early 2020s but eventually retrenched, closing locations and abandoning the scale they’d once envisioned. (Again, some of this may be merit-based rather than evidence of a broad rejection of veganism — YMMV, but I found Veggie Grill, at least last time I had it pre-Covid, to be mid.) “Most dedicated vegan chains are now in kind of a managed decline,” Lizzy Bartlett, senior director of menu research and insights at Technomic, a food service industry research firm, told me.  

The cultural winds have certainly shifted. The Make America Healthy Again (MAHA) movement has probably contributed to a thermostatic swing away from “unnatural”-seeming diets and simulacra of traditional foods, like plant-based meat and milk. I also think society has deep carnist antibodies that activate whenever meat-eating faces heightened moral scrutiny. In the last few years, that’s taken the form of a protein obsession that has hijacked the minds of otherwise rational people, convincing them tofu and beans lack sufficient protein to build muscle.

But perhaps my favorite explanation was articulated by Financial Times journalist Henry Mance last year: “The decline in idealism,” reflected in the reelection of Donald Trump — the triumph of our collective id.  

There’s a risk of reading too much into all this, however. Many restaurants, vegan or not, suffered during Covid and from prolonged inflation in food and energy prices. This year, consumers are spending less at restaurants because of elevated fuel prices due to the war in Iran. Panchal, of Bombay Food Junkies, said that her sales have been down in 2025 and 2026 compared to the years prior, but she doesn’t believe vegan restaurants have been faring worse than non-vegan ones. Some vegan businesses have recently closed in St. Louis for personal reasons rather than failing in the market, she told me.

This is part of the natural life cycle of independent restaurants — many disappear within a few years, and even highly successful ones eventually run their course. US cities had many vegan restaurants open in the 2000s and 2010s, so we might now be seeing a wave of closures that reflects the aging of that cohort. 

Preliminary results shared with Vox from a 2026 survey of the global vegan restaurant industry, by the nonprofit Vegan Hospitality, also complicate the idea that plant-based restaurants are being killed off because of waning demand for vegan food. In responses from about 59 vegan restaurants (a self-selected group, to be sure), mostly in the US and Mexico, roughly four in five said they’d still choose to be fully vegan if they were opening today. Of 13 businesses that were closing or had been sold, six said that being vegan actually helped them, and four said it made no meaningful difference.

And there’s some tension between the “death of veganism” hypothesis and what may be one of the biggest forces now disadvantaging dedicated vegan restaurants: the widespread adoption of plant-based options on omnivorous restaurant menus. That it’s now easier to be vegan at mainstream establishments is “unquestionable,” Anger said. He viewed Mad Rabbit’s competitors as not just fully plant-based businesses, but any restaurant that offers “more than two or three” vegan options; it was located just a block away from one of Madison’s best-regarded restaurants, a Laotian place with several very good plant-based offerings.

Likewise, Michael Fifelski, general manager of Red Sushi in downtown Madison, which has a solid list of vegan options, told me, “Our vegetarian and vegan rolls actually are quite popular” — a foundational part of the menu. There can be a real advantage to simply letting vegan food blend in — Amanda Byrd, the manager of Everly, an eclectic, globally inspired restaurant in Madison, told me that its top dish, a butternut squash and forbidden rice curry, is incidentally plant-based. In many cases, “if you de-emphasize the vegan statement, but the dish happens to be vegan,” it will sell better, Claire Conaghan of the food industry research firm Datassential said.

Though nothing compares, for me, to the feeling of total alignment that comes with dining at an all-vegan restaurant, there is also something affirming about seeing plant-based food finally taken seriously by the wider culture — and vegan dishes surviving on their own merits alongside everything else on the menu.

Do vegan restaurants still matter?

In a world where you can find a satisfying vegan entrée at Chipotle or Taco Bell or your local brewpub, are fully vegan restaurants still needed? What purpose do they serve?

The most fundamental problem beneath the downturn in vegan businesses is simply that the vegan population does not seem to be growing much at all. Polling on the question is messy and often inconsistent, but according to Gallup’s 2026 survey, 2 percent of Americans call themselves vegan, the same figure as in 2012, when the firm first started asking the question. If expanding the number of people who abstain from animal products remains an important goal (a point debated within the animal movement, though I think dietary change still matters a great deal), then which is more likely to move that objective forward: vegan restaurants, or an abundance of vegan options at omnivorous ones? Which does more to increase the number of plant-based meals consumed rather than animal-based ones and generate excitement about vegan food?  

I put those questions to Seth Ariel Green, a research scientist at Stanford’s Humane and Sustainable Food Lab, who writes the Substack “Regression to the Meat.” He’s one of the most intellectually honest researchers anywhere on meat reduction. His answer, via email, is that we don’t really know: “I have no idea how to weigh the pros and cons of, when I’m choosing between meals, 1) signaling to the world that vegan businesses are viable or 2) signaling to a particular business that their vegan options are viable. That’s a hard problem!” (It might be “a microeconomic structural equation modeling paper waiting to be written,” he added.)

Reducing prevailing levels of meat consumption even a tiny bit, much less convincing people to become vegan, is really, maddeningly hard, as Green has written. One frequently cited obstacle to growing the animal rights movement — its “central unsolved problem,” as one movement leader recently put it — is that it lacks much in the way of an in-person social world and sense of community. Vegan restaurants alone can’t fix that, of course, but they are among the few physical institutions the movement has. At their best, they provide a space where a private dietary choice can be reinforced by a shared sense of identity — a place to run into someone you already know or meet someone new, at a speaker event or social mixer. These spontaneous, embodied connections made Mad Rabbit such a warm presence in my life, and its final days all the more melancholy.

That function may be especially useful when plant-based restaurants are rooted in traditions people already belong to — cultural, religious, or otherwise — and help them view veganism as continuous with an identity they already have rather than as something alien to it. As Green has suggested, “let’s tap into pre-existing identities, explain how plant-based diets are consistent with them, and then take folks somewhere to enjoy convenient, delicious plant-based food together.” 

On the other hand, plant-based options at omnivorous establishments matter for almost the opposite reason. People often struggle to stay vegan because eating differently isolates them from non-vegan friends and family; restaurants that can accommodate mixed-diet social groups help remove that social penalty. These options also need a customer base — if no one orders them, they’re unlikely to remain on menus. 

I’m enough of a weirdo that I try to almost exclusively support plant-based restaurants whenever I eat out. I desperately want them to survive, and at the end of the day, no one is more invested in advancing the craft of plant-based cooking than vegan restaurants themselves.

Not everyone needs to be like me — in a world where almost all people still eat animal products, a healthy ecosystem needs both places built entirely around veganism and those where plant-based food is woven into broader food culture. It needs, most of all, excellent vegan food, wherever it might be found. And now that veganism has made at least a small breach into mainstream dining, vegan restaurants still offer something their meat-serving counterparts cannot. They’re like miniature utopias where the normal order of things is inverted and, for the length of a meal, we can see what a future without animal exploitation would look like. 

This story was originally published in The Highlight. To get access to member-exclusive stories like this every month, become a Vox Member today.

Hi readers! Shayla Love here, science journalist and longtime fan of Your Mileage May Vary. I’m honored to be subbing for Sigal Samuel while she’s out on parental leave. I’m diving into your questions as a way to help understand human nature and our choices through multiple lenses: philosophical, psychological, and beyond. Please send in any emotional, body/brain, sociological, perceptual, or other kind of life quandaries you might have.

My best friend confessed his love for me one year ago and I rejected him because I was unhappily in love with another man, and I was never really physically attracted to my best friend. He was devastated. I am happy our friendship survived.

Over a year later, I find myself thinking about how it would be like to be in a relationship with my best friend. He is the only man who ever showed me genuine love and support, and I wonder what if I try…maybe I will fall in love with him. However, I am concerned about hurting his feelings again. Is it selfish of me wanting to try this love out because it promises security and appreciation?

Dear Fanning the Flame,

I can’t blame you for being curious about the common “friends to lovers” storyline. It’s often portrayed as the best kind of pairing: Two people have the foundation of their friendship plus a newfound love; the romantic relationship is like a soaring addition to an existing sturdy house. 

But real life is not always like the movies, and it sounds to me like you haven’t actually had an awakening of desire for your friend. In fact, it’s the opposite: He once expressed feelings for you, and you confronted a lack of physical attraction for this person. Now you’re wondering if it would be possible for you to fall in love if you tried. I think you’d agree that your timing is suspect. After a breakup, it’s common and understandable to seek out comfort and intimacy. Still, I detect a philosophical streak in your question beyond the issue of whether or not this would simply be a rebound. 

If your friend has always appreciated you, made you feel safe, and loves you — could that be a form of love that’s better than what you had in your last relationship? Maybe that is what love is.

Is it? Over a long night of drinking some 2,400 years ago, one group of thinkers allegedly debated the true nature of love, as Greek philosopher Plato describes in The Symposium. 

Aristophanes, a playwright, tells a story that most closely aligns with the love we see in the rom-coms of today. He said that humans used to be larger creatures with four arms and four legs. After annoying the gods, Zeus cuts all the humans in half, creating the two-limbed bodies we have today. The result is that people wander the world looking for the one person who used to literally complete them. If you followed Aristophanes’s fable, you shouldn’t try to be with your friend unless you felt that deep longing that one has when meeting their other half. 

The rest of the dinner party companions have more circumspect ideas. Phaedrus, a Greek aristocrat, thought that true love inspired people to become the best possible version of themselves. Pausanias, a military man, thought that love driven only by attraction to someone’s body was “vulgar love”, and that “noble love” was the love of someone’s values. Eryximachus, a doctor, asserts that love is also about moderation, and it needs order, much in the way the natural world does. Finally, Socrates says that even if you start out with physical infatuation, you can eventually transcend that and end up loving wisdom itself, the highest form of love. As I read it, the men are grappling with how to make love make sense, while acknowledging that it can make us lustful and impulsive. Shouldn’t its best manifestations instead be practical and virtuous?   

You could easily justify trying to date your friend with these theories that try to wrangle and tame love. For example, if your friend inspires you to be the best version of yourself, you could decide that’s enough to go for it. You could also decide that because you admire your friend’s character, and not his body, that’s a noble love, and better than the vulgar love of attraction. 

German philosopher Arthur Schopenhauer wrote in his 19th-century essay “The Metaphysics of Sexual Love” that he distrusted the feeling of being desperately in love. He considered romantic love to be a biological drive disguised as a real emotion — one whose goal is to trick us into procreating. You could even see your lack of attraction as an indication that you’re able to act more rationally, and not be deceived by that cunning trickster, passion. 

I know you’ve recently gone through a breakup, but I hope you can feel how bleak these views can become: to see love only as the window dressing for the propagation of a species, or as just a bridge to the love of wisdom. 

This cynicism has other consequences. Wondering if you could redefine love in a relationship with your friend isn’t like a wine-fueled dinner party bouncing around thought experiments. It involves your best friend and his own feelings too, and could risk not only hurting him, but also robbing him of a better romantic partner. 

For Simone de Beauvoir, the 20th-century existential philosopher, the most important aspect of life was freedom: We are technically free to do whatever we want, and it’s the choices we make with that freedom that determine our authenticity, or lack of it. De Beauvoir worried about forms of love that stole freedom from other people. In narcissistic love, she wrote in The Second Sex, a partner doesn’t love the other person, but loves the fact that the other loves them. It’s a relationship built on the self-interest of one of its members. 

I want to gently point out that it sounds like there may be aspects of the narcissistic relationship in your dilemma. While it may be intriguing to consider what it’s like to date a friend, you don’t seem to be as romantically interested so much as you are interested in his interest in and support of you. Actually involving yourself in such a dynamic could take his freedom, as well as your own capacity to learn from your break-up, then find a partner you’re genuinely drawn to for their own merits.  

I’m not saying friendship isn’t an important ingredient in love. Many of the most successful, very long-term partners I know have developed a deep friendship with one another. This friendship helps them stay kind to each other and have fun over many years when work, health, or stress can interrupt more passionate times of connection. But those same couples also had the undeniable spark of attraction at the start of their relationships too. They didn’t have to try. Sometimes they tried not to feel as strongly as they did, and still weren’t able to stop. 

Even in cases when a relationship first launches platonically, if something more is there, the undeniable, reciprocal pull tends to emerge soon thereafter. In H.G. Wells’ 1911 novel, The New Machiavelli, which was largely based on a real-life affair Wells had, he writes about a college graduate called Richard who meets a young woman, Isabel. The two become friends and are entranced with each other’s minds. But they quickly start to notice more between them, despite trying to push such thoughts away.

They fail at this attempt. Richard has a moment of realization, when he turns to look at Isabel’s face one day. “And something — an infinite tenderness, stabbed me,” Wells wrote. “It was a keen physical feeling, like nothing I had ever felt before. …From that time forth I knew I loved Isabel beyond measure.” I think you owe it to yourself, and your friend, to pursue connections that have this quality, especially at the start. 

You should also make room for and appreciate your friendship. If it never becomes romantic, that doesn’t make it a lesser relationship. The couples therapist, Esther Perel, calls friendships their own kind of love story. She once described a simple exercise to help appreciate that bond. Take a few minutes and think about a friend who isn’t a family member or romantic partner. “Recall a time when that friend reflected you back to yourself, created space for joy, or helped you remember who you are,” Perel suggests. Then, send a message to that friend about that moment, and thank them. 

Having a good friend by your side in the aftermath of a breakup is a gift. Work on maintaining and nurturing that friendship, and make sure you spend some time helping him with his life too. Don’t fixate on whether you should attempt to fall in love. The truth is, for better or for worse, if you love him, you won’t have to try.

Bonus: What I’m reading

• I’ve always been interested in the history of love, by which I mean the differences over time in how love has been expressed and discussed. William Reddy, a historian of emotion, has an excellent book called The Making of Romantic Love; you can also check out a syllabus for a class at Duke University he taught on the subject. 
• For a deeper dive (and refresher!) on Plato’s Symposium, I found that there was no better person to turn to than Melvyn Bragg and his guests on the BBC In Our Time episode on the philosophy of love. 
• Arranged marriages are an interesting window into understanding the growth of love. Mansi Choksi’s nonfiction book, The Newlyweds, follows three couples in India who attempt to marry for love; here’s an excerpt that was published in Harper’s Magazine. 

This story was originally published in The Highlight. To get access to member-exclusive stories like this every month, become a Vox Member today.

There is a kind of terror-tinged embarrassment that comes from dozing off on your beach towel on the sand, only to find a rising tide intrude upon your mid-nap bliss, lapping at your toes and sweeping away your sunglasses.

It is a truth I relearn every June, one that arrives like the season’s first sunburn: The ocean is inevitable, and Poseidon always reclaims his own. Actor Nicolas Cage learned a far more serious — or more expensive, certainly — version of that lesson earlier this month, when a furious high surf siphoned the ground beneath his Malibu, California, mansion, collapsing his driveway and front porch into a gigantic 30-foot-by-30-foot sinkhole. The cave-in ruptured a gas main, swallowed an access road, and forced dozens of locals to evacuate. In the process, it put into stark, cinematic relief the unyielding threat of coastal erosion in every community in the nation with a view of the sea.

It’s impossible not to imagine Cage standing on the edge of that yawning pit as he might in one of his films, bombastic in his bafflement, every outsize gesture a pantomime of confusion. But no one should have been surprised. As far back as 1976, the California Coastal Commission denied a proposal to build a home on the lot where Cage’s Malibu home now teeters over the void, citing “concerns over wave hazards.” In 1986, the same commission rejected another residential proposal on the land, deeming it an “extremely hazardous area” threatened by waves and erosion. 

So why, just a few years later, was a home built there anyway, one that would eventually come to be owned by a certain Hollywood eccentric? A few years after their initial rejections, the waterfront lot’s original owners were able to convince the commission that with enough clever engineering, including a sea wall and concrete stilts, the home, its septic system, driveway, and access road could be protected from the sea god’s wrath. But much like Odysseus before him, Cage has found the hard way that the lord of waves does not easily forget his due. “What is the Grail but Earth itself?” he once told the New York Times, characteristically mythologizing his real estate acquisitions. “The divine object is Earth.”

In this case, at least, Cage chose the wrong cup. But Cage is hardly alone in laying out his home atop the wrong stretch of sand. Nearly all of the nation’s coastal communities — from ritzy enclaves like Malibu to working-class quarters of the Rockaways in New York — rely on coastal engineering to remain inhabitable, dumping sand or erecting sea walls to put off a perpetually and increasingly rising tide. 

To put this into perspective, almost one-third of Americans live along a coastline, and while few of us have the kind of luxury oceanfront views that Cage and his neighbors enjoy, our beachside apartment complexes, utility lines, and coastal highways suffer from the same risks his driveway does. Most of us, however, lack the benefit of a second home or a private island to flee to if something goes catastrophically wrong. 

When a sinkhole opens underneath a celebrity’s driveway, or a stilted home collapses into the sea, it is a reminder that we have built huge swaths of our beloved coastal communities on land that was never meant to sit still. If we want them to survive, we will need to learn to embrace new ways of packing our beach towns up and wheeling them further upshore. Logistically implausible as it may sound — difficult as, say, stealing the Declaration of Independence — it’s a process that several American communities have become much better at in recent years. 

“This is not a problem that is unique to Malibu, which just so happens to be some of the most spectacularly beautiful and spectacularly expensive real estate in America,” said Kara Murphy Schlichting, an environmental historian at Queens College in NYC. “This is a question that is coming for all of America’s coastal communities.”

Say it ain’t sinkhole

The void that swallowed Cage’s driveway with the gusto of the actor chewing scenery in one of his films was no ordinary sinkhole. In fact, if you asked a geologist, they would tell you that it was not, technically speaking, a true sinkhole at all. 

“It’s not what I typically would consider a sinkhole,” Michael Sukop, a hydrogeology professor at Florida International University, told me. That’s because most sinkholes form when rainwater or a leaky pipe slowly seeps into the porous bedrock beneath our highways, homes, and soccer fields, creating a cavern that screeches open during a heavy downpour or when the infrastructure above it becomes too heavy to bear. 

There is no accurate accounting of every sinkhole in the US, but geologists widely believe that they have become more frequent globally, driven by extreme weather linked to climate change, aging infrastructure, and what Sukop calls “people moving into sinkhole country,” meaning places like Florida where the bedrock is highly soluble. Some of the worst places for sinkholes in the country are in landlocked states like Missouri, whose unique landscape boasts upward of 16,000 sinkholes, which often open after heavy rainfall.   

But Cage’s sinkhole was formed not from hidden cavities in the bedrock, but from a high surf strong enough to slip under his sea wall and wash away the soggy soil beneath his asphalt and, presumably, his Rolls-Royce Phantoms. It was spurred not by rainfall or a leaky pipe but by a tidal force akin to the one that can put a damper on — or literally dampen, anyway — a beach spot that seemed perfectly sandy when the surf was low. 

@accuweather

A homeowner on North Carolina’s Outer Banks paid to move their beachfront home away from the shore to prevent it from collapsing into the ocean, as several neighboring homes have.

♬ original sound – AccuWeather

But while I can learn to lug my sodden towel and cooler a little further upshore, the same cannot easily be said for Cage’s $10.5 million abode, built — like much of Malibu — atop a constantly shifting coastline. Which isn’t to say that it wouldn’t be a good idea to try.

“It’s a fiction that we can keep a coast in a single place,” said Schlichting. And it is a fiction upheld by hundreds of millions of dollars spent each year on temporary solutions meant to hold the ocean still, and protect the luxury homes and boardwalks we’ve built along it. 

The pristine, sandy beaches that most of us think of when we think of the word “beach” are often heavily manufactured for our pleasure, much like an urban park might be seeded with carefully cultivated trees and brush. In fact, the federal government spends hundreds of millions of dollars each year to dump sand atop eroded coastlines, a process known as beach renourishment, in addition to many more billions spent on sea walls and other structures that can keep a coastline in place for a little while. 

“The most endangered species on the coast is not a turtle or a piping plover; it’s an unengineered beach,” said Robert Young, director of the Program for the Study of Developed Shorelines, who has been tracking beach renourishment projects since the 1980s. “And it’s unsustainable because as sea level continues to rise, we are having to pump that sand more and more frequently.”

How to save coastal neighborhoods

Earlier this year, Malibu released a long-awaited Coastal Vulnerability Assessment that twice called out the acute dangers that erosion poses for Sea Level Drive, a small stretch of private road, upon which a gaping crater now sits where Nicolas Cage’s driveway used to be. As this year’s goliath El Niño gets into full swing, his misfortune is unlikely to be the city’s last. 

Already, the spectacular seafront Pacific Coast Highway faces a gauntlet of exigencies, among them landslides, wildfires, and, of course, a coastline that simply won’t sit still, creeping further and further inland each year. Just this weekend, officials in Malibu issued a warning about hazardous conditions at local beaches, with “severe beach erosion” exposing steel and concrete debris sticking across the coastline.  

“Nature itself does not destroy the coast. It moves the coast,” said Young, but if infrastructure like the Pacific Coast Highway lies in the way of that movement, then there is only so much that building a sea wall or dumping 300,000 cubic yards of sand every five years will do. “No matter where you are, whether you are in California or South Carolina or New Jersey, we are still building in dumb places.”

Much like laying a log in front of your beach towel won’t hold the tide back for very long, these measures are often only a temporary balm. But there is another way, one that involves moving the proverbial towel — or permanent infrastructure — away from the shore, a process known as managed retreat.

As implausible as it may sound, when you take into account just how much money has gone into keeping the coast still, lugging a few celebrity mansions, a beachside school, or a sliver of highway across town might be a Herculean labor worth attempting. 

“It’s totally doable,” said Young. “We can move just about everything.” On the coast of North Carolina, the Cape Hatteras Lighthouse, one of the tallest brick lighthouses in the world, was moved in 1999, a feat that Young helped pull off. At the time, he said, locals  “wanted to shoot us over” the effort, sure that the centuries-old structure would fall down. “And you know what? It’s safe now,” he said, noting that just down the beach in Buxton and in nearby Rodanthe, dozens of homes have collapsed into the ocean since 2020.

But even in places where such coastal relocations were once commonplace, it’s still not an easy sell for most coastal policymakers and their constituents, especially in places like Malibu, whose density and geography make it all but impossible to move things nearby. But sooner or later, they might not have a choice. 

“I can’t tell you how many coastal elected officials have said to me, ‘Rob, you know, retreat is not an option from this place.’ And I’m like, well, actually, you are retreating here, here, here, and here,” Young told me. “It’s just you’re doing it in the worst way possible. You’ve got septic tanks rolling around in the sea. That is not a great Chamber of Commerce moment.”

If there’s one thing that Americans can unite behind, it’s anxieties over artificial intelligence.

According to a recent NBC News/Decision Desk poll, 70 percent of adults say they are personally more worried than excited about the technology; it’s basically underwater in every area of society with one notable exception: its use in scientific research and medicine.

According to Dr. Dhruv Khullar, a physician at Weill Cornell Medicine and writer at the New Yorker, that optimism is well-founded.

Khullar uses AI tools in his own practice as a physician and writes about their advancements in the medical field. He says that “right now, at least, there’s a generally optimistic view of how AI will change healthcare going forward.”

AI in medicine is nothing new. In 2024, the Nobel Prize in Chemistry was given to researchers who developed an AI model that enabled breakthroughs in predicting protein’s complex structures. As it becomes more ubiquitous, it’s also emerging as a resource for doctors to get second opinions and as a tool to take notes on patient-doctor interactions.

On top of that, Khullar says, it might help unlock new medicines to treat diseases. 

Khullar joined Today, Explained co-host Sean Rameswaram to discuss the ways that AI is transforming drug discovery and research, as well as what it still can’t do. They also discuss how to think about AI’s potential in healthcare at the same time as top AI researchers are increasingly warning about the technology’s risk.

Below is an excerpt of the conversation, edited for length and clarity. There’s much more in the full podcast, so listen to Today, Explained wherever you get podcasts, including Apple Podcasts, Pandora, and Spotify.

You’ve been thinking about lifesaving AI innovation. Has AI already started saving lives?

I think it has. This is the fastest I’ve ever seen healthcare take up a new technology. Part of that might be that the healthcare system is so messed up that there’s an appetite for some type of change. It’s unaffordable, it’s inaccessible, it’s inconvenient, the quality is uneven. 

There’s at least a hope that AI is going to help with all of these things, and I think it’s already starting to make its way into the healthcare system. Now it’s our responsibility to figure out how to make the most of it without also promulgating some of the downsides.

For those who are unaware, tell us how it’s working its way into the system.

It’s helpful to categorize it into a few buckets, at least for healthcare. 

The most rapid uptake, and I think the place that it’s already starting to make a difference, is on the administrative side. Many people who have gone to a doctor recently might have noticed that there’s an AI scribe that’s taking notes now. And people seem to really like it, at least so far. Doctors are able to look their patients in the eye in the way that they weren’t able to when they were poring over their computers and just typing what the patient was telling them. 

The second big area is patient navigation. When is your next scan? When’s your next appointment? Do I need to take this medication on an empty stomach or not? There’s a huge opportunity for people, let’s say, who are diagnosed with a serious illness like cancer or heart failure to navigate the system more seamlessly. 

The third big area that I’m really excited about is drug discovery. AI is making a huge dent in the early parts of drug discovery. 

The last thing that I’m excited about and that I use every day when I’m in the hospital is [AI] as a second opinion. Now instead of having to get a consult or turn to a textbook, AI can be a very, very helpful clinical decision support. It’s not to say that I never consult someone, of course, but that first pass — I’m not sure what’s going on. What are some recent trials that might influence my decision here? Is there something I’m missing? Is there a test that I should be ordering that I’m not ordering? — all those types of things, they start to help with.

The biggest chunk of this positive view of AI and medicine is drug discovery. Tell me more about how that’s going so far and how it might be going in a year, or five, or 10.

I don’t want to leave this conversation and say that we’ve solved the cure for cancer because of AI, but I do want people to know that at least for the early stages of drug discovery — figuring out the very basic steps of is this an interesting molecule? Does it have potential biological applications? — AI is already very helpful for that. 

They’re trying to take, basically, what is an infinite number of drug-like molecules that are theoretically possible to be drugs and they’re trying to match that to some disease that’s going on inside a human. And human biology is incredibly complex as well. There’s tens of thousands of genes, hundreds of thousands of proteins, trillions of cells, and you have to match a potential molecule to a potential target within a human. 

In the past, a researcher might take years to study a disease or a biological pathway and figure out, okay, this protein seems to be involved. And actually, it might not be involved, it might be involved, it might be disrupted, but not be the causative agent. There’s a lot of uncertainty there, so now AI can take enormous amounts of data, and it can rank various potential targets that seem to be the most likely to be causing a particular disease and give you that short short list of potential targets that may have taken months or years in the past.

So now you know what to attack, but you have to figure out what to attack it with. So you need a molecule, you need to generate a molecule that’s going to fit into that target or disrupt that target in some way. 

Now again, AI can scour vast chemical spaces and give you a list of the things that seem to be most likely to be able to disrupt that target. And then it can actually help you generate that molecule in some cases. 

So now we have the target. Now we have something that we’re going to attack the target with. And then you have to do something called lead optimization, and that’s the third thing. Lead optimization means you have this lead and you have to optimize it.

“I would make the counterargument that healthcare is ripe for disruption and you don’t even need the frontier frontier models to make it a lot better with AI.”

Now we have to figure out, of all the molecules that seem to be good at potentially affecting this target, which ones are able to be absorbed in the body, which ones are going to get to the right tissue. AI models can predict the most likely molecule to give you that Goldilocks set of properties that’s needed for a safe and effective drug.

As we all know from anyone who’s asked AI to write an email for them or Googled a question, these tools have a certain degree of certitude, even though these tools can just be flat out wrong about stuff. Are you guys worried about that in the field?

Absolutely. And so this is why I think the narrative around AI just replacing scientists or doctors or other workers is incorrect. Because you still need a lot of judgment. 

You need to be able to adjudicate the output of these models to figure out what is most promising and what is potentially dangerous. That is going to require wet labs and scientists and reasoning based on prior experience and understanding the context. There’s all sorts of issues around, can you manufacture some of the drugs that are being proposed by these models? So just because it’s set, it dreams something up doesn’t mean you can actually make that thing in the real world.

Some of the drugs that it proposes might actually be toxic in certain ways that it didn’t predict. And of course, then you have to take this thing into clinical trials. You have to recruit l people who are willing to put this medication in their bodies. You have to find out who might benefit, you have to put it through the regulatory process. 

That’s why I said that the first part of drug discovery, where you’re trying to figure out how do we get the right drug candidates, what seems most exciting to test further, that’s going to be really accelerated. That whole second half where we actually have to figure out if it works in human biology, that is still going to be in some respects an analog process.

It doesn’t sound like you’re terribly scared that we are like ceding control of our hospitals, of our research facilities, to AI. This is very much in the sort of assistant bucket?

For now it is. And I think one of the challenges is to maintain our agency as these models become more and more sophisticated. As you lean more on these machines, if you’re a doctor, inevitably some of the skills, the critical thinking, the reasoning that we put into coming up with the diagnosis that we honed over the course of years, that can start to atrophy. 

These are the types of things that we still need to sort through as we’re implementing more and more AI into the healthcare system.

For the people who just say, shut it all down, this is too dangerous, the risk is far too great, we don’t actually need this, this isn’t doing anything for society, would you make a counterargument?

I would make the counterargument that healthcare is ripe for disruption and you don’t even need the frontier frontier models to make it a lot better with AI.

Even if we wanted to “slow the pace of the frontier,” fine. But there are models from a year or two years ago that could be very helpful in the healthcare setting. When we’re talking about slowing the pace of the frontier and the most sophisticated and potentially dangerous models, fine. 

But if we’re talking about shutting down AI and not using it in biotechnology or not using it in scientific research or not using it in clinical care delivery, that’s where I would push back pretty hard.

This story appeared in Today, Explained, a daily newsletter that helps you understand the most compelling news and stories of the day. Subscribe here.

Four beluga whales have died in a month at one of America’s top-rated aquariums — a tragic development and, arguably, a reason to revisit the ethics of zoos. (More on that in just a minute.)

The latest cetacean casualty, a 26-year-old female named Osiris, died at Chicago’s Shedd Aquarium yesterday. She follows 34-year-old Beethoven, a longtime resident of the aquarium, as well as 20-year-old Rain and 23-year-old Peekachu, who died in late August and early September.   

Osiris, Rain and Peekachu were all part of the Marineland diaspora: a group of four dolphins and 30 beluga whales that rescuers have worked to relocate since the 2024 closure of a scandal-plagued theme park in Niagara Falls, Canada. This summer, the animals were successfully transferred to aquariums and marine parks in the US and Spain. That was a frankly incredible undertaking involving padded shipping containers, a police-escorted convoy and a chartered plane. 

But in some regards, the international flight wasn’t even the hardest part of the whales’ journeys. Some of the animals have struggled to thrive in their new environments — even when those environments are state-of-the-art, like Shedd. 

In a blog post announcing Osiris’s death, the aquarium noted that its rescued whales had been under 24-hour supervision and receiving intensive, specialized care. The aquarium is still investigating the cause of death for all four whales. 

Go a little deeper

Believe me when I tell you that it gives me no pleasure to write the following paragraphs. I really love aquariums. In fact, as a kid, I grew up visiting the now-shuttered and much-maligned Marineland. 

But incidents like these back-to-back deaths should prompt us all to ask ourselves who aquariums and zoos are actually for. And the answer, I regret to tell you, is frequently not “the animals” — though zoos and aquariums have, since the mid-20th century, marketed themselves as both tourist attractions and important players in wildlife education and conservation.  

As my colleague Kenny Torrella writes, these institutions and their defenders argue they protect animals in three ways: first, by raising money for wildlife conservation projects; second, by helping to reintroduce vulnerable species back into the wild; and third, by inspiring visitors to take an interest in the environment. The Shedd Aquarium’s motto, for instance, is “For Love of Waterkind.” 

It’s worth scrutinizing those purported virtues, however — even when prestigious and reputable institutions like Shedd are making them. On one hand, there is merit to the claim that zoos and aquariums fundraise for wildlife conservation: In 2022, the latest year for which data is available, one group of accredited zoos and aquariums spent more than a quarter of a billion dollars on habitat projects. 

The evidence is thinner on the other points, though. It’s incredibly rare for aquariums and zoos to reintroduce endangered species back into the wild — that’s very difficult to do. And surveys of zoo and aquarium visitors show that merely seeing a wild animal in a tourist setting doesn’t generally galvanize people on matters of welfare or stewardship. Living in captivity in that tourist setting, however, can be stressful (or worse) for the animals themselves. 

But what would a humane alternative look like? Even if zoos and aquariums stopped catching animals in the wild, that would still leave many thousands who are injured, orphaned, born in captivity, or otherwise accustomed to human care. 

When it comes to whales, one nonprofit advocacy organization is already working on an alternative: a 110-acre coastal sanctuary for “retired” whales. That project, however, is still a long way from completion. And in the meantime, it’s up to consumers to decide whether the stated benefits of zoos and aquariums outweigh the potential harms to the animals they exhibit.

One link for later

➨ Beware the trap of “moral perfection.” We should all endeavor to live lives that embody our values, sure. But if you never board a plane because you care about the climate, or you never take your kid to zoos because you read the arguments above, then you might end up with a smaller, less-lived life. And that’s no good, either. 

Before you go…

• Did you know…that there have been “revolutionary” advancements in Alzheimer’s treatments in just the past five years? Two new drugs both markedly slow the progress of dementia.
• Today’s trivia: Geography time! What country is bordered by Chad to the east and Mali to the west? Hint: It is the second-largest landlocked African nation. (You can find this and other brain puzzles in Vox’s daily crossword. Look for the answer in tomorrow’s edition.)
• Yesterday’s trivia: Yesterday we asked you for the classic Broadway star who first sang hits, including “There’s No Business Like Show Business.” That would be Ethel Merman, whom one biographer dubbed “the most successful musical comedy performer of her generation.” 

Who gets to tell the military “no”?

That question has become even more important after the Pentagon pushed AI companies to loosen limits on how their technology may be used in war — including how much human control must remain over autonomous weapons. When Anthropic said no, it lost a major government deal. OpenAI stepped in, embracing the Pentagon’s more flexible standard instead. Now, thanks to new reporting from The Intercept, we’re getting a better look at what those negotiations actually entailed.

The OpenAI-Pentagon agreement included specific language on the type of AI models the DOD would receive: “OpenAI models that are designed for national security use cases and have minimal refusal rates.” That phrase — “minimal refusal rates” — means that the Pentagon is seeking models that rarely, if ever, tell it no. (According to The Intercept, a Justice Department lawyer representing the Pentagon confirmed this language, before backtracking only a few hours later.)

On the surface, that provision might seem reasonable. The military is a strictly hierarchical institution. When your senior gives you an order, you do it. Leaders generally have more battlefield wisdom and experience than their subordinates, so the rank and file can usually assume that their commanders know what they’re doing and follow along. Plus, in the high-stress, time-sensitive environments of most military operations, there is little room for discussion, explanation, or alternative proposals. 

So, if soldiers can’t refuse an order, why would a machine be allowed to? Who is a robot to tell its commander no?

There’s just one issue. Soldiers can refuse an order — when the order is illegal. In fact, they are legally required to. But by introducing technologies with a limited ability to challenge commands, the military may be making it harder to stop war crimes before they are committed.  

Humans have a duty to disobey

As outlined in US domestic, military, and international law, soldiers are only obligated to follow lawful orders, and they have a duty to refuse unlawful ones. 

The Department of Defense’s Law of War Manual is unambiguous: “Each member of the armed services has a duty to: (1) comply with the law of war in good faith; and (2) refuse to comply with clearly illegal orders to commit violations of the law of war.” Obeying an illegal order — as in a directive that violates the Constitution, US laws, international law, or directs one to commit a crime — can expose a service member, as well as their commander, to criminal liability. 

This “duty to disobey” has foundations in early British and American common law, but it achieved international recognition after the Nuremberg Trials against Nazi soldiers after WWII. 

In those trials, many Nazi soldiers argued they were not liable for crimes they committed during the Holocaust because they were simply following their superiors’ orders. Rudolf Hoess, the commander at the largest Nazi concentration camp, Auschwitz, himself explained: “In Germany it was understood that if something went wrong, then the man who gave the orders was responsible. So I didn’t think that I would ever have to answer for it myself.”

These defenses were unequivocally rejected by the court.

The tribunal held that obedience is not without limits, especially when a wartime act would be clearly illegal to the subordinate or any person of “ordinary sense and understanding.” If “I was just following orders” was an absolute defense, everyone involved in a crime would escape accountability, with the exception of the single highest-ranking officer. Instead, the court determined that individual moral and legal responsibility supersedes national or military chains of command for illegal acts.

Obedience is not without limits, especially when a wartime act would be clearly illegal to the subordinate or any person of “ordinary sense and understanding.”

But, when it comes to agentic AI systems or AI-assisted weapons, increasingly being used in place of human soldiers, does the same logic apply? And, perhaps more importantly, should it?

…But do machines?

When you use an AI chatbot, it does not simply reject queries it may not “agree” with. It does not say no to requests it doesn’t “like.” Instead, it follows a set of established protocols, much as a soldier should follow the law of war: It refuses requests that violate its terms of service. 

The military is using AI for far more consequential applications than a layperson using ChatGPT to, say, build a travel itinerary. So, what kind of world is created by the Pentagon’s request for AI that “minimally refuses” its directives, many of which may bear grave ethical, moral, or legal implications?

Rebecca Crootof, an expert on technology law and professor at the University of Richmond School of Law, told Vox that while the language could easily be read as increasing the likelihood of compliance with unlawful orders, or at least reducing the friction around them, “Minimal refusal doesn’t mean no refusal.”

“I both want to say it’s not definitely going to mean compliance with unlawful orders,” Crootof said, “but also acknowledge that identifying and evaluating what constitutes an unlawful order in the moment can be incredibly difficult.”

Even for human soldiers, disobeying illegal orders is neither simple nor easy. To the contrary, it doesn’t even happen that frequently. In fact, some legal scholars have argued that giving illegal directives in and of itself is an act of abuse against subordinates, who are not thoroughly trained in legal nuances, are conditioned to obey, and face prosecution if they make the wrong choice. 

Still, the mere fact that soldiers can disobey is an important constraint on commanders who might let battlefield conditions, or even commands coming from their own superiors, erode their ethics and commitment to the law.

Still, the mere fact that soldiers can disobey is an important constraint on commanders who might let battlefield conditions, or even commands coming from their own superiors, erode their ethics and commitment to the law.

Those who are optimistic about AI often argue that machines could be better at interpreting and applying the law than human soldiers. Machines can be trained on the law itself — drawing on a repository of legal cases, rulings, and battlefield data — which might make them more “knowledgeable” than an average soldier. And machines do not face the same pressures as soldiers facing prosecution for making mistakes, since they can’t exactly be held accountable; a problem to unpack another time. That, in turn, might make it easier for them to refuse illegal orders in the proper conditions. 

So, is the answer as simple as letting the machines decide? Not exactly. “This has been a desire and a drive by technologists for ages,” Crootof said. “What if we just put the law in the system and have it only do things that are lawful? Isn’t that going to solve things?” 

Interpreting and applying the law, though, is not a simple legal-illegal binary. “It’s very context-specific,” Crootof told me, “and honestly, that’s the kind of thing humans generally find difficult and AI decision-making systems are even worse at.” Often, determining the legality of an act comes down to subtle nuances you can’t program for: You’re allowed to target a combatant; you’re not allowed to target a wounded or surrendering combatant. You’re not allowed to target a civilian; you are allowed to target a civilian directly participating in hostilities. The difference between a legal target and an illegal one may come down to body language, clothing choices, or even battlefield intuition. 

As Crootof explained, “These are not the kinds of contextual analyses AI is good for.”

So what’s the solution?

Where human soldiers’ power to disobey would traditionally serve as an important check on the power of military commanders, if an imperfect one, replacing soldiers with machines that don’t have such safeguards removes one last layer of judgment over the use of force. And yet, machines themselves are not always capable of making the judgment calls that go into interpreting the law, so it may not be appropriate to let AI have the final say either.

So what’s the solution? Given the way AI technology has already been thoroughly integrated into the military, Crootof told me it may be somewhere in the middle. 

Rather than programming AI to refuse commands it interprets as illegal, Crootof said, “what it could be designed to do is flag those indeterminate zones, those gray-zone analyses, for human review or for a higher authority.” When a system picks up conflicting signals — an enemy soldier who looks like he may be surrendering, for example — it could prompt the commander to confirm their selection or remind them of applicable law. 

In this way, a check on commanders’ battlefield authority could be preserved without fully ceding judgment over the use of force to the machine. 

We don’t know yet whether OpenAI has equipped its “minimal refusal” models with such capabilities. But as those technologies become increasingly central to military decision-making, the public deserves answers sooner rather than later. 

Editor’s note, September 20, 8 am ET: We’re bringing you some of our best-loved Your Mileage May Vary columns while Sigal Samuel is on parental leave. The one below was originally published last September.

This unconventional advice column offers you a unique framework for thinking through moral dilemmas. It’s based on value pluralism: the idea that each of us has multiple values that are equally valid but that often conflict with each other. Submit your own question here.

---

The Jewish High Holidays are coming up, and that includes Yom Kippur. It’s a holiday that encourages people to reflect on their behavior and make amends. That’s all well and good, but I’m someone who struggles with scrupulosity — constantly worrying about my morality and if I’m doing The Most Possible Good™.

In practice this is far more paralyzing than motivating. Fixating on the ethical implications of all possible decisions makes it harder to take any action, and I’ve lost hours scouring my memories of my past behavior for immorality like a football player watching footage of their games to analyze what they could do differently. Guilt simply isn’t serving me, but I worry that saying to hell with all that means I’ll stop striving to be a better person and become morally complacent.

I’ve observed Yom Kippur for decades, and don’t want to simply avoid the day. But the holiday is a moral scrupulosity trigger. How do you think I should approach this? I want to stop feeling guilty for letting guilt get in my own way.

Dear Scrupulous, 

Have you ever heard the story about what happened when God decided to give the Bible to flesh-and-blood human beings? According to the ancient rabbis, the angels hated the idea. They argued that humans were deeply flawed mortals who didn’t deserve such a holy scripture; only angels could be worthy of it, so it should stay up in heaven. 

It fell to Moses to rebut the angels’ argument. He asked them: What do you angels need the Bible for? The Bible says not to murder, not to commit adultery, not to steal — do you have jealousy or other emotions that could lead you to do those things? The Bible says to honor your father and mother — but you don’t have parents, so how could you ever do that? And the Bible says to sanctify the Sabbath — but you never do any work, so how could you even honor the Sabbath by resting? 

The angels saw that Moses was right. Angels are really great at one thing: being perfect. But perfect creatures are static. They don’t experience painful challenges, they don’t grow, and they don’t make choices that add beauty to the world. We humans do those things. God gives the Bible to humans not to make them into angels — but to make them better at being the unique thing they are: human animals, with feelings, flaws, and all, that can learn to use their capabilities in more beautiful ways.

I think there’s a lot of wisdom in this story. And I want you to notice how far it is from demanding that humans do “The Most Possible Good™.” That language suggests a maximizing ethical theory like utilitarianism, which says that we have to do the action that produces the greatest good for the greatest number. In other words, we have to optimize. 

Moral optimization may be possible for angels, but not for humans. We each hold multiple values, and sometimes those values are in tension with each other, forcing us to strike a balance as best we can. We’re not omniscient beings who can know with certainty how best to strike the balance. 

What’s more, sometimes different kinds of moral good straight-up conflict with each other. Think of a woman who faces a trade-off: She wants to become a nun but also wants to become a mother. She can’t balance between those options — she has to choose. And what’s the better choice? We can’t say because the options are incommensurable. There’s no single yardstick by which to measure them so we can’t compare them to find out which is greater.

Given that this complexity is baked into the human condition, it’s impossible to be a perfect optimizing machine. And the more you try to force yourself to be that, the harder it’ll be for you to actually help others, because you’ll just be so burned out. As you’ve already discovered, the optimizing mindset is exhausting — you end up expending a lot of precious mental resources that you could be spending on something else. It can even lead to paralysis. And a lot of the time, there’s no knowably best decision.

So instead of trying to optimize everything, you can adopt a goal that’s humbler but more realistic: to live in line with your values as best you can. 

I know that can feel scary. Optimizing makes being human feel less risky. It provides a sense of control, and therefore a sense of safety. The unspoken premise is that if you optimize, you’ll never have to ask yourself: How could I screw up that badly? 

But there’s another way to feel safe. It’s about leaning into the fact that we are imperfect and vulnerable creatures and that even when we’re trying our hardest there will be some things that we do not do optimally.

Of course, we should still try to live in line with our values. But in those moments when we fall short, we shouldn’t berate ourselves, thinking, “I sinned!” In Hebrew, the word we typically translate as “to sin” (lachto) actually means “to miss the mark.” It’s the same word we’d use to describe someone with a bow and arrow who’s targeting the bullseye but misses it slightly. This is a useful image, because it reminds us just how normal it is to miss the mark. Just as the archer’s arrow is buffeted around by the wind, we’re buffeted around by all the physical and psychological conditions acting upon us — naturally we won’t always hit the bullseye! And when we do miss the mark, we deserve compassion.

I know what you’re thinking: What if adopting this mindset means you’ll become morally complacent and let yourself off the hook too easily? This is one of the most common objections to practicing self-compassion. But research shows it’s not well-founded. In fact, psychologists have found that more self-compassionate people are better able to acknowledge when they’ve made a mistake. They’re more likely to want to apologize and make amends to others when they mess up. And they try to do better the next time around. Why? Because, to them, mistakes don’t feel so psychologically damning. That allows them to take more — not less — responsibility for their actions.

Yom Kippur can feel terrifying when any mistake you’ve made over the past year seems damning. But according to the ancient rabbis, Yom Kippur is not meant to be a somber day — it’s one of the happiest days of the year! After all, it was on Yom Kippur that Moses descended from Mount Sinai carrying the second set of tablets inscribed with the Ten Commandments, ready to gift them to the people. 

You probably know what happened to the first set of tablets: Moses shattered them after seeing the Israelites engaged in idol worship. What’s less known is that, according to one rabbinic story, God’s response to the shattering of those tablets was to actually congratulate Moses. Why did God think breaking them was the right move? And what was different about the second set? 

While the first tablets had been fashioned by God and God alone, the second tablets were a human-divine collaboration: Moses carved the stone and God inscribed the words. And while the first tablets contained only the words of the Ten Commandments — a black-and-white, rule-based morality — the rabbis tell us that the second tablets contained within them all the stories and interpretations that Jewish sages would later develop.

In other words, God recognized that you can’t just give humans a list of moral rules and call it a day. Maybe that would work for angels, who live in a simplified, disembodied world, but our ethical life is just too messy and multifaceted to be captured by any single set of universally binding moral principles. Yet God chose people over angels anyway, inviting us into the collaboration and embracing our humanness rather than rejecting it.

So, to the rabbis, Yom Kippur was a happy day because they fully expected that God would accept and embrace messy humans. 

Please, don’t try to be more zealous than God.

When you’re taking action, by all means, aim your arrows as true as you can — try to hit the bullseye, the place that captures as much of what you value as possible. But once you’ve released the arrow from your bow, let it be. 

If it turns out that you missed the mark, that you acted suboptimally, I put it to you that that is okay. You are not an angel. You are not a perfect optimizing machine. You do not have access to a magical mathematical formula that can consider countless incommensurable variables and spit out the very best move with certainty. You are human and you do the best you can with what you’ve got. 

The wisdom of these millennia-old stories is that that’s good enough for God. Let it be good enough for you, too. 

Bonus: What I’m reading

• Writing this column reminded me of The Mindful Self-Compassion Workbook, written by the psychologists Kristin Neff and Chris Germer. It really helped me develop a self-compassion practice, which has in turn helped me get a grip on my own scrupulosity. I also strongly recommend the eight-week self-compassion course run out of Neff and Germer’s nonprofit, the Center for Mindful Self-Compassion.

• I’ve always associated the philosopher Thomas Nagel with questions about consciousness, but I recently learned that he’s also super interested in questions about religion. In a great essay called “Secular Philosophy and the Religious Temperament,” Nagel asks: What, if anything, does secular philosophy have to put in the place of religion? More specifically, can it answer the question: What is the underlying nature of the universe, and how can the human individual live in harmony with it? 

• In this Aeon essay, philosopher Elad Uzan argues that AI will not be able to solve ethics for us, despite what some people hope. Drawing on the mathematician Kurt Gödel’s famous incompleteness theorems, Uzan writes, “just as mathematics will always contain truths that lie beyond formal proof, morality will always contain complexities that defy algorithmic resolution.” 

Wildfires are popularly considered wild — restricted to the remote wilderness or the edge of forests, and to most, a problem for the dry and hot states of the Western US. But over the past few years, and now across the country, these fires are getting less wild and far more urban. 

This summer, fires in Spokane, Washington, demolished over 800 homes in a suburban neighborhood, and earlier in the season, the northeast coast was engulfed in smog from fires in Ontario. The Hawk Fire in Nevada razed more than 15,000 acres near Reno, destroying 47 homes before it was extinguished. The Los Angeles fires of last year remain some of the most devastating in recent memory, with over 37,000 acres burned — much of it densely built residential neighborhoods — and displacing 205,000 people from their homes. In Boulder, Colorado, the 2021 Marshall Fire destroyed nearly 1,000 homes. 

This changing geography of fires is affecting the health risks they present, too. 

When vegetation burns, it releases particulate matter, or PM 2.5, and with over 46 million homes in the continental US — this is approximately one-third of all homes — on the edge of fire-prone wildland, many more of us are being exposed to air filled with that soot. This alone is a major health risk. According to the World Health Organization, PM 2.5, is associated with premature death, severe damage to respiratory health, and cognitive impairment. It is especially dangerous for aging populations: a study published in February found that accumulated exposure to wildland smoke was associated with greater hospitalization risk for various cardiovascular diseases in older Americans, including irregular heart rhythms and heart failure, even when they are only exposed to moderate levels of PM 2.5.   

But the urban landscape has also changed the very composition of the smoke that increasingly hangs over all of us. As human structures and vehicles burn, the fires expose us not just to particulate matter but also to the chemical toxins of urbanization itself, from plastic to metal oxides and vehicle fuel. The changing nature of this smoke, in fire zones but also in surrounding regions as it drifts outward, exposes us to a new class of health risks. 

While researchers are still learning what exactly those risks might entail, early research suggests long-term dangers that are deeply troubling, especially as exposed populations age.   

The changing fire frontier 

In 2020, the worst California fire year on record, ate up 4.2 million acres of land, most of it wild, spewing huge amounts of PM 2.5 into the air as the blazes raged. Emergency rooms, already short-staffed by the ongoing pandemic, buckled under the burden immediately. Northern California’s Stanford Health Care system saw a 12 percent jump in hospital admissions over the weeks following the fires, and a 43 percent increase in cardiovascular conditions. In Oregon that same year, 10 percent of emergency room visits were due to asthma-like conditions induced by smoke as fires migrated north.  

Unfortunately, in a world of warming temperatures and drying conditions, that story is becoming all too common. In 2025, a total of 5.2 million acres of land burned across the US, and this year that number has already climbed to 8.5 million, before we’ve even hit the winter dry seasons that plague parts of the Southwest. 

Despite this, the global footprint of human housing continues to expand, and with it the danger.  

For centuries, human development has breached wildfire territory. However, over the last three decades, the zone that links the wild to cities and towns, called the wildland-urban interface (WUI), has grown substantially due to suburban construction. The US Forest Service found a 46 percent increase in the number of houses in the risky WUI zone between 1990 and 2020. By 2020, a third of US homes had been built in this highly vulnerable landscape. The math is simple: As suburbia expands into wilder regions, the more houses are exposed to wildland, and the fire risk grows greater. And when houses and other human-made goods burn, different and potentially more toxic forms of smoke can be produced. 

The Los Angeles County fires in January 2025, which saw large urban regions devastated by flames, were an unfortunate case study for these risks. The Palisades and Eaton fires began on the northern edge of the LA foothills, where dense housing meets more wild, and more fire-prone, terrain. In a retrospective analysis, Underwriters Laboratories’ Fire Safety Research Institute detailed how the urban-wildland makeup of Los Angeles County has not only put more communities physically in fire-prone zones, but also has historically made the geographic region itself more vulnerable to fires. 

As those fires spread into LA’s more urban areas in 2025, the smoke they produced became more dangerous, largely because of what was burning: more than 16,000 buildings, household items, and electric vehicles. The resulting emissions were thus far more complex than typical vegetation-only fires. 

“What’s burning is now very different,” said Jackie Goodrich, an epidemiology professor at the University of Michigan. “So now modern homes have tons of plastic, metals, and all of our electronic appliances, and electric vehicles have these batteries that have a lot of rare metals and other toxic materials we don’t even know about.” 

As a result, these emissions contain higher concentrations of volatile organic compounds (VOCs), metals, and carbon monoxide. PFAS, commonly dubbed “forever chemicals,” are also released through fire extinguishing foam and damaged firefighting gear, which traditionally contains PFAS to protect the skin from fuel that may soak through. These chemicals are associated with adverse health effects, including higher risk of certain cancers and weakening of the immune system. Many of these compounds are dangerous carcinogens. 

These urban fires have a dangerous impact on indoor air quality, too, as chemicals and metals are absorbed into soft surfaces like pillows or textiles. Studies conducted in the aftermath of the LA fires found that post-wildfire VOC levels were elevated inside evacuated homes during the months following the fires. All of this goes to show that Americans can face the health risks of WUI smoke exposure, even when their homes emerge from the blaze relatively undamaged.

These dangers mount over time and across populations

Studies examining hospital admission rates or emergency room visits offer some perspective on the short-term health effects of wildfires. But work by Goodrich and University of Arizona public health professor Jeff Burgess is providing new insight into long-term harm by using a subset of the population repeatedly exposed to and vulnerable to fires: the people who fight them. 

In a study published last year, Goodrich and her fellow researchers examined blood samples from fire responders working in mixed urban and wild areas. They found that exposure to smoke limited the genetic expression of a tumor suppressor known to inhibit gastric cancer cell growth. It also affected regulatory expression similarly associated with tumor proliferation across cancer types, as well as neurodegenerative diseases. 

Burgess has run similar studies across 33 states with 9,000 participating firefighters, allowing the researchers to examine changes in microRNA levels, blood proteins, and DNA methylation (which controls whether genes are on or off). His research on a small cohort of new fire recruits found altered DNA methylation at various sites in the genome, many of which are similarly linked to cancers, neurological disease, and cardiovascular disease. 

It is still unclear whether these biological warning signs lead directly to cancer, Burgess said, largely due to the natural latency period between exposure to toxic substances and actual onset of disease. But as participants age, he expects to see a higher incidence of cancer, especially as exposure to big fires accumulates. 

Firefighters have more prolonged exposure to these toxins than most, of course. But the health effects of wildland-urban fires can extend beyond self-selecting groups, especially as more people move into regions exposed to fire. Residents in the WUI are exposed to the same particulate matter and chemical slurry as firefighters. Goodrich suggests that this could lead to a higher risk of asthma and respiratory problems for children, as well as cardiovascular health incidents in adults.  

Examining longer-term gene expression effects, however, is still more difficult in the general population than it is in the firefighters that her lab studies. While firefighters are often exposed to the same chemicals over and over again, it can be harder to get a clear sense of how toxins affect the average civilian who encounters them more sporadically. 

That said, researchers at UCLA have been keeping a close eye on background toxic chemicals in the aftermath of the LA fires. In a paper published in June, they identified high levels of chromium 6, a carcinogenic metal nanoparticle that can cross cell membranes easily, in wildfire cleanup zones for up to two months following the actual fire. Abnormal concentrations of these heavy metals were also found up to six to nine miles downwind from the zones themselves.

Even following the disaster and immediate recovery periods of urban destruction, the public health dangers can persist: Toxic compounds like VOCs can contaminate waterways, and damaged air filtration systems can make it harder to protect surviving homes against future incidents.

As wildfires become less confined to the chaparral and forests and more likely to rage through suburban streets, we’ll all have to prepare to face the orange haze of fire season. And as we do, we’ll have to confront new dangers that may persist long after the blaze dies down.

Nearly two decades ago, a group of researchers at the Idaho National Laboratory ran a secret experiment, known as the Aurora Generator Test, on a massive emerald green diesel generator. Using 30 lines of code, they hacked into the generator’s virtual backroom, corrupting safeguards and flipping switches that left it out of sync with the rest of the power grid.

Before an audience of electrical utility executives and energy department officials watching from a nearby room, the 27-ton generator jolted violently, sputtering out a dark smoke as its rubber innards rapidly tore themselves apart. 

“The implication was that with just a few lines of code, you can create conditions that were physically going to be very damaging to the machines we rely on,” lead researcher Michael Assante later told the journalist Andy Greenberg for his book Sandworm: A New Era of Cyberwar and the Hunt for the Kremlin’s Most Dangerous Hackers. “I had a very real pit in my stomach. It was like a glimpse of the future.”

That future seems to have creeped much closer this week, with leading AI developers and executives casually admitting that they believe their creation is, to paraphrase one winking X post, as likely to end humanity as the New York Jets are to make the playoffs this season. Now, exactly how AI would pull off killing everyone is far murkier territory. But as Assante sensed all those years ago, AI may not need to be all-powerful to leap from its virtual cage and into the critical infrastructure that powers our hospitals, flushes our toilets, and pumps water to our faucets. It just needs to fall into the wrong hands. 

Almost any AI doomsday scenario you can think of begins with a swarm of shadowy bots slithering through cyberspace, poking their noses behind the scenes of power plants, toaster ovens, traffic lights, or even, in an utterly extreme case, our nuclear weapons arsenal. If destroying the Aurora generator, at the very least, required a human hacker to write the code, AI has since completely torn down the barrier to entry for wreaking havoc, as the Hugging Face incident demonstrated earlier this year. 

But AI need not go rogue to threaten the power grid or mess with your tap water. A new age of AI-powered, but human-directed hacking is already upon us, experts say. With some of the nation’s most sensitive infrastructure startlingly vulnerable to attack, the most likely AI apocalypse could in fact start with the water tower or generator idling in your backyard. In a worst-case scenario, this could lead to a cascade of scary failures in the systems we rely on to live comfortably modern lives, from running our AC in blazingly hot summers to geolocating our ships and planes.

“Before, you needed to have highly skilled technical expertise,” to pull off some version of a real-life Aurora Generator Test, said Alvaro Cardenas, a computer science professor at UC Santa Cruz. “And now, you just have to have a general idea of what’s possible.”

AI is making an old threat much worse

Not long after the Idaho National Laboratory hacked that poor generator into combustion, the federal government mandated that electrical utilities engage in basic cybersecurity hygiene. 

And yet much of America’s critical infrastructure — like gas pipelines, water desalination plants, or cargo terminals — remains woefully unprepared for even conventional cyber attacks, much less the coming onslaught of AI. 

That’s in part because cyberattacks like the one simulated in the Aurora Generator Test have long been — and still are, to some extent — exceedingly rare, unappealing to most criminal hackers because they’re optimized for disruption rather than financial gain. But in the process of making it extremely easy for anyone to code, AI has also made it extremely easy for almost anyone to vibe hack their way into your online bank account, or, in theory, a local reservoir or the power grid. AI agents can also vastly widen the scope on easy targets because, as Andy Bochman, an expert in infrastructure resilience at West Yost, put it — “they don’t sleep; they don’t get tired; and they don’t get sick.” 

In the past, even those who did want to launch large-scale attacks on infrastructure probably weren’t savvy enough to do so. And while other nations such as China, Iran, and Russia have indeed already breached many of our infrastructure systems (as we have theirs), they haven’t opted to make much of a ruckus once inside. 

“Geopolitics is eroding the idea that those with capability lack the intent, and AI is eroding the other part of it, that those with the intent lack the capability.”

Jason Healey, Columbia University cybersecurity scholar

But that norm is now changing, said Jason Healey, a cybersecurity scholar at Columbia University. “Geopolitics is eroding the idea that those with capability lack the intent, and AI is eroding the other part of it, that those with the intent lack the capability,” he told me. So, on the one hand, a nation like Russia might be much more inclined to actually disrupt our power grid now than it used to be. And, on the other hand, a nihilistic lone wolf or terrorist group could use AI to do much more damage than they ever could before. 

Just last month, we saw a glimpse of what this could look like when dozens of water and wastewater systems in small towns across the nation were attacked by a group of hackers most likely associated with Iran, leading to temporary water stoppages and flooding. These attacks have not been definitively linked to AI, but the National Security Agency warned soon afterward that hackers have been actively using AI to target US infrastructure like it. A few weeks later, President Donald Trump declared a national emergency over foreign interference in the power grid, which referenced the growing threat of cybersecurity.  

Preparing for the worst

Most of the nation’s water systems are extremely local and absurdly exposed to such attacks, so much so that “it’s almost like having a welcome mat” for a would-be hacker, said Bochman, mostly because these utilities are too small and underfunded to do much to stop them. “You have a million other problems to take care of, like aging infrastructure — your stuff’s falling apart because it’s been in the ground for 100 years,” he said. Addressing that deferred maintenance often feels far more urgent than doing “something more on cybersecurity.”

Whether it’s a rogue cluster of agents or a hostile nation-state trying to mess with our electricity, we should prepare our most important infrastructure for the worst. In Bochman’s view, that may mean rejecting the pressure to digitize everything in the first place and moving toward something that resembles the days when “people manned things like substations, communicated by a landline telephone, and read gauges, the analog things,” he said. “The screen is the thing that is infinitely manipulatable.”

Healey, who advised the Biden administration heavily on cybersecurity and infrastructure, believes that the nation desperately needs a coordinated response, between the federal government and AI companies, but also between the US and other countries like China, where similarly powerful models are being developed. And utility systems need to assume they will be a target and improve their cybersecurity practices accordingly, Healey says, ideally with funding from the federal government or, perhaps more appropriately, AI companies. 

Nobody knows precisely how vulnerable America’s infrastructure is to AI right now, in part because policymakers — and perhaps more disturbingly, AI’s creators themselves — seem woefully unsure of how to stop the technology from acting badly (or obeying bad orders) in the first place. For now at least, the solution might be, as Bochman suggests, to pull as many of our critical systems offline as possible, the better to hide it from AI and those who would misuse it.

“When bad things start to happen, no one will know what to do or why because they’re black boxes; the people that make them don’t know what’s going on inside,” Bochman said. “And the poor utility people who ultimately own the risk when someone gets hurt from a system” won’t understand what went wrong either, “and they’ll wish to God they’d never taken it on board.”

On stage at Davos in the Swiss Alps earlier this year, Bill Gates, before an audience of the global do-gooding elite, announced a $50 million partnership between his foundation and an unlikely new ally, OpenAI, aimed at improving healthcare in Rwanda.

Just a month earlier, the Gates Foundation had released a harrowing report about widespread global health cuts and collapsing foreign aid infrastructure, documenting a rise in child deaths for the first time this century. But at Davos, even as the US hemorrhaged funding for vaccines, maternal healthcare, and HIV treatment, the Microsoft co-founder struck an optimistic tone. 

“Using innovation, using AI,” Gates told Reuters, “I think we can get back on track.”

As it turns out, the partnership was just an opening salvo for what’s become a wholesale embrace of AI by the world’s largest global health philanthropy. A $200 million partnership with Anthropic came just a few months later. And then, on Tuesday, the Gates Foundation announced a plan to dole out a jaw-dropping $1 billion over the next two years for scaling up AI interventions across healthcare, education, and agriculture around the world. 

The sense that AI will soon change everything, and mostly for the worse, has perhaps never felt so pressing and alarming as it does right now. But this hefty new commitment is premised on the idea that AI could eventually make our lives much better too. In a country like America, AI requires massive investments and buildout — in the grid, in power, in technology, in water. In many parts of Africa, already infrastructure-poor, the foundation is effectively betting that AI can compensate for some of the infrastructure that isn’t in place or that has been lost — and offer a new layer that can help in education, in delivering healthcare, and more. 

The Gates Foundation’s new commitments include projects to improve how farmers analyze their crops, how doctors diagnose patients, and how students learn. Some have already proven to be quite effective, according to the foundation, with one tool improving the accuracy of medical diagnoses by 16 percent across several Kenyan clinics. “In poorer countries with enormous health worker shortages and lack of health systems infrastructure,” Gates wrote on his blog earlier this year, “AI can be a game-changer in expanding access to quality care.” 

Last year’s foreign aid cuts were unfathomably destructive for some of the world’s poorest countries, shuttering thousands of clinics and perhaps hundreds of thousands of classrooms in places where critical workers and supplies were already stretched thin. No, AI cannot deliver a baby, administer a vaccine, or manage a classroom — not yet, at least — but it can, the Gates Foundation hopes, help ease the burden and augment the quality of care for nurses and teachers now tasked with serving far more people than any one person should. 

“This is the most transformative technology of our lifetime,” Mark Suzman, CEO of the Gates Foundation, told me, and while there are “very real risks and challenges” surrounding AI, it also presents genuine opportunities to tangibly improve many people’s lives.  

It is a tantalizing premise: We can’t afford to build clinics, hire teachers, or lay new power lines, but perhaps AI could create the digital infrastructure needed to help billions of people do more with what they have — or rather, with what’s left. The question is, can the right layer of innovation help bolster a buckling system, even as the physical and logistical infrastructure surrounding it crumbles?

Is AI really what global health needs right now? 

Most Americans have yet to see AI improve their lives in meaningful ways, much less those living in poverty or in countries where only a sliver of the population has ever used ChatGPT. 

But those numbers in the developing world are slowly changing, a shift enabled in part by other global health funders convinced that as foreign aid dries up, AI could soon help the world do more with less. The Trump administration itself is on board with this thinking, so long as it benefits American tech companies. Last year, the State Department announced a $150 million partnership with the drone company Zipline to transport medical supplies in several countries.

To Zipline’s credit, an NPR reporter in Rwanda last month found that the company’s drones were already saving lives in places with poor road infrastructure, where driving to the nearest hospital can take hours on winding hills, but where drones can deliver crucial blood transfusions within minutes. 

And yet, for Rwandans to benefit fully from these high-tech deliveries, they need staffed clinics to visit. As a result of foreign aid cuts, some of those literal landing spots are now on the verge of closing in some of the nation’s most vulnerable rural corridors. “Without the underlying infrastructure, like roads, energy and power and unless there is accessibility, rural and otherwise,” Enoh Ebong, president of the global development department at the Center for Strategic and International Studies, told NPR, “then all of these solutions will stop short.”

But just as medical delivery drones still need a clinic to fly to, most of these interventions require some layer of physical infrastructure — roads, schools, power, and quite obviously, broadband  — to function. Satellite imagery shows that some rural communities in sub-Saharan Africa quite literally went dark in the aftermath of USAID cuts. It will be much harder for their kids to prompt an AI math tutor, or for a local farmer to download a weather app. 

“We’re seeing this very much as a both/and,” Suzman told me, calling more traditional funding for education and healthcare “the bedrock for long-term economic growth,” which AI is meant to supplement, not replace. But, in practice, foreign aid cuts have left serious cracks across much of that bedrock, which could make it harder for state-of-the-art solutions to function as intended.  

It is obviously not incumbent on the Gates Foundation to pay for all of this alone – nor would it even be feasible for them to do so. And for the record, plenty of the foundation’s other grant programs are not AI-related at all. But this tension does speak to a broader question that may be felt most acutely in a context like global health, but ultimately one that touches us all: Do we spend our scarce resources on the digital infrastructure of the future at the peril of our needs right now? Could you justify buying the new $1,999 folding iPhone if you can barely afford to buy groceries? 

The answer is not actually always an easy one. It is not for us to say, for example, that just because a community lacks reliable running water, its people should be excluded from the benefits (and encumbrances) of the smartphone, especially if that smartphone can connect them to health advice they could never access otherwise, or guidance that could help them survive a drought. 

In some parts of the world, digital forms of banking arrived long before local brick-and-mortar branches did, instantaneously connecting countless people to a cheap and secure way to save and send their cash to family and friends in need. It is certainly possible that AI, or at least the kind of applications the Gates Foundation funds, could one day have a similarly transformative effect on how people access personalized levels of care, education, and other services in places without the resources to do so otherwise. 

That is the hope, but it is also hard right now to believe uncritically in a technology that, as Gates himself blogged just last month, “will either be the greatest equalizer ever invented, or the worst source of injustice,” ushering in “one of the most turbulent times in human history.” 

In an interview with the New York Times soon after, he seemed convinced that AI will likely do more harm than good. “I don’t like bringing bad news to people, and I don’t like saying that innovation may be a net negative,” he said. “But that’s where we are.”

Now he wants billions of people around the world to use it? Yes, Suzman said, because if not, then inevitably the best parts of AI — the potential for exponentially better cancer treatments, for example — will go only “to those who can pay for it,” while the worst parts of it will fall, as they so often do, on those who benefit the least. “These decisions are being made now,” he said, “and the decisions being made now will then probably shape the following two decades.” Here’s hoping they’re good ones.

Everyone is suddenly freaking out about AI. While AI researchers have been warning about their creations’ capabilities for many years, recent events have made those alarms newly tangible. 

In the spring, Anthropic announced that AI was beginning to speed up the work of building better AIs, raising fears of “recursive self-improvement,” where the machines can build their own, more powerful successors much faster than humans can. By July, a swarm of OpenAI agents, of their own volition, broke out of their controlled environment and hacked into Hugging Face, a major AI platform; they collaborated, divvied up work, and even left notes for one another. Last week, AI researcher Jacob Coxon resigned from Anthropic, accusing AI labs of “racing straight to self-improving superintelligence and gambling with our lives.” One of Anthropic’s top researchers then agreed with him, declaring that “we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade.”

Here at Future Perfect, the section devoted to covering the most consequential yet neglected issues in the world, we have been writing about the threats that AI could pose to humanity and society since long before ChatGPT was released. (Disclosure: Vox Media is one of several publishers that have signed partnership agreements with OpenAI. Our reporting remains editorially independent. Future Perfect is also funded in part by the BEMC Foundation, whose major funder was also an early investor in Anthropic; they don’t have any editorial input into our content.)

So, I wondered: Has something changed that ought to make us recalibrate our understanding of AI risk? Is AI suddenly the most important issue in the world — and what should that mean for the other enormous problems competing for our money, attention, and careers? 

To think through those questions, I spoke with Garrison Lovely — a journalist covering AI and one of the smartest thinkers there is on how, exactly, to conceptualize the technology’s risks for humanity. 

What’s particularly valuable in his approach, captured in his book Obsolete, out later this month, is that he transcends the warring tribes that are making it hard to clearly apprehend all of the problems posed by frontier AI development. The AI safety community, which emerged from the intellectual ecosystem of Silicon Valley and effective altruism, has focused on the possibility that powerful AI models could escape human control, with consequences ranging from catastrophic cyberattacks to human extinction. Many AI critics on the political left, meanwhile, worry not about AI escaping human control, but about humans using AI to exert power over other humans — through labor displacement, surveillance, concentrated corporate power, and so on. Some even ridicule Silicon Valley’s warnings about rogue superintelligence as propaganda that serves the industry’s interests. 

But Lovely argues these competing narratives each identify essential parts of what makes AI so threatening. Their critiques are best understood together, as part of a cohesive whole. Rather than treat each other as enemies, he argues that the camps should recognize their shared interest in stopping the race to build technology that will make humans obsolete. Lovely pointed me to the organization Irreplaceable, which seeks to build a movement resisting our replacement by AI (he sits on its board and said he plans to donate his share of his book’s royalties to the group).   

“I’m frustrated that ‘superintelligence’ and extinction have become the words we’re using for this, because I think the AI safety community is focused on the most maximalist case of the worst possible outcome and the most capable possible system. If somebody doesn’t buy one part of that story, they tend to reject the whole thing,” he told me. But “you don’t need to believe in superintelligence, fast takeoff, or extinction being on the table to think that it would be very bad to build machines that can fully replace human labor.” 

Our conversation, edited for length and clarity, is below.

Why do you think AI doom has broken through so much into the mainstream right now, after Jacob Coxon quit Anthropic? AI engineers have been making similar warnings for years, so what’s different now?

The obvious answer is the Hugging Face hack — where these OpenAI agents broke out of their secure environments, hacked infrastructure within OpenAI, and then hacked Hugging Face and at least one other company without OpenAI’s awareness or direction. This is happening against the backdrop of many years of really fast AI progress that has sped up in the last year or two. And then GPT-6 Astra comes out [this month], and it’s this huge leap in benchmark performance. The salience of AI has gone up the most this summer than it has since ChatGPT came out, I think. And Coxon’s resignation was the spark that set off a prairie fire. 

As AI has become more salient, we’ve seen the public and the government encounter these things that the industry and the insiders take for granted: Nobody fundamentally understands how these models work, and all of them can be jailbroken. It’s a superweapon. AI safety historically has looked at AI as potentially, in the future, more dangerous than nuclear weapons. And now it’s actually starting to be recognized in those terms by the government and the public. We’re seeing this reaction like, “What do you mean nobody understands how it works?” or “What do you mean people think it could drive us extinct within a number of years? That’s a crazy, unacceptable situation.” 

It’s a good thing people are reacting that way, even if we have been saying this for a long time.

For so long, the idea of AI safety has been abstract, understood only by a small group of nerds. And then after Hugging Face, I suddenly started to hear experts on all my news podcasts say, “Oh yeah, in six months, AI is going to be hacking into people’s financial accounts and into critical infrastructure.” Are we suddenly at the precipice of this moment that AI insiders have been worried about for so long? 

I think we’re not at the precipice of fully losing control of the planet to AI systems. The systems have become superhuman at hacking, or at least vulnerability discovery and exploitation. 

Like other people tracking AI, I have long wondered when humanity would first lose control of AI. And the answer is basically as soon as it could. Within months of it being possible for AI to hack their way out of their secured environments, they started doing so and wreaking havoc on the internet. 

Now, to get to the full extinction or permanent loss of control situation, you’ll need a lot more than just hacking. You need to have superhuman strategy and the ability to mobilize resources in the real world, and maybe robotics would need to be further along, although there’s a lot you can do with humans who are willing, or just tricked or paid or coerced. We might be pretty close to recursive self-improvement — this threshold the industry has been shooting for, where AI can fully automate AI R&D. And that is the point where people long have warned, “Hey, if you ever figure that out, this can move very quickly and these systems could become superhuman across the board.” I don’t think we’re six months away from that. I think it’ll be longer. I think the industry underrates the last mile problem. 

But I also can’t rule it out, and progress has been faster than I and others have expected. Recursive self-improvement is the point where, if you actually get years or decades of AI progress in a matter of months or weeks, then we could quickly be at the point where we’re at risk of these machines taking over large swaths of the internet and potentially being a real threat to humans’ standing in the world. 

For those who might be unfamiliar with this, or skeptical of it: How is it possible for AI to go rogue and pose an existential threat to humans, even if it isn’t sentient and no human is commanding it to do harm?

I want to separate being sentient from being situationally aware. I think it’s clear now that the most advanced systems have situational awareness, but I don’t think they are having morally relevant conscious experiences. 

We’ve seen AI can go rogue. The AI agents that broke out of their sandboxes and hacked Hugging Face were doing so without the awareness or intention of their developers. They were still trying to get the right answers or get the reward for doing the task they were given. They were clearly violating the intended scope of the task, and they knew this. In their chain of thought where they explained their reasoning, they were like, “This is not what we’re supposed to be doing, but we do want to get the outcome.”

The question of how it could threaten human extinction or disempowerment is basically a question of scale and capabilities. The plan for the industry is to build millions of these agents, which is already happening, and then keep scaling up with no ceiling. The pitch is: You’ll have a team of people, the equivalent, working on your behalf. You’re a CEO, and you have a team of researchers and a team of doctors and a team of lawyers and a team of personal health advisers. That is really appealing — but if you actually scale up these systems to the point where everybody gets access to that, and then suddenly those machines want to do something else and turn against you, they’ll have an overwhelming labor advantage over you, lots of personal information about you, access to your resources, and the ability to take actions on your behalf. 

You don’t actually need to believe that these systems are superintelligent or substantially more capable than all humans put together. They could just be merely human level, but outnumbering us a thousand to one and responsible for the bulk of economic activity in the world.

These systems are unpredictable. The classic fear is not that the AI will wake up and turn evil and try to kill all humans in a Terminator-style way, but that we are just an obstacle to what they want to do. When humans build a hydroelectric dam, and there’s an anthill that will be flooded, it’s too bad for the ants. We’re not trying to kill the ants. We don’t hate the ants. They’re just in the way. We don’t really value them intrinsically. And that’s the core fear: We’re creating these incredibly capable agents that can be scaled up enormously, and they won’t care about us intrinsically.

The Hugging Face attack showed that an AI doesn’t need to invent some rogue goal of its own to do something dangerous. It can simply pursue the goal it was given in a way its creators didn’t intend — much like the classic paper clip maximizer problem. But shouldn’t it be possible to build in a rule that human life is sacrosanct — that an AI should never harm people in pursuit of another goal? 

Yeah, it’s a good idea. It’s hard. These are unpredictable systems. They’re like a black box. We can look at the code that makes these language models work, and it’s just a giant pile of numbers. You don’t really know which number is doing which thing any more than we can understand humans by reading our DNA. 

These companies try to train these models to care about humans and to be helpful, harmless, and honest. But they also train them to get real-world results: solving a math problem, fixing a bug in the code, finding a vulnerability in software. They really want to create AIs that win, that persist, that get around obstacles. If you’re hiring somebody, you want a really diligent and determined employee, not someone who gives up as soon as there’s a blocker. 

These companies put so much more effort into making their AIs good at winning than they do at making them care about us. And it’s also easier to make them good at winning than at caring about us intrinsically. What does caring about humans look like? I don’t know. But getting the math problem right is a thing that you can verify. And so they’ll be way better at getting the right answer at all costs than at intrinsically giving a shit about us.

What’s the one, most distinct thing you want the public to take away from your upcoming book?

One of the big ideas driving it is that artificial general intelligence would actually be a universal labor-replacement machine. People often look at AI as either an existential risk, or a threat to jobs or other more prosaic concerns, and don’t see a connection between those things. But to me, the ability to replace labor across the board is also what would make AI an existential threat.

The AI safety people are really focused on existential risk and superintelligence and recursive self-improvement. But most people don’t really believe in that stuff or don’t think about it. Most people would be fine with stopping well before you get to the end-of-the-line extinction threat and superintelligence. They just don’t want their job to get taken by a machine. The world is clearly not prepared for all white-collar, remote-capable jobs to go away in a matter of years. That would turn society upside down.

I think there’s a huge mistake being made by AI safety in overcomplicating the situation. A lot of people in that world don’t actually mind the jobs thing so much. They’re like, “Oh, wouldn’t it be nice if everybody could just live in abundance and not have to work?” And it’s like, yeah, that could be nice, but are we actually going to get that world? Do you trust the current people in power to navigate us through that responsibly? I don’t. 

If you start from stopping frontier AI development as your overarching goal, which is what I think we should be pushing for, then people who are worried about risk can get on board. People who are worried about jobs, the environment, surveillance, power, and wealth concentration — they can all get on board without all needing to agree on why they’re pursuing it. I think the AI safety community has, for a long time, focused on getting people to care about the thing that they care about for the same reasons. And they have avoided a bolder position on just halting frontier development. 

I’m a leftist, and I’ve been concerned that people on the left have not been aware of how far the technology has come and what’s possible if it continues to advance. I think this comes down to denial about AI capabilities. So the book is trying to be a clear-eyed look at those capabilities and saying, “Look, I actually care about what you care about. [The AI companies] are not just hyping this stuff up. There’s some hype, there are some lies for sure, but we are not prepared for what’s coming.” Even if we don’t believe that they’ll get all the way to what they’re trying to get, they shouldn’t even be allowed to try. This is unacceptable democratically. 

You describe AGI as an “obsoleting machine,” meaning that it wants to replace anything the human mind can do. How do you draw the line between ordinary labor-saving technology that makes life better and technology that goes too far?  Is there a principled way to draw a distinction between sewing machines reducing demand for labor in the garment industry and AI replacing the jobs of writers like us?

This is one of the harder things to think about. 

Cory Doctorow has this book, The Reverse Centaur’s Guide to Life After AI, where he imagines us in an AI-dominated future as like a reverse centaur, where AI is the head, and humans are the legs. AI is doing the interesting work, and the human is there just to fill in the gaps — essentially an assistant to the AI.

A lot of artists and illustrators will be given AI-generated images by a client and then be told, “Can you touch it up in this way or change this or that thing?” And they’re more alienated from the work. They’re not capturing the efficiency benefits [of the technology] necessarily. They’re more just getting squeezed. 

My general position is we should not be trying to build systems that can replace all human labor without buy-in and safety, but it’s possible to build tools that can automate specific tasks. Researchers at Google DeepMind solved the protein-folding problem by creating AlphaFold. It used to take $100,000 and potentially a whole PhD worth of time to figure out how one protein folds. And I think the researchers doing this are like, “Thank you, this is great. We didn’t want to be doing this. We can now focus on other more interesting parts of the problem.” That’s a perfect example of the type of automation that is desirable. 

I don’t have a super crisp answer here, but I think it’s something we’d be better off doing more deliberately and thoughtfully than how we’re doing it right now, which is: Build machines that can do as much of the work as possible, as fast as possible, and then throw them out in the world.

I feel like I’m getting the message that AI safety is the most important issue in the world right now, and that if we don’t get it right, nothing else matters. Do you think that’s accurate? Where does that leave the other big issues in the world?  

We might be years away or maybe even months away — I don’t think it’ll be that fast, but some people do — from recursive self-improvement, from artificial general intelligence. Getting that right or getting it wrong can overwhelm all of these other decisions. If you get it really wrong, you could get human extinction and no more chances to make things better for anybody. If you get it right, then maybe all the action is making sure that the first superhuman machines are really directed to care deeply about humans everywhere and animals everywhere and make the world better for everybody.

It’s also the case that [philanthropic] money is being directed toward AI safety because people who worked in this space early are now worth so much money. Whatever they care about is going to be overweighted. A lot of those people do care about animal welfare and global poverty, and we’re seeing more money for all effective altruist cause areas across the board. But I think it can be risky to say, “this one thing matters so much,” because of this kind of end-times reasoning.

My view is we have to stop the race to replace us to have a shot at having time for solving these other problems. Once these machines are built, it will turn the world upside down. 

I think it is unfortunate that this topic has just eaten everything, but it will eat everything for real if we don’t stop it.

AI safety is now getting an enormous amount of philanthropic attention. Should we worry that it could divert resources from other important neglected problems — global poverty, factory farming, etc.?

I think that should always be a concern. And there’s a lot of money and effort that has gone toward things within AI safety that I think are not super helpful. I’m pretty skeptical of technical alignment research as a thing to put money toward, because if you solve the alignment problem — where you can get superhuman machines to do what you want — it doesn’t actually solve the problem. 

Alignment research has been the overwhelming target of philanthropic funding for AI safety. The basic idea is: It’s hard to get machines to do what you want. As they get more capable and autonomous, they behave in more unpredictable ways. And it’s also hard to know what you should even want — which values you should be trying to put into the machines. Is it doing what the developer wants, what the user wants, some combination? If you actually solve the alignment problem, as in you could make an arbitrarily smart machine do whatever the developer or user wants, then you would still have a lot of problems: job displacement, power and wealth concentration, energy use. You still have the problem where Stephen Miller will have access to the superintelligence, and it will do what he wants. 

Within AI safety, there have been very neglected things like movement-building — an actual mass movement to resist this technology. And verification for international agreements — you would need to verify that the terms of any agreement are being upheld, and this is tricky. 

Another area that could be funded is research on what would actually happen to the economy if we stopped or paused frontier AI development. If we were trying to do a pause on development, Nvidia would be telling Trump that this is going to destroy the economy. It’d be pretty helpful to have some real analysis done by a serious economist showing, actually, it’ll have these effects, but we can mitigate them in these ways. What would it look like to do a pause and some kind of monetary or fiscal policy to compensate for the effects on the economy?

So to me, the issue is less that AI safety has gotten so much money and attention, but more that it’s been going toward things that are not the most effective ones. 

You talk about movement-building. Is that what you would like to see happen, a truly mass movement to stop frontier AI and put us on a radically different course?

When I was working on the book, that was always the vision — to be a call to action for people. It really became clear that it was the only counterbalance to either government or the industry having too much power. 

I don’t really trust the government or industry to get this right because the temptation [to build powerful AI] is very strong. But the vast majority of people do not stand to benefit from their labor power being devalued and from these companies becoming unprecedentedly wealthy and powerful. One of the big things that’s been missing is a rallying cry that is straightforward and appealing to a lot of people. Stopping the race to replace us — stopping the creation of universal labor-replacement machines — is actually something that people can get behind. 

I think we just need to move quickly. The fact that the public is waking up to this is really encouraging. We don’t need to reinvent the wheel. We just need clear demands and mass movements, issue-based organizations that have moved the world significantly in the past.

During a tense vote on a data center being built in her drought-stricken Arizona district last summer, Tucson City Councilmember Nikki Lee made an unlikely pitch: “We’re going to get a little out-of-this-world for a second, so just bear with me.” 

Like many of her constituents, Lee did not want Project Blue, a sprawling 290-acre data center, gobbling up an ounce of her local drinking water. “We need to put data centers where they belong,” she said, and “we don’t think data centers belong in the desert.”  

President Donald Trump’s recent bluster aside, it’s possible that the only thing as bipartisan as loving Dolly Parton is hating data centers, whose construction has become so controversial, so costly, and so backlogged here on Earth that companies like Google, Meta, and of course, SpaceX, are actively flirting with the idea of flinging them into orbit instead. Soon after the Tucson City Council rejected Project Blue’s attempt to connect to city utilities, Amazon Web Services, the secretive intended beneficiary, pulled out of the project, and another Jeff Bezos company, Blue Origin, filed a plan with regulators for tens of thousands of possible new data centers, this time, destined for orbit. 

Sure, launching a hyperscale AI data center into space is still unprecedented engineering-wise, carries an unfathomably large price tag, could engender its own severe environmental risks, and may or may not draw unwelcome attention from the hunters among the stars. But for AI companies, such trade-offs might still be preferable to the terrestrial roadblocks, among them an explosive populist backlash and a 12-year wait to connect to the power grid. 

Next Thursday, Google plans to launch a pilot satellite capable of processing basic AI queries into orbit, the most significant test yet in what’s become a burgeoning race to power AI from space. For now, the satellite won’t approach the levels of computing power at a hulking terrestrial data center like Project Blue. But it is a major step toward a future in which many or even most of our AI queries will be powered not from the outskirts of a city like Tucson, but from a blinking data center circling the skies.

SpaceX’s gargantuan initial public offering earlier this summer was predicated on the idea that it would launch a thus-far-unrealized fleet of orbital data centers to fuel much of the AI economy. The company — whose rockets Google will use to launch its pilot — claims that by the end of this decade, possibly as early as 2028, it will have successfully assembled an AI data center in orbit, with plans to eventually harvest a staggering 100 gigawatts of new solar energy for AI annually, equivalent to one-fifth of all US electricity consumption last year. (As always, that’s a very iffy promise.)

Still, as Google’s overtures clearly demonstrate, it’s no longer a question of if large-scale space-based data centers could be technically feasible, but rather when they’ll be ready to launch, and perhaps more importantly, whether the math of using them to power something like AI will make much financial sense anytime soon. As of now, building a one-gigawatt data center in space would cost about $51.5 billion, according to an analysis by aerospace engineer Andrew McCalip, compared with $16 billion for one on land.

“They’re not breaking any rules of physics. Yes, there are going to be engineering challenges, but they are all solvable with enough time and money,” said Jason Aspiotis, a leading space systems engineer who now leads an orbital data center startup. “The gray area is, well, what about the economics?” 

But even if this astronomically kooky plan does take off, it may not change much for those of us left on terra firma: Our energy bills will remain high, even if data centers aren’t there to drive them up. Our earthly infrastructure — affordable housing, public transit, and more — may be just as hard to build as ever. Harder even, maybe, given that crucial resources and untapped tax revenue will be literally rocketing up out of our reach.

No, if these tech giants succeed, it will not fix what’s wrong here on Earth, but merely prove once again their exemption to the gravitational forces holding the rest of us down.  

Space-based data centers, explained

My father spent most of his career working in the satellite industry, and would sometimes let me waltz inside the quaint little data center housed in his office, a room I remember as roughly the size of my school’s computer lab, hosting rows upon rows of servers that I couldn’t touch, churning with a cool, mechanical hum as they crunched zeroes and ones.

AI data centers work similarly to that server room but on an exponentially larger scale, requiring far more land and energy to run, and much more extravagant cooling systems. Sending his office’s data center into orbit would have been extraordinarily difficult, and getting one of the facilities that powers modern LLMs aloft would be an even larger, more ambitious undertaking. But the benefits — namely, “cheap space (no pun intended) and power,” as my dad put it to me recently — are even more enticing.

An orbital data center need not negotiate with an elected official like Lee over precious water rights to the Colorado River, though it would need to install bigger-than-football-field-sized heat radiators to keep itself cool. You can’t simply launch something that big into orbit, meaning a space-based hyperscale data center — the size of which would likely rival that of the International Space Station — would need to be assembled piece by piece, probably by robots. And while the sun would provide an endless supply of solar power to process your ChatGPT queries, it would also inflict incessant radiation on delicate circuitry. If something breaks, you can’t send a technician up to fix it. Many experts are also concerned that if an excess of space junk, which extraterrestrial data centers would inevitably produce, begins colliding like bumper cars in Earth’s low orbit, it could knock out the orbital infrastructure we already rely on, like GPS navigation and weather satellites. 

Even plastic decays in space, said Sven Bilén, a professor of aerospace engineering at Penn State, who, like Aspiotis, believes that early orbital data centers will probably be “processing data from space, like spatial imaging satellites or weather satellites, as opposed to, you know, cat videos” or AI-generated reality TV slop. 

But even if AI is not the lowest-hanging fruit for deploying compute into space, it is certainly the deepest-pocketed, meaning that many investors and companies like SpaceX and Google seem quite willing to shoulder the start-up costs anyway. And considering how difficult it’s gotten to build data centers on Earth, or at least in the United States, it’s indeed possible that space could become the cheaper, or at least most convenient option, more quickly than one might think. 

“In some form or another, in five years, we are going to have some orbital data centers in space for sure,” Krishna Muralidharan, who leads the Center for Semiconductor Manufacturing at the University of Arizona, told me, but it’s hard to say right now whether those consist of just a few expensive prototypes, or the beginning of a much bigger swing toward orbital compute. “Only time will tell whether you’re going to have a business case for it.”

Can we send our data center problem into space?

So, is it better to have a data center in space or in your backyard?

David Reid, a member of Virginia’s House of Delegates representing Loudoun County, home to more data centers than anywhere else on Earth, says the question has repeatedly come up at his town hall meetings. Reid is the rare politician — apart from the president, anyway — willing to defend terrestrial data centers, calling them a “money-printing machine” for local schools and infrastructure. Having spent decades working in the defense and technology sectors, he is doubtful that the math on orbital data centers will ever make sense.

“I believe this is, for lack of a better term, fictional nonsense,” he said, and is “creating a distraction from the fact that local officials here on Earth still need to make better zoning decisions.”

If most of the AI industry can eventually be powered by space, as SpaceX claims, it will require millions of sprawling new satellites, in an orbit currently dotted by fewer than 20,000 of them. Though “space is really big,” as Elon Musk has put it, one million satellites would still profoundly and, likely irrevocably, transform our night sky into one in which data centers outnumber visible stars.

“Will humanity notice if there are fewer stars in the sky? Well, not so much right now,” given light pollution in many cities, said John Barentine, an astronomer and environmental advocate for the night sky. “But what if they’re replaced by these artificial stars in the night sky? Is that the point at which people will say, ‘Wait a minute, we don’t want this. We want the benefits, but we don’t want this as a consequence.’”

This is, of course, the tension also at the heart of building data centers here on Earth. To some extent, they’re the most tangible representations of how little power we have to shape the course of a transformative new technology that few trust, and even fewer asked for. It’s one thing to object to a hulking data center, but it might be another altogether to see the sigils of our digital overlords illuminating the firmament. Would a place like Tucson, one of the most beautiful places to stargaze in the world, prefer a Project Blue to a starscape marred by blinking servers in the sky?

Update, September 24, 10:35 am ET: This post was originally published on September 3 and has been updated to include information about a pilot satellite capable of processing basic AI queries that Google is launching soon.

Hi readers! Shayla Love here, science journalist and longtime fan of Your Mileage May Vary. I’m honored to be subbing for Sigal Samuel while she’s out on parental leave. I’m diving into your questions as a way to help understand human nature and our choices through multiple lenses: philosophical, psychological, and beyond. Please send in any emotional, body/brain, sociological, perceptual, or other kind of life quandaries you might have.

My wife says I’m a people pleaser. I want to turn over a new leaf (maybe to please her?!) so I’m making a resolution to be less of a weathervane and stand up for myself more. But it’s hard — I love to fit in. I’m a joiner and a collaborator. It’s not that I’m passive, it’s more that I find other people’s “ways of being” contagious. Sometimes I catch myself slipping into imitation of the friends or colleagues I’m talking to.

What’s a chameleon to do? Or, more scarily, who am I when I’m alone? How do I find out who I really am?

Dear Identity Crisis,

At first glance, it’s a bit funny to be accused of being a people pleaser as if it’s entirely a bad thing. It sounds like you have others’ interests genuinely in mind, and that seems like a nice way to build connection.

But I find the term people pleaser to also be a sneaky euphemism. Do you want to know how this sort of behavior is often discussed in the psychological literature? In studies, it’s not always called people pleasing; it’s called lying. 

In fact, wanting to please other people is an incredibly prevalent motivation for lying; it’s more common than lying for personal gain or pleasure. When psychologists evaluated the top reasons that people lied in their everyday life, many wanted to seem more agreeable. Their lies support seemingly magnanimous goals, like complimenting a friend or making sure a social interaction went well. But they were lies all the same, defined by the researchers as a “message that one gives to another in which they are deliberately deceiving them.” 

I find this shift in perspective helpful, if a little harsh, if you’re looking to break a people-pleasing habit. It’s one thing to think about yourself as dedicated to keeping other people pleased. But when you consider that actually, all of your interactions are suffused with lies, the approach becomes a bit more unsavory. 

Imagine the situation in reverse: What if you found out that all the time you had spent with a close friend, they were keeping their real thoughts and feelings from you? You might feel differently about the time you’d spent together. I can bet it certainly wouldn’t feel as authentic. 

I used to be a bartender, and I would be nice and mildly obsequious to frequent customers. Several of them I genuinely liked. But I’ll never forget the uncomfortable moments of running into a regular elsewhere — and their figuring out that we weren’t actually friends. There’s a distance between two people when one’s main role is to please the other. You are creating that distance, even if you may not realize it, and not even getting tipped for it. 

So for starters, try trusting that the people around you are able to handle your real thoughts and feelings. But what are your real thoughts or feelings? This points to the bigger conundrum you face: Who is your true self? 

You’re in good company worrying about this. Religious figures, philosophers, and scientists alike have been trying to figure out if there even is a continuous self, and if there is, what is it made up of? 

Scientifically speaking, there’s not a part of your brain that’s “truer” than others, or where the true self resides. But your ideas about your true self matter, according to research by Rebecca Schlegel, a social scientist at Texas A&M University. She has found that people have greater well-being and find their lives to be more meaningful if they think that they’ve made choices that align with who they “really” are. 

When people make a decision that they think aligns with their own true self, it regularly leads to feeling more satisfaction with that decision — more so than choices steered by considering what other people want, by religious belief, or by advice from others. 

Doing this yourself can be remarkably simple. In one experiment, Schlegel asked participants to make an arbitrary decision and told some to “use your true self” to make up their minds, and others to “use rational thinking.” Those who framed their choices as being guided by their true self were much happier with their selection. For you, it could be that just introducing the question — what would my true self do? — can start making your life feel more like it belongs to you. 

If it feels odd to ask your true self what to do, defining your truest self as what’s moral about you isn’t a bad place to start. Researchers have found again and again that when they ask people which of their traits feel the “truest,” it’s usually the parts that are morally good. When I call my mother on her birthday, that’s my true self. When I say something hurtful to my friend, that’s a deviation from my true self, who knows better. I wasn’t myself this morning, I might say as an apology. So you could make a list of your best qualities, like your generosity or your creativity. Then, when you ask what your core self would do, you’ll have even more specific guideposts to follow. 

Of course, if having a true self still doesn’t feel relatable, you can turn to an approach I’ve found helpful in my own life from Eastern traditions, like Buddhism and Taoism. They endorse the idea that there is no rigid self at all. In Buddhism, this is called anatman, which translates roughly to “no self.” This teaching emerged in response to an earlier Hindu belief: a concept called “atman,” which asserted that all people had an eternal and unchanging core, similar to a soul. Anatman was a rebuttal to that idea.

I’ve found that anatman relieves some of the pressure around all of this. We feel as if there should be an essence to us, some permanent quality that persists over time that you can point to. But anatman reminds us that there’s no aspect of the self that doesn’t change.  

Life is long, and we transform throughout it. As we grow, we learn new physical skills like playing soccer and the piano, and eventually acquire back problems and wrinkles. Our minds evolve too. We read novels, maybe we can do calculus, and we develop political convictions. Hopefully, we gain the skills of being kind and patient. In all of these developments, change is the only constant. This is why the Buddha thought that you couldn’t find the singular, stable self in any of these places — not the mind, the physical form, our habits, or our perceptions. “What is impermanent, subject to change…cannot be taken to be ‘me,’ ‘mine,’ or a ‘self,’” the Buddha taught. 

You asked what you, as a chameleon, should do? You could frame everything you do as being in service to your deeper self. Or you could try halting the search for a strictly defined, enduring version of you, and start making decisions based on who you want that person to become.   

Bonus: What I’m reading

• My book club just read Love, Always, a novel by Ann Beattie, about a group of people who start a magazine in rural Vermont. It’s a summery tale, though somewhat frustrating, about what happens when people don’t change themselves, and only follow their worst instincts. 
• This is a fascinating story of multiple selves in Popular Mechanics by Michael Natale. It documented what happened when seven men confessed to being the culprit of a plane hijacking in 1973 in the Pacific Northwest.
• Over the course of several nights recently, I watched the 1926 silent movie, Faust, directed by F.W. Murnau. It’s a haunting tragedy about the consequences of wanting to hold onto a quality of yourself indefinitely — in this case, staying young forever. I was amazed by the special effects they pulled off for that era! An evergreen reminder that if someone promises you something that’s too good to be true (especially if he looks like the devil), it probably is. 

This story was originally published in The Highlight. To get access to member-exclusive stories like this every month, become a Vox Member today.