Ian Bogost | The Atlantic

The Romance of the Gas-Station Sign

2026-04-11T08:00:00-04:00

Adorning urban intersections and rising high at countryside interstate exits, the gas-station sign announces the mood of the consumer economy. For the past several weeks, the economic ramifications of the Iran war have been more or less universally represented through photographs and videos of them. It’s easy to see why: The price of gasoline is always displayed on the sign, in huge numbers that overwhelm the rest of the scene. That design, which is unlike anything else in the economy, makes the gas-price sign a kind of key to understanding American life.

Long before financial data were easily trackable in real time, gasoline offered a view of shifting market forces, seen while commuting to work or driving home from Kmart. In the analog era, workers replaced the numbers on reader boards multiple times a day, occasionally from high up on ladders. Eventually, the signs were digitized and prices were lit up in LED displays—easily changeable and neon at night.

Joe Sohm / Visions of America / Getty

FPG / Hulton Archive / Getty

A pump attendant filling up a Chrysler at an Amoco station in 1958

Drivers—which, in America, is really just to say citizens—learn to watch gas prices closely, to compare them constantly. Unlike with automobiles themselves, no status is conferred by splurging on gas. Groceries such as eggs and milk can be associated with lifestyle choices or socioeconomic striation, but gas is just gas. Even premium gas is just gas. It is a commodity you pay for, whose price is universally known and more or less equally charged. It is the closest ordinary people come to directly interfacing with the pure chaos of the market.

Normally, when you buy something—a loaf of bread, a rib-knit polo dress—you get a finished, usable thing. These are known as goods, and they are differentiated: The bread could have seeds or contain refined or whole-wheat flour; the polo dress might be branded Ralph Lauren or Lacoste. You do not buy the commodities that are used to create goods—wheat, cotton—in almost any circumstance. They are undifferentiated, all the same.

Technically, gasoline is a finished good too. A petroleum company refines crude oil, a commodity, into the fuel you pump into your automobile. And yet, gas works more like a commodity than a good. Gas is essentially nothing more than the price you pay for it.

[David Frum: Why Trump didn’t predict the gas-price spike]

Left: H. Armstrong Roberts / ClassicStock / Getty. Right: Bettmann / Getty.

Left: The OPEC oil crisis of 1973 led to empty pumps at service stations. Right: Motorists line up for gas on the first day of gas rationing imposed on nine California counties following the 1979 revolution in Iran that caused a shortage of crude oil.

Don Norkett / Newsday RM / Getty

“NO GAS” is written on signs placed on gasoline pumps at a station in Deer Park, New York, on December 29, 1973.

Almost no other consumer purchase can be boiled down so purely to its price. Other goods have become somewhat undifferentiated, such as bottled water and store-brand sugar, but they possess other meaningful properties: your perception of the taste of the water, or the way the packaging fits in your hand or looks on the shelf. Even electricity isn’t a good whose price you encounter as part of the built environment; it is a service whose usage cost is averaged for later abstraction into a bill. Gasoline has no packaging, and you do not perceive it (beyond the initial smell)—but you do perceive its changing cost.

Most industries, over the course of their development, move from commodification to brand differentiation, and from products to services. Coffee went from, well, coffee to Folgers to Starbucks. Gasoline has done the opposite.

When motorcars were new, gasoline worked more like a service than a product, let alone a commodity. You’d pull into the station and someone would pump gas for you; you typically wouldn’t know the price beforehand. The petroleum companies—Standard Oil and Texaco, for example—competed on service. An attendant might be unusually friendly and efficient, or check out your car, which was probably fairly unreliable at the time.

Joe Sohm / Visions of America / Getty

Collage of gasoline prices and Bush-era protest signs

Don Emmert / AFP / Getty

A man changes gas prices at a station in Queens, New York, in April 2006.

Price competition for gasoline arrived only after enough cars, stations, and brands came on the scene, by the late 1930s. This is when prices started appearing more prominently on signs and station windows. Early gas-station signs were smaller than today’s, and still subordinated to the oil-company name and image. Price signs began doing political work too: The 1932 Revenue Act introduced a one-cent gas tax, and with it the still-universal 9/10-cent price notation that the oil companies used to show that they were passing on almost all of the cost.

From the ’50s to the ’70s, price displays moved from pumps and windows to roadside signs, which grew larger so drivers in fast-moving vehicles could see them from a distance. But the truly massive gas-station price signs we know today didn’t arrive until after the 1973 oil shock. The price of crude oil quadrupled in the six months after October 1973, when OPEC embargoed sales to countries that supported Israel in the Yom Kippur War. A second shock, in the aftermath of the 1979 Iranian Revolution, sent prices still higher. In photographs of the era, you see lots of gas-station signs announcing no gas, but few that look like today’s signs. Throughout the 1970s, prices remained mostly mounted atop pumps or in modular boards located at ground level.

That changed by the early ’80s. Thanks to the two oil shocks of the previous decade, gasoline became cemented in the American mind as an essential commodity subject to enormous volatility, not a component of automobile maintenance. Gas stations shifted from selling a trusted service to a price-indexed commodity, barely distinguishable from the crude oil from which it is made.

Luke Sharrett / Bloomberg / Getty

A Shell gas station in Louisville, Kentucky, in 2016

Stefani Reynolds / AFP / Getty

A vehicle drives past a Marathon gas station in Washington, D.C., on March 31, 2022.

Gas, once cheap and easy to get, was now regarded as scarce, unstable, and politically charged. The largest, most visible part of the sign that advertised where to buy gas was no longer a company logo but the current price per gallon.

Gasoline’s status as the observable, contentious price of transportation freedom has been universal for 45 years. Despite its many downsides, gasoline unites Americans in a common plight. It gives us a local window into global affairs. It offers a common thread for pocketbook discourse, the economic equivalent to the weather. It provides a convenient touchstone for politics, because someone in charge can be blamed for bringing about or failing to prevent conditions that caused life to become—obviously and on display—more expensive.

[From the May 2020 issue: An ode to driving in America]

But gasoline’s role in this dance is on the wane. Electric vehicles don’t require fill-ups, and they create far fewer emissions. So EVs also signify the end of the gas-station sign’s universal meaning. Almost nobody knows what rate they pay for electricity—or even what a kilowatt-hour of the stuff is. They certainly don’t hang a sign advertising the cost in their driveway.

Richard Jordan / Hulton Archive / Getty

A sign advertises “Gas for Less” to travelers on Historic Route 66 in Twin Oaks, Missouri, in 1989.

Carolyn Van Houten / The Washington Post / Getty

The Vega Truck Stop Punjabi Restaurant, in Vega, Texas, on February 16, 2023

EVs allow the drivers who can afford a costly car ($11,000 more on average up front than a traditional vehicle, though lifetime operating costs are lower) to no longer concern themselves with the geography, politics, and common culture of gas stations. A culture with no consumer worry about gas prices cannot gripe about them in solidarity at breakfast. A society of EV drivers doesn’t need Big Gulps or Buc-ee’s. Where will we buy our processed dessert hand pies?

Because the current oil shock is the first one to take place since EVs became widely available, the energy crisis caused by the Iran war marks the first time that a gas-price spike cannot be construed as a universal condition. Gasoline is dirty, smelly, toxic, and environmentally reprobate. But it is also strangely romantic. Even though we must give it up—even though there has never been a better time to buy an electric car—gas has long connected our automobile-dependent nation. Now the gas-station sign no longer represents a shared life and its laments.

Perhaps it can absorb a new meaning, just as its placards expressed earlier changes. Once a symbol of the everyperson’s fraught but common relationship with a commodity and the global economy it rests atop, the gas-station sign now also represents the capacity to opt out of that economy, speeding past unfazed by everything it stands for.

People Are Thinking About Looksmaxxing All Wrong

2026-03-31T09:12:07-04:00

Perhaps you’ve heard of looksmaxxing, the online trend in which young men strive to become supposedly attractive, often through self-harm. Thanks to Clavicular, a young, fringe manosphere influencer, this term—and others modeled after it—has proliferated. You can be a looksmaxxer by soft maxxing (skin care or exercise) or by hard maxxing (plastic surgery or self-mutilation). Looksmaxxers often find themselves jester-maxxing, that is, using humor to gain the attention of women.

Maxxing can be specialized, too, and even modest, maximally speaking. A dude might be personality-maxxing instead of jester-maxxing. Less incel-maxxing versions might entail health-maxxing—what people called wellness approximately 10 minutes ago. Want your gut to be more regular? That’s fiber-maxxing. Want to build bulk? You’re protein-maxxing. Some so-called tradfem women want to bear more children through fertility-maxxing—a process our culture once understood as getting pregnant again. Maxxing goes the other way too, maximizing harm instead of benefit: Maybe you’ve got a drug habit, in which case you might be pill-maxxing. Anorexia, for some, is now starve-maxxing.

Everything worth doing seems to be worth maxxing. Want to use technology less and pursue human connections more? That’s friction-maxxing. What about relaxing or zoning out? You’re nothing-maxxing. Reading is book-maxxing. Going to bed is sleep-maxxing. Buying a pair of denim shorts for the welcome spring is probably jorts-maxxing. Think you’re reading an article right now? Nah, bruh, you’re Atlantic-maxxing.

The trend is irritating and stupid, but it also betrays a truth: The online life is an extremist one, and the result is fatigue-maxxing.

As with any trend, but especially in the depths of YouTube, Reddit, 4chan, Discord, Kik, or any of the other very online places, maxxing is overblown. Not many people are saying any of these things—at least not in large numbers. Instead, they are consuming protein, getting pregnant, and even reading books.

[Thomas Chatterton Williams: ‘Looksmaxxing’ reveals the depth of the crisis facing young men]

But an idea can become potent through its rapid depiction in culture—including in articles such as this one, which maxes maxxing even as it attempts to minimize it, somewhat. Online, our odd verbal tics—“Do better,” the figurative literally, “I can’t”/“I’m dead,” “THIS”—can seem like a mania. News about a mania can help make it a reality.

Writing at The New York Times, Nitsuh Abebe argues that -maxxing as a suffix is the love child of two cursed parents: the idea of optimizing a resource, which he attributes to video gaming, and incel culture, where he locates the origin of looksmaxxing specifically.

These two origins—gaming and incel culture—make maxxing seem perverse and fringe. Critic-maxxers hope Clavicular is, as my colleague Will Gottsegen put it, a curiosity—in other words, a freak. But a freak who might be dangerous, because his words and actions could spread. Perhaps they already have: Gottsegen’s concerns, for example, were affirmed by the hateful, violent, and anti-Semitic abuse he received from Clavicular acolytes after requesting an interview with the looksmaxxer in chief.

This version of online extremism’s story is a comforting one. A bad actor becomes radicalized online by other, prior bad actors, who then spread their particular gospel of badness further, online.

Radicalization really does work this way, sometimes. Colleen LaRose, a.k.a. Jihad Jane, became immersed in jihadist recruitment forums online, joined an al-Qaeda cell in Ireland, and became entrenched in a homicidal conspiracy. Anwar al-Awlaki, an American al-Qaeda propagandist, spread Muslim extremism among the English-speaking online world. Dylann Roof internalized white-supremacist messages from websites surfaced by search algorithms, and later carried out the 2015 shooting at the Emanuel African Methodist Episcopal Church in Charleston, South Carolina. Tamerlan Tsarnaev, one of the Boston Marathon bombers, was inspired by YouTube. So were Brenton Tarrant, who killed 51 people at Christchurch, New Zealand, mosques in 2019 after absorbing far-right ideologies from online forums such as 4chan, and Jake Angeli, the “QAnon Shaman” known for his shirtless, head-dressed appearance during the January 6, 2021, attack on the U.S. Capitol.

Though distressing, these well-known individual extremists offer consolation for aspiring normals, because they represent a perversion of norms, even while also threatening to erode those norms. So do categories of softer radicalism unattached to specific figureheads: “grindset” work extremists, manosphere red-pillers, disinformation-addled elder Facebookers, or “HODL” investment perverts. Life online appears to be neatly divisible, and indeed divided, into freaks on the one hand and, on the other, reasonable people such as yourself.

It is not. Online life is extremist in deep and broad ways. The intensity of internet life, even under ordinary circumstances, pushes us toward extremes on any topic—and every topic. No idea, belief, purchase, product, or event can be ordinary or innocuous. Everything must be done with absolutism, and those extremes must be performed in public, online.

When an influencer announces that she is “obsessed” with a makeup palette, slop bowl, or imported low-cost garment, her verbal tic spreads. Suddenly ordinary people also become “obsessed” with cleansers. They discover peptides they “can’t live without.” Mild enthusiasm doesn’t travel online. Intensity does. The algorithm, and the economy that supports it, first escalates speech that overstates the emotions that underlie it. As the speech spreads, the feeling follows. People truly come to believe that they cannot live without a particular peptide or a palette.

Over time, everything becomes either life-changing or irrelevant. Moderate positions can certainly be held, but only in the ordinary life one leads quietly and calmly offline—insofar as living offline is even possible anymore.

Fans no longer simply like a television show, film, comic book, author, quick-service restaurant, or any other cultural product. Instead, their enjoyment has become ramped up to a level formerly reserved for unhinged, extremist fandoms, such as Trekkies and Beanie Baby collectors. Fans feel the need to defend a work’s true meaning. They attack deviations from agreed-upon “canon.” They cease to interpret but instead draw lines in the sand. For every cultural good, identity has become fused with the object of interest, turning previously normal people leading unremarkable lives into Steak ’n Shake beef-tallow purists, Harry Potter moralists, or cast-iron-pan-cleaning radicals.

[Charlie Warzel: This is what it looks like when nothing matters]

On Facebook or Nextdoor, a missing package can never represent a simple misfortune or misunderstanding; it must instead become urban decay or a racist incursion. A loud car heard blocks or miles away represents social breakdown. A dog let off the leash becomes indicative of moral rot. An opinion offered offhand suggests a secret wickedness that must be exorcised. Platforms reward escalation with attention, and the audience also often responds. And so everyday life becomes overinterpreted, as local forums become symbolic battlegrounds.

You no longer buy or use or encounter goods, services, or events, but endure them with your whole person: I am a Moleskine journaler; I am a Stanley-cup commuter; I am a barefoot-shoe jogger. Advice or even just notions—only check email after noon; never do 10 reps of crunches—solidify into absolutism or vanish.

The speed, urgency, and constancy of online life amplify extremism because posting, replying, and generally participating in the discourse is its own virtue enrobing all the rest, an internet-maxxing to rule all the others. Liking became fixation, watching became safeguarding, asking turned to prosecution, trying devolved to optimizing, noticing twisted into diagnosis.

Maxxing declares this state of affairs honestly. Finally, we can shed the pretense that internet life is reasonable, level-headed, or healthy. The whole internet is a machine for extremist thought, belief, and action. Maxxing could amount to its endgame—the final victory of full-throated extremism of any form and kind. But immoderacy online always ratchets up. Eventually, and probably soon, the max-maxxers will seem temperate in hindsight.

The Myspace Dilemma Facing ChatGPT

2026-03-18T07:00:00-04:00

My colleague Perry was clutching a Ben & Jerry’s ice-cream bar. Our office, which made websites at the turn of the millennium, had decided to try Kozmo.com, a new site that promised to bring DVDs, books, and treats to your door within an hour. We paid a couple of dollars for the ice-cream bar—delivery was free. Yet Perry’s eyes opened wide, the way they did when something bad was about to happen.

Kozmo.com collapsed a year later, in April 2001, along with much of the rest of the dot-com economy that employed us at the time. Today, of course, the Kozmo.com model is everywhere—DoorDash, Uber Eats, Grubhub, and other services allow you to order almost any foodstuff to your door. Kozmo.com wasn’t a bad idea so much as a badly timed one. The marketplace wasn’t mature, and the business model for delivery hadn’t been established.

The technology industry believes in the gospel of original ideas and disruptive innovation. But Kozmo.com, just one of many examples, shows how likely the first mover in a sector is to founder or fail or fall out of favor: Friendster and Myspace were first in social networking, but Facebook eradicated them. Treo and BlackBerry owned the smartphone market until the iPhone stamped them out. Early winners seem obvious and enduring, until they do not.

Will the AI market be any different? No and yes. A year ago, OpenAI’s ChatGPT was a generic name, the Coke or Kleenex of generative-AI chatbots. Today, its competitors, especially Anthropic’s Claude, are advancing quickly. OpenAI’s fall from favor looks inevitable.

[Read: OpenAI is in trouble]

But AI appears to work nothing like delivery apps, smartphones, social networks, or even computer operating systems. If ChatGPT becomes outmoded, it won’t be the result of OpenAI losing ground or failing to innovate. Instead, the entire generative-AI sector will have become a commodity, like soft drinks or facial tissues. That process has already begun.

The logic of first-mover advantage is essentially colonialist: a land grab across a newly discovered commercial wilderness. In the early days of American industry, some of that wilderness really was untamed. DuPont ushered in the modern synthetic-fiber industry with nylon; Bell Labs invented the transistor, the foundation of modern electronics.

A more boring path to success, made from standards, licensing, and distribution instead of frontier-crossing invention, proliferated in the 20th century. In the early 1980s, Microsoft pushed MS-DOS (and then Windows) onto various PC manufacturers to get the software into as many machines as possible. JVC’s VHS cassette beat Sony’s (in many ways superior) Betamax format because JVC licensed its format faster and more broadly, meaning more VCRs used the format and more video stores stocked it. In the early days of cable television, channels such as CNN and MTV sought to be carried on as many cable systems as possible. Retailers such as Walmart started saturating regions with stores, helping solidify supplier and distribution networks while slowly eroding competition from local businesses.

By the 1990s, lessons from Microsoft, JVC, CNN, and others suggested that network effects—increased value derived from the number of users of a product or service—were central to land-grab victories. At the same time, technology shifted economic growth toward immaterial goods, such as software and online services. If a platform could scale fast and first, it could own the market. This rationale drove venture-capital investment during the entire internet era, from the dot-coms through Web 2.0 and the smartphone. User growth, market share, ecosystem lock-in, and brand awareness became more important than profit or even revenue.

But success didn’t always come from being first. In many cases, what we now consider the big winner became the most familiar only after the market was established. Google was not the first search engine—Lycos, AltaVista, Ask Jeeves, and others filled that role years earlier. Google invented a better way to catalog the web, and then it built a viable advertising model atop that success. Facebook did the same with social networking; Apple did the same with smartphones. Zimride, which later rebranded as Lyft, offered ride-share services first, but Uber became dominant by expanding aggressively. Reversals of early success turn out to be far more common than first-to-market entrenchment.

Given that history, OpenAI might seem more likely to cede ground than to retain it. The company’s apparent advantage has turned out to be more fragile than expected. (The Atlantic entered into a corporate partnership with OpenAI in 2024.) But something else has changed too. Just as the online-services market differed from the electronics market that preceded it, so too does the market for AI services differ from what came just before.

A single AI victor may never emerge. Different AI companies are proving to be good at different things. Anthropic’s Claude, for example, is particularly effective at coding and at analyzing long documents. Sophisticated AI end users are using multiple services for different tasks. The various models—ChatGPT, Claude, Gemini, and so on—tend to trade places rapidly on various metrics of performance.

ChatGPT won over many ordinary consumers—college students, say, or your aging parents. But workplaces, schools, and organizations adopt technology as a part of interoperating software suites such as Microsoft Office and Google Workplace. And unlike Uber and Amazon, AI has quickly become embedded in those. The network effects that lock users into an AI service are situated inside these software packages, up the value chain from GPT, Gemini, Claude, and other AI models. The AI is a (supposedly!) thinking utility that powers old business services in new ways, but it isn’t the service—or won’t remain so, at least.

[Read: Even Silicon Valley says that AI is a bubble]

At work, AI is not a choice but a tool—welcome or not—to which employers subscribe that becomes a default. Initially, ChatGPT did the hard work of educating users on what large language models are and how they can be used. Now many consumer and business users understand what AI is and what it can do, and they have moved on to trying to figure out how to use it. Which AI they use has become less important—and perhaps less of a choice, besides.

The technology of LLMs feels disruptive and world-changing—more like nylon or iPhones than like Kozmo.com. But the big AI companies’ models have reached parity fairly quickly. Switching costs are fairly low in the AI marketplace, compared with those of videocassette recorders, smartphones, or even social networks. Whether or not AI is ethical, effective, or even fit for purpose, the act of using it appears to be naturally promiscuous.

Microsoft, Google, Facebook, Uber, and other platform-centric companies relied on winning users and holding them hostage. AI doesn’t seem to work that way. Instead, it looks more like corporate infrastructure or commodities. Companies don’t really care which cloud-computing provider they use, so long as the service is reliable and competitively priced. Firms will procure whatever hard disks, printer paper, or lunchtime slop bowls are convenient, suitable, and priced right. They seem likely to end up doing the same with AI.

OpenAI CEO Sam Altman made the provocative statement last week that in the future, intelligence will be “a utility, like electricity or water.” Instead of taking that claim as hubristic—as Altman claiming that smarts will be owned by OpenAI—consider a far more mundane and probable idea: AI could become, in just a few years, a commodity as invisible and anonymous as power or plumbing. Nobody cares what company makes the lights work or the toilets flush, so long as they do.

The Accidental Winners of the War on Higher Ed

2026-01-29T07:00:00-05:00

Updated at 11:07 a.m. ET on January 30, 2026

In the waning heat of last summer, freshly back in my office at a major research university, I found myself considering the higher-education hellscape that had lately descended upon the nation. I’d spent months reporting on the Trump administration’s attacks on universities for The Atlantic, speaking with dozens of administrators, faculty, and students about the billions of dollars in cuts to public funding for research and the resulting collapse of “college life.”At the same time, I’d been chronicling the spread of AI-powered chatbots that have already changed undergraduates forever.

This article was featured in the One Story to Read Today newsletter. Sign up for it here.

Initially, I surveyed the situation from the safe distance of a journalist who happens to also be a career professor and university administrator. I saw myself as an envoy between America’s college campuses and its citizens, telling the stories of the people whose lives had been shattered by these transformations. By the summer, though, that safe distance had collapsed back on me. The wave of campus crises was causing immediate effects in my own academic life. People I knew were losing their jobs. Yearslong projects were stranded. I heard from students cheating their way through every class with AI, and from faculty following suit. I watched professors lash out at other disciplines with misplaced blame. I listened to friends discuss whether they should send their kids to school abroad. I had personally devoted 25 years to higher education. Now it felt like the whole idea of the university might be ending.

Sitting in my office, I began searching for some cause for hope, some reason to believe that higher ed could stanch the damage for the next generation of students. It occurred to me that I’d been hearing less despair from colleagues at certain smaller schools that offer undergraduate study in the “liberal-arts tradition,” a broad and flexible approach to education that values developing the person over professional training. I wondered if these schools—especially the wealthy ones that cluster near the top of national rankings—might enjoy some natural insulation from the fires raging through the nation’s research universities.

I texted, emailed, telephoned, and Zoomed with friends in higher-education leadership. Current and former heads of both research universities and liberal-arts colleges confirmed my intuition: Well-resourced and prestigious small colleges are less exposed in almost every way to the crises that higher ed faces.

To dig deeper, I decided to embark on a fact-finding mission in the guise of a traditional college tour. In November, I went to visit four elite liberal-arts colleges—Amherst, Davidson, Smith, and Vassar—where I joined prospective students and their parents for their campus tours, sat in on classes, and spoke with undergraduates, professors, and administrators. Might schools such as these emerge as the accidental winners of the war on universities?

I came to Amherst College too late in the autumn to observe peak foliage, but amber and ocher remnants still spread across the Pioneer Valley in western Massachusetts. This is the scene Emily Dickinson, a local resident, would have taken in before composing the lines, “The maple wears a gayer scarf, / The field a scarlet gown.”

My tour began at the Colonial Revival residence that serves as the school’s admissions office, but we soon arrived in the new science building, a modern structure with oddball green furnishings. Nate Scharf, our sophomore guide, sat us down and gestured toward a small classroom where he had taken a psychology class. At a typical research university, an undergraduate course like this would meet inside a lecture hall, perhaps with separate breakout sections for discussion led by graduate students. But neither Scharf nor any of his classmates had ever been exposed to this scenario—because Amherst, like most liberal-arts colleges, has no graduate students.

No graduate students: The American public may not fully grasp the implications of this statement. At a university, grad students may end up leading class discussions, but that’s only part of the story. Their presence on a campus telegraphs a more important truth: that in terms of mission, teaching undergrads comes second to research.

[Read: Grad school is in trouble]

And at most universities, grad students play a crucial role within the research system: They perform the frontline work of science. Faculty members get federal grants, which are used to pay for doctoral students, who in turn serve as laboratory staff. Professors’ feeling of worth and productivity may be a function of how many doctoral students they advise—because that helps determine how many studies they can carry out, how many papers they can publish, and what sorts of new grants they can win to keep the process going. This endless competition is a major feature of the modern university—whether it’s an elite private school (such as Princeton or Duke), a public flagship (such as the University of Michigan or UC Berkeley), or a land-grant institution (such as Texas A&M or Virginia Tech).

A school like Amherst, though, which has no doctoral programs whatsoever, is free of the rat race of research productivity and expenditure. As these colleges like to point out, that’s good for undergrads, because faculty must focus on education. The lack of doctoral research programs also makes the schools more resilient to bullying from Washington. In 2025, the Trump administration made a point of suspending hundreds of millions of dollars in research grants to Columbia, Harvard, Northwestern, and other schools. It also threatened to impose a tighter limit on what are called “indirect costs”—the portion of each grant that covers all the broader costs of running a large enterprise for research. With so much funding endangered all at once, targeted universities had little choice but to negotiate—which is to say, to accede to some portion of the Trump administration’s demands.

At Amherst, this level of pressure simply couldn’t be applied. In 2024, the college took in around $3 million from all of its federal research grants put together. (For comparison: Washington University in St. Louis, where I teach, received $731 million that year from the National Institutes of Health alone.) Christopher Durr, an associate professor of chemistry, told me that his research is “designed to be done mostly in-house,” by which he means without external funding, and with the help of undergraduates. He studies biodegradable plastics in a lab that is mostly funded by the college itself, and he gets no research money at all from the federal government. Settling into an Eames-style chair in his office, I asked him what indirect-cost rate Amherst applies to its grants, and he said he didn’t know—an innocence that would be unheard of at a research university.

The research that can be done at a small liberal-arts college, with its more modest labs and equipment, is necessarily constrained. Even with Amherst’s generosity, Durr said “there’s a limit” to how far he can push his research. In truth, the most important scientific and medical discoveries aren’t likely to be made at a place like Amherst or Smith, the nearby women’s college, which tend to pay their own students to work on faculty research. But this need not be a limitation for undergraduates. The conditions that produce landmark discoveries are not necessarily the same ones that produce a serious education.

Perched along the shores of North Carolina’s Lake Norman, in the Charlotte suburbs at the northern edge of Mecklenburg County, Davidson College’s location is a crossroads for red and blue America. In the last election, two-thirds of Mecklenburg’s voters chose Kamala Harris; just a mile up the road, in Iredell County, two-thirds of voters went for Donald Trump.

Connor Hines, the current student-body president and a senior majoring in political science, is a self-described moderate conservative who grew up in the area. He told me that he chose to come to Davidson in part because its students are able to disagree with one another in thoughtful ways. His good friend Nina Worley, an educational-studies major who was raised in Harlem, said the same. “I knew that to be a good advocate for education policy, I needed to meet a Republican for the first time,” Worley told me. She and Hines do not see eye to eye on everything—the two do not agree, for example, on the policies of Alexandria Ocasio-Cortez. But they said they carry out that disagreement with respect.

[Read: The end of college life]

According to higher education’s many critics on the right, elite universities have tried to stifle disagreement; allegedly they’re pushing students into echo chambers instead of broadening their minds. The Trump administration made this accusation central to its war on research universities: With billions of dollars in withheld grants as leverage, it initially pressured nine institutions into signing an agreement, called the “Compact for Academic Excellence in Higher Education,” that would commit them to “fostering a vibrant marketplace of ideas on campus” and protecting conservative viewpoints.

In the meantime, under local pressure, reactionary speech suppression is taking hold at public universities: Many schools, including Auburn and the University of Michigan, have shut down their DEI offices; Texas A&M censored courses based on race and gender content; and Indiana University halted the print edition of the student newspaper, an act its journalists said amounted to censorship. Some private schools, including Columbia, Northwestern, and Cornell, have struck deals to restore funding that variously included provisions regarding civil-rights standards, disciplinary processes, DEI rubrics in admissions and hiring, and policies related to sex and gender identity.

Davidson, which, like Amherst, is far less exposed than research universities to the threat of withheld money, has not been explicitly asked to sign the compact. (Nor have any other liberal-arts schools, including the ones I visited.) Chris Marsicano, an associate professor of education studies who had introduced me to Hines and Worley, told me that the school was way ahead of the national conversation on campus dialogue in any case. In 2019, it created a Deliberative Citizenship Initiative, which runs programming to help students and faculty engage on issues about which disagreement is likely, such as immigration and border control. “We’re the hipsters of civil discourse,” Marsicano said. “We were doing it before it was cool.”

Certainly these sorts of efforts can be found at lots of universities: Ohio State has a Civil Discourse for Citizenship initiative, for instance, and WashU has just launched an Ordered Liberty Project, which includes associated coursework on the topic of “disagreement.” But they may be most likely to succeed when they’re applied to smaller student bodies, like Davidson’s. At lunch in the school’s campus commons, I watched President Douglas Hicks greet a dozen students by name. My tour guide explained that students and faculty tend to mingle there; in fact, they can’t avoid it, because everyone eats in the same place. Some might find this stifling or claustrophobic, but being forced to live together day by day may also stimulate openness and mutual respect. “You can’t hide,” Hicks told me.

[Read: Texas sends Plato back to his cave]

At a small liberal-arts college, where a cohort may number fewer than 500 people, admissions officers can also take a stronger hand in assembling a group of students who match the institution’s culture and its vibe while also having very different backgrounds. And the fact that almost everyone at a small liberal-arts college tends to live on campus, or very close to it, adds to the sense of intimacy. “It’s just much easier for me to get to know faculty here, much easier for me to get to know students, much easier for me to hear what’s on their minds,” Amherst’s president, Michael Elliott, told me.

One effect of this, he said, is that professors actually show up to faculty meetings to talk about the future of their institution. They participate in budgeting conversations, debate the creation of majors, and approve new courses. This is decidedly not the norm at many larger universities, where professors may not see these meetings as a core part of the job, and where administrators can ignore them altogether. (Last year, Northwestern reached a settlement with the Trump administration despite 98 percent of its voting faculty having recommended otherwise.) Elliott admits that all of this faculty participation can make the college move more slowly in decision making. But done right, faculty governance is a form of healthy campus discourse, too.

Perhaps no threat to higher ed is more acute than the recent, rapid spread of generative AI. “It’s just about crushed me,” one college teacher told me in the spring of 2023. “I fell in love with teaching, and I have loved my time in the classroom, but with ChatGPT, everything feels pointless.”

The remarkable test-taking, paper-writing power of AI is hardly under control at liberal-arts colleges such as Amherst or Davidson. But in my conversations with students and faculty, I witnessed a struggle to understand and respond to it with greater earnestness and complexity than I have encountered elsewhere in higher education. In my experience, students and faculty at research universities tend to have a Manichaean view of the technology: Either AI is an existential threat to be cast out, or a vital tool to be embraced as a professional virtue. Little room is left for nuance, which means that little progress can be made on shaping policy.

Davidson prides itself on having an unusually deliberate honor code. (Students I spoke with said this code is taken so seriously that they can leave their belongings anywhere on campus without fear that they will be stolen.) But the seductions of ChatGPT are hard to resist, and Marsicano noted that the college has seen an increase in code violations due to AI. That sounded like less of a problem here than elsewhere, though. If the students are availing themselves of the technology, then at least they appear to be doing so with some reservations.

Hines told me that he uses AI tools, though not for academic work. Worley said she’s avoided AI so far, because “I’m still struggling with what makes me comfortable about it and what doesn’t.” At Davidson, each professor can set their own approach to AI in the classroom, but that can be confusing. In the meantime, Worley and her roommates ponder what the rules should be. They ask, for instance: If it’s now okay to have AI write your emails for you, why is that okay? And they talk about AI data centers’ environmental costs. “Everyone’s wrestling with it,” she said.

Learning how to wrestle with an issue like AI, which is ambiguous and undecided, is the whole point of an education undertaken in the liberal tradition. It’s also made easier in a context where faculty-to-student ratios can be much smaller than they are at even elite research universities. (At a school like Davidson or Amherst, some classrooms will contain just a handful of students and a professor—a scenario in which siphoning work through Claude would not be easy.) Carina Cole, a Vassar media-studies student, told me that a supportive culture on campus also makes it possible to treat AI with greater care. Her fellow students are more likely to ask one another for help than turn to technology, she said. But Worley admitted that the situation feels fragile no matter what: “I’m holding on for dear life.”

At Amherst, I sat down with Matthew McGann, the dean of admission and financial aid. I asked how his school stacked up against the nation’s most prestigious universities. If a student also gains admission to Yale, do they go there instead? “They do,” he said. Amherst has a sterling reputation, but it still cannot compete against a famous Ivy.

The relative obscurity of even the best small liberal-arts colleges may be helping them to avoid the Trump administration’s war on higher ed. But that doesn’t mean that they’re immune from any risk of being targeted. Haverford College, in Pennsylvania, for example, has been made the subject of a civil-rights investigation for allegedly failing to address anti-Semitic harassment against Jewish students. (This is the same charge, and the same government action, that has been directed or threatened against Columbia, UCLA, Stanford, Cornell, Rutgers, and many other big universities.) Haverford may not rely on the federal government for giant research grants, but even just defending itself against investigation would be costly, as would, say, losing federal Pell Grants that subsidize low-income students.

Small colleges’ endowments could also be attacked, even if the schools are never targeted specifically. In 2025, Congress voted to increase the tax on endowment income, which private colleges and universities draw on for financial aid, from 1.4 percent to as much as 8 percent. (The wealthiest liberal-arts colleges spend a large portion of their endowment earnings to fund students who can’t afford to spend up to $100,000 a year to attend.) For the moment, institutions with fewer than 3,000 tuition-paying students—which means most liberal-arts colleges—are exempted from this increase, so their budgets remain untouched. But that could change.

Then there is the nagging question of practicality. Even if you believe that a liberal-arts college offers the best education, going to school to learn how to think might seem like a luxury today. In the end, you’ll still need to earn some kind of living. If the paths for getting there—which may include postgraduate study in a doctoral program or professional school—are diminishing, then college itself will follow suit.

Still, after spending several weeks on my tour of wealthy, liberal-arts colleges, I grew to think that the pitch they’re making to prospective students and their parents for the fall of 2026 was convincing. All things considered, when the time comes, I might rather see my own preteen daughter attend a school like Amherst or Davidson, Smith or Vassar, than a research university such as my own. The form of higher ed that they provide seems poised to be the most resilient in the years to come.

In a way, that resilience may pass on to the students who attend them. On my last night in western Massachusetts, I was set up to have dinner with a group of graduating Amherst seniors. By all measures, these were tremendously successful young adults. Hedley Lawrence-Apfelbaum intends to complete a fellowship at the University of Oxford on his way to Harvard Law. Ayres Warren, who has already published a journal article on breast-cancer disparities, will pursue a career in environmental health-care management. Shane Dillon is plotting out a career in politics. All three of them told me they were worried about their futures—not because they had attended Amherst instead of Yale or Emory or Michigan State, but because the world seems so uncertain on the whole.

These seniors weren’t panicking, though—this was very clear. “We’ll see,” Dillon said when I asked what was next, what he feared, and what he hoped for. I’ve known a lot of brilliant undergraduates during the course of my career at research universities. Some have become wealthy in business; others have become famous; still more have made the world better through research and invention. But Dillon and the other students whom I met during my college tour struck me as being unusually attuned to the facts of their predicament. They were already living in the thickness of their own lives, aware of who they are and not just what they might become. They seemed to understand, by virtue of the education they’d received, that uncertainty is to be expected, and that the future lasts a long time.

This article originally implied that Hedley Lawrence-Apfelbaum had already received admission to Oxford.

The Roomba Was a Disappointment

2025-12-17T12:47:18-05:00

The home-vacuum robot began, like most things, with war. In August 1990, the same month and year Saddam Hussein invaded Kuwait, three MIT roboticists incorporated the company that would eventually become iRobot, the maker of the Roomba.

In its first decade, iRobot began to assemble a small-droid A-team for the theater of combat. The Ariel defused mines; the PackBot handled bomb disposal. (Later they would be joined by the Warrior, which breached obstacles; the camera-encrusted SUGV, which handled recon; and the palm-size FirstLook, which could be thrown through a window to investigate hazardous materials.) These machines weren’t weapons, but they facilitated weaponry’s consequences. At the turn of the millennium, iRobots might be seen on cable news kicking up Iraqi dust, investigating suspicious domestic packages, and probing the ruins of the World Trade Center.

From their armature, iRobot built Roomba in 2002—a domesticated robot that surveilled, detected, and removed materials from wood floors and carpets. Roomba created the U.S. market for home vacuum robots, and home robotics in general, insofar as a general home robotics was ever realized. Eventually, things went south. Competitors took over. A 2022 acquisition by Amazon failed on antitrust grounds, and this week iRobot filed for bankruptcy. Roombas will continue to meander across wool and laminate, but under the control of iRobot’s Chinese manufacturing partner.

Roombas never really worked, not well. They got stuck on the lips of rugs. They tumbled down stairs their sensors failed to see. Cluttered spaces—which is to say, homes and apartments—caused Roombas to struggle. They tripped over electrical cords, they failed to clean in corners, and they failed to pick up debris beyond fine dust. Because of all this, Roomba owners sometimes felt the need to tidy up in advance of robotic cleanings, or to add a round of human cleaning afterward. The high-tech vacuum could seem like a wayward pet.

The Roomba also shared some DNA with its military forebears. The device builds an internal map of the home it struggles to clean, and iRobot disclosed that these ghost maps could be shared with partners. In 2022, images leaked from Roomba test units outfitted with cameras, which showed real people in their homes, including one woman on the toilet. Depending on the whims of its new owner, the innocuous robot could become a spy. Some worry that Roombas will become a swarm of covert, foreign operatives.

Roomba’s affliction inverts the hopeful spirit that it and other robots of the early aughts once held. In 2001, Segway promised to reinvent cities with its electrified stand-up scooters. Sony’s AIBO, an expensive robot dog, was meant to offer both companionship and entertainment. LEGO Mindstorms promised to teach kids to program robots. But most such products either failed or drifted into unexpected niches. (The Segway lingers as a vehicle for sightseers and mall cops.) New ones have appeared—the costly Roomba-style lawn mowers I now see in my neighborhood, for example, or preposterous humanoid AI butlers such as the 1X Neo and the as-yet unreleased Tesla Optimus, which cost as much as a small car. By any reasonable standard, the dream of everyday robotics has largely failed.

[Read: A pocket guide to the robot revolution]

Why? Because software was easier. After 2004 or so, when the economy had recovered from the dot-com crash, consumer growth and corporate profit came from scaling low-cost, data-driven software services: Google, Facebook, LinkedIn, Salesforce, YouTube, Skype, even WordPress and Blackboard. Innovation in machinery was redirected toward personal-media and communication devices. Once the smartphone was invented, it produced another surge of data-driven software products.

For most consumers, robots also promised (and delivered) far less utility or joy than, say, a social network, which was much simpler to build and run. But the robots still had value in the contexts that begat them: warfare and espionage. iRobot continued selling military products, and spun off that business in 2016. Sony discontinued its original entertainment robots, and repurposed their technologies for other consumer products. Industry found applications, too: Larger-scale robots became commonplace in factories and warehouses, where the environments are standardized and automated labor can be overseen. And iPad-on-a-stick telepresence robots became popular in hospitals, where eventually they helped deliver last rites to COVID-19 patients in quarantine.

The humble Roomba is important as a symbol of everyday robotics’ promise and its failure. Whether in The Jetsons or Star Wars, the idea that cute, autonomous devices could live inside your home has persisted. It is time to abandon that dream for good. Robots are for battle, and robots are for spying, and robots are for places where humans will not or cannot be. They are not your friends, and they will never clean your floors.

‘Six-Seven’ Is Six Feet Under

2025-12-12T08:11:06-05:00

This article was featured in the One Story to Read Today newsletter. Sign up for it here.

It goes like this: When six and seven appear together in print, or in speech, or on television, or in a YouTube video, or even just when you write them down on loose-leaf paper, that’s “six-seven.” “Six-seven!” you say, you probably being a middle-school-age child. Such is the youth phenomenon known by this name. Now you know, but chances are you already did, especially if a preteen has lived in your house anytime since this spring.

Six-seven-ing seems to have peaked around Halloween, and now, as the holidays descend, its days are numbered. My own 11-year-old never liked it (a culture of contrarianism pervades the Bogost residence), but now she actively scorns it. “The memes will reset on New Year’s Day,” she recently announced. I hear the same from other parents of kids her age. Worse, parents are now saying “six-seven” (as are sports leagues and fast-food chains), which is of course fatal for anything that kids find cool. But six-seven might have been doomed by design. It just wasn’t built to last, not as a staple of kid-dom anyway.

People have referred to six-seven as a meme, but that’s not a helpful way to understand it. Yes, the tic spread widely, starting in March, thanks to a few popular online videos, but what doesn’t spread online these days? Better: Take six-seven as an example of “childlore,” a name for folk culture that arises or spreads among children. Jump-rope rhymes such as “Mary Mack” are childlore, as is doggerel slander such as “Fatty Fatty Two-by-Four.” So is the concept of cooties, a fictional disease caught when someone gross touches you. Likewise the “cool S” doodle, that squared-off figure that looks like a vertical, pointy-ended infinity symbol; paper fortune tellers (a.k.a. cootie catchers); the soothsaying paper-and-pencil game MASH; playground rituals (“No take-backs!”); and received superstitions (“Step on a crack, break your mother’s back”).

[Read: Rage bait is a brilliant word of the year]

Childlore thrives when two conditions are met. One is a murkiness of origins. Nobody knows who first folded a cootie catcher or drew a “cool S.” Some childlore feels like it has always existed, stretching back into eternity and reproducing itself as legend. Other trends arrive as if air-dropped by aliens. This doesn’t bother kids, because everything is new for them. Learning an arbitrary playground ritual or a chant for the numbers between five and eight is not much different from discovering that red pandas exist, or how to plot points on coordinate planes, or that Goldfish crackers also come in pretzel flavor. The 10-or-so-year-olds who latched on to six-seven did so just because it wormed through their ears into their brain and then departed out of their mouth.

But obsessional online culture has sought to answer questions of origin rather than waving them off. Anything can be researched and explained today, whether as historical fact or invented conspiracy. Doing so misses the point. Six-seven supposedly originated in a Skrilla song, but knowing that fact offers no further understanding. To trace a children’s phenomenon to its explanation also ruins it. Who cares where it came from? For kids, the point is that it is here, and it is theirs.

The second condition is a network. Adults may construe a child’s life as unencumbered, but kids actually have very little control over what happens to them. They thrive or wither at the whims of teachers, parents, and other adults who control their hours and days. Within those constraints, childlore offers kids an opportunity to develop and control a local culture. Another kid teaches you the jump-rope rhyme. You watch one draw a doodle and then do it yourself. Six-seven started this way, and then it circulated through the tender, sticky hands of children.

Alas, what started in their network soon became a social currency for adults online. Gen Xers and Millennials—the parents of today’s middle schoolers—couldn’t leave six-seven be; they’re accustomed to burning through such memes like their parents or grandparents did cigarettes, as if culture writ large can withstand habitual abuse. Social-media-influencer culture also latched on to the phrase for its own attentional ends, as did brands, the indefatigable scavengers. These forces stole six-seven from the kids who had nurtured it.

But six-seven’s fragility is just as much a product of its structure. The numbers have no meaning of their own. Six-seven is just a lasso looped by fate around two adjacent integers on the number line. It hides no secret payload of violence, sex, sacrilege, or anything whatsoever. This emptiness surely helped six-seven’s rise. Hearing it might irritate parents or teachers, but that irritation has no cause and therefore merits no reproach. What are you going to say—Stop naming whole numbers? (Some schools have said just that.)

Some successful forms of childlore—like bubble-shaped handwriting, or the way to fold a notebook page into a note for passage to a friend—were more resilient because they were material, and bounded by circumstance. They had to be inscribed onto or made from paper, for example, or to arise in particular social contexts. Six-seven just floats loose. It can be uttered for any reason or for no reason. My older daughter, who is in her early 20s, admitted that she recently found herself saying “six-seven” involuntarily upon hearing a co-worker refer to “six or seven” of something. Like one must say “cow” upon seeing a cow along a roadside, prior knowledge of six-seven demands releasing its utterance when numerically provoked. “I didn’t put my heart into it,” my older daughter said. “But what’s important is: It wasn’t a choice. It just came out of my mouth.” Six-seven was so universal by nature that it was bound to spread into the world of adults, which suffocated it.

Teachers, parents, and even some kids are pleased to welcome six-seven’s end. “No one really says it anymore,” a friend’s 9-year-old reports; my own kid longs for this report to be true. Others are more circumspect. It’s not dead but “slowly dying,” according to another friend’s preteen daughter. “No offense, Mom, but parents don’t understand,” she added. Instead, both kids explained, their peers have started saying “41”—a competing meme that carries six-seven’s ghost (six times seven, plus six, minus seven)—as if to mourn it. Six-seven’s end brings relief, but also sorrow: Childlore, a staple of boy- and girlhood for centuries, has become tenuous and fragile. Being a kid is hard, and it isn’t getting any easier. Even the silly trappings of youth now feel as fleeting as everything else.

Get Your Kid a Watch

2025-11-28T08:00:00-05:00

Because of time’s arrow, my daughter, who was once a toddler, is now a preteen. A new question thus arises: When should I let her get a smartphone? This problem isn’t new to me. I have two older kids, now in their 20s. Back in the day, I bought each of them an iPod Touch—essentially, a smartphone without the phone—when they were about her age, and then the full device at around the start of high school. But online life was different then. There was less pressure to be smartphone-connected all the time. Social media wasn’t yet as ubiquitous, or worrisome, as it is today. Now the stakes seem higher.

Today smartphones are as widespread as the concerns about their effects on young people’s brains. Psychologists have written best-selling books about how bad phones are for kids, and many schools have banned their use. Despite all this, no one can dispute the fact that phones and phone apps have entered every aspect of contemporary life. Even Jonathan Haidt, who aims to end the phone-based childhood, floats policies that would allow for a phone-based adolescence. The question is not whether your kid will ever get a smartphone, but rather how to manage its adoption in a way that will preserve the integrity of child, parent, school, and home life. And to that end, I believe I’ve found a good solution: Get your kid a watch.

That idea had not occurred to me until my daughter brought it up. She’d been FaceTiming with a friend who had just received an Apple Watch. Now my daughter wanted one, and it didn’t take long for me to acquiesce. After all, as a small device with fewer features, a smartwatch would have to do less damage than a standard smartphone. Maybe it would also do substantial good. The smartwatch might allow her to connect with friends and family, while keeping her away from social media.

[Read: You’re getting ‘screen time’ wrong]

I ordered her an Apple Watch that very day. In theory I’d been open to another sort of product—a smartwatch that is specifically designed for kids—but the competition barely registered. The market for children’s smartwatches has been flooded for years with garbage. Many products of this type are toys, and crappy ones at that: hunks of cheap plastic with poor displays and valueless software; Dick Tracy novelties for a generation that has never heard of the guy. The next tier up includes more functional devices with network connections, such as the Gizmo Watch. But that product, like many others in the category, caters to adult control. Technically, the Gizmo can be used to exchange text messages and calls, but only with a contact list that is managed by a parent. The device’s main function for a kid is passive: It allows her to be called or texted by her parents, and tracked by them via GPS. This is a house-arrest bracelet, not a smartwatch.

At the risk of devolving into “when I was a kid”–ism, when I was a kid, we learned how to use technologies through actual use. There were few phones or televisions or stereos for kids—instead, just phones, televisions, and stereos. The ownership, location, and operation of these devices was subject to the oversight of parents, who also gave their children direct and deliberate instruction on the devices’ proper use. I was taught how to dial a phone, but also what to say or not say on one, for example. And parents spent considerable thought on questions such as whether telephones should be in children’s rooms. Then, as now, their minds were on potential harms. What’s new today is the sense that nothing can be done to mitigate these harms aside from wholesale prohibition.

If I was going to do this, I wanted to get my daughter a fully operational smartwatch, and not some kiddie version that wouldn’t really help her learn how to navigate the computerized world. To some extent, I wanted her to confront the capabilities, confusions, and risks of online life, so she could learn how to manage them herself. I have owned and used smartwatches for some time, and I surmised that their many limitations compared with smartphones—and the uselessness of most of their apps—would make one a perfect candidate for this process.

We’re Apple users in my house, so the Apple Watch made sense, but similar options are available for Android, including Samsung’s Galaxy watches. The Apple Watch SE was the cheapest option, and as with any Apple Watch, you can set it up for a family member who does not own an iPhone. For that to work, you need to buy the more expensive cellular model, which permits your kid to call, text, and email from almost anywhere. It also lets you track their location. The latter function has a quirk: My kid also has an iPad, and Apple seems to treat that device, which stays home all the time, as her default location. At first I found this defect annoying, but soon I came to appreciate it. I almost never really need to know where she is, and the habitual pursuit of her geospatial data would feel like an invasion of the autonomy that the watch was meant, in part, to increase.

I’ve written in the past about the pleasures of installing a landline—a home phone that could be used by the family as a whole, rather than its individual members. For my daughter, the landline was a source of confidence that she could contact her mother or me, or a neighbor—or, God forbid, an emergency service—if she needed to. Our home phone played a similar role for me as well.

[Read: America gave up on the best home technology there is]

The smartwatch offers something more. Most communication is not done in emergencies, but in ordinary life: I’m running late or Meet me at the other door or Dinner’s ready. The ability to exchange mundane information from afar—even from across the street at a friend’s house—is part of being a whole person in the world today. Ashley James, the mother of my daughter’s friend, told me that she’s been delighted by her daughter’s usage of the smartwatch: When her daughter sees an Apple News story that she thinks might interest James, for example, she sometimes sends it in a text. James also said that her kid now texts extended-family members, developing connections that might not have materialized otherwise. Just having the device, James told me, makes her daughter feel included in the world of technology “that kids want to be a part of so badly.”

In a way, it is strange to talk about a 10-year-old this way. When I was 10, a newspaper would have been sitting on the breakfast table, and I could have shown an article to my mother at any time. But then life became digitized, and now you need a device of some kind just to see the news. Like it or not, becoming a person in the 2020s means becoming a user of computers. It also means figuring out how to express yourself online.

I’d experienced my own revelation about my daughter once she started using the Apple Watch. Back when she had just her iPad, I’d concluded that she was terrible at texting. We have a family group chat, and she would either respond to messages with a single word, or not respond at all. But after she got her watch and learned to tap out texts across its tiny screen, her messages exploded into wry quips and fully formed ideas. She turned out to be a killer texter. I quickly surmised the prior problem: She mostly uses her iPad to watch streaming shows. All those texts were interrupting her! Imagine if your text messages kept popping up on your television. She was already old enough to express herself online in sophisticated ways, but until she got the smartwatch, she didn’t have the tools to do so.

I have since concluded that the smartwatch is an unalloyed good. James seems to agree. With these devices on our daughters’ wrists, our children feel a part of the world of portable, personal technology, even as the devices offer them just modest access to that world. They’re connected, but also free of the social-media posting and scrolling that is the real cause of anxiety about kids and phones.

I find it startling that Apple and other tech companies haven’t leaned even further into this obvious opportunity, to bill the watch as a sort of training tool for life online. (I did see an advertisement in one of my daughter’s magazines for a children’s-smartwatch brand called Cosmo—described, a little weirdly, as “the perfect first phone.”) What a shame that so much effort is devoted to providing parents with all manner of controls for their kids, but scarce support. The well-timed and thoughtful introduction of a smartwatch could help mitigate concerns about children’s smartphone use while also providing them with a scaffolding on which to learn basic digital-life skills.

For the moment, though, the smartwatch is too often lumped together with the smartphone, as if they were different causes of the same disease. On this logic, many schools ban both. But such prohibitionism is reliant on magical thinking: It assumes that kids of some arbitrary age can be suddenly trusted to use smartphones, so long as they’ve spent their prior years in full digital quarantine. That’s not how things work. Kids must be introduced into connected life, one step at a time.

Welcome to the Slopverse

2025-11-26T07:30:00-05:00

This story is part of a series marking ChatGPT’s third anniversary. Read Charlie Warzel on the precarity that ChatGPT introduced to the world, Lila Shroff on the people who can no longer make decisions without ChatGPT’s input, or browse more AI coverage from The Atlantic.

Bill Lowery, a sales executive, is confused when a workmate asks where he should take a date out for dinosaur. “You’re planning to take this girl out for dinosaur?” Lowery asks. “That’s right,” the colleague responds, totally nonchalant. Lowery presses him, agitated: “Wait a minute. You’re saying dinosaur? What is this, some sort of new-wave expression or something—saying dinosaur instead of lunch?” When Lowery returns home later in the day, his wife reports on their sick son while buttering a slice of bread. “He’s so pale and awfully congested—and he didn’t touch his dinosaur when I took it in to him.” The salesman loses it.

This is the premise of “Wordplay,” an episode of the 1980s reboot of The Twilight Zone. As time progresses, people around Lowery begin speaking in an even more jumbled manner, using familiar words in unfamiliar ways. Eventually, Lowery resigns himself to relearning English from his son’s ABC book. The last scene shows him running his hands over an illustration of a dog, underneath which is printed the word Wednesday.

“Wordplay” offers a lesson on the nature of error: Small and inconspicuous changes to the norm can be more disorienting and dangerous than larger, wholesale ones. For that reason, the episode also has something to teach about truth and falsehood in ChatGPT and other such generative-AI products. By now everyone knows that large language models—or LLMs, the systems underlying chatbots—tend to invent things. They make up legal cases and recommend nonexistent software. People call these “hallucinations,” and that seems at first blush like a sensible metaphor: The chatbot appears to be delusional, confidently asserting the unreal as real.

But this is the wrong idea. Hallucination implies that a mistake is being made under a false belief. But an LLM doesn’t believe the “false” information it presents to be true. It doesn’t “believe” anything at all. Instead, an LLM predicts the next word in a sentence based on patterns that it has learned from consuming extremely large quantities of text. An LLM does not think, nor does it know. It interprets a new pattern based on its interpretation of a previous one. A chatbot is only ever chaining together credible guesses.

[Read: The AI mirage]

In “Wordplay,” Lowery is driven mad not because he is being lied to—his colleague and wife really do think the word for lunch is dinosaur, just like a chatbot will sometimes assert that glue belongs on pizza. Lowery is driven mad because the world he inhabits is suddenly just a bit off, deeply familiar but jolted from time to time with nonsense that everyone else perceives as normal. Old words are fabricated with new meanings.

AI does invent things, but not in the sense of hallucinating, of seeing something that isn’t there. Fabrication can mean “lying,” or it can mean “construction.” An LLM does the latter. It makes new prose from the statistical raw materials of old prose. The invented legal case and the made-up software are not actual things in the real universe but credible—even plausible—entities in an alternate universe. They are, in another word, fictional.

Chatbots are convincing because the fictional worlds they present are highly plausible. And they are plausible because the predictive work that an LLM does is extremely effective. This is true when chatbots make outright errors, and it’s also true when they respond to imaginative prompts. This distinctive machinery demands a better metaphor: It is not hallucinatory but multiversal. When generative AI presents fabricated information, it opens a path to another reality for the user; it multiverses rather than hallucinates. The fictions that result, many so small and meaningless, can be accepted without much trouble.

The multiverse trope—which presents the idea of branching, alternate versions of reality—was once relegated to theoretical physics, esoteric science fiction, and fringe pop culture. But it became widespread in mass-market media. Multiverses are everywhere in the Marvel Cinematic Universe. Rick and Morty has one, as do Everything Everywhere All at Once and Dark Matter. The alternate universes depicted in fiction set the expectation that multiverses are spectacular, involving wormholes and portals into literal, physical parallel worlds. It seems we got stupid chatbots instead, though the basic idea is the same. The nonexistent legal case that AI suggests could exist in a very similar universe parallel to our own. So could the fictional software.

The multiversal nature of LLM-generated text is easy to see when you use chatbots to do conceptual blending, the novel fusion of disparate topics. I can ask ChatGPT to produce a Charles Bukowski poem about Labubu and it gives me lines like, “The clerk said, they call it art toy, / like that explained anything. / Thirty bucks for a goblin that grins / like it knows the world’s already over.” Even as I know with certainty that Buk never wrote such a poem, the result is plausible; I can imagine a possible world in which the poet and the goblin toy coexisted, and this material resulted from their encounter. But running such a gut check against every single sentence or reference an LLM offers would be overwhelming—especially given that increasing efficiency is a major reason to use an LLM. Chatbots flood the zone with possible worlds—“slopworlds,” we might call them, together composing a slopverse.

[Read: AI’s real hallucination problem]

The slopverse worsens the better the LLMs become. Think about it in terms of multiversal fiction: The most terrifying or uncanny alternate universes are the ones that appear extremely similar to the known world, with small changes. In “Wordplay,” language is far more threatening to Bill Lowery because familiar words have shifted meanings, rather than English having been replaced by a totally different language. In Dark Matter, a parallel-universe version of Chicago as a desolate wasteland is more obviously counterfactual—and thus less uncanny—than a parallel universe in which the main character’s wife had not given up her career as an artist to have children. Parallel universes that wildly diverge from accepted reality are easily processed as absurd or fantastical—like the universe in Everything Everywhere All at Once where people have fingers made of hot dogs—and familiar ones convey subtler lessons of contingency, possibility, and regret.

Near universes such as the one Lowery occupies in The Twilight Zone can create empathy and unease, the uncanny truth that life could be almost the same yet profoundly different. But the trick works only because the audience knows that those worlds are counterfactual (and they know because the stories tell them directly). Not so for AI chatbots, which leave the matter a puzzle. Worse, LLMs are functional rather than narrative multiverses—they produce ideas, symbols, and solutions that are actually put to use.

The internet already acclimated users to this state of affairs, even before LLMs came on the scene. When one searches for something on Google, the resulting websites are not necessarily the best or most accurate but the most popular (along with some that have paid to be promoted by the search engine). Their information might be correct, but it need not be in order to rise to the top. Searching for goods on Amazon or other online retailers yields results of a kind, but not necessarily the right ones. Likewise, social-media sites such as Facebook, X, and TikTok surface content that might be engaging but isn’t necessarily correct in every, or any, way.

People were misled by media long before the internet, of course, but they have been even more since it arrived. For two decades now, almost everything people see online has been potentially incorrect, untrustworthy, or otherwise decoupled from reality. Every internet user has had to run a hand-rolled, probabilistic analysis of everything they’ve seen online, testing its plausibility for risks of deception or flimflam. The slopverse simply expands that situation—and massively, down to every utterance.

Faced with the problems a slopverse poses, AI proponents would likely make the same argument they do about hallucinations: that eventually, the data, training processes, and architecture will improve, increasing accuracy and reducing multiversal schism. Maybe so.

But another worse and perhaps more likely possibility exists: that no matter how much the technology improves, it will do so only asymptotically, making the many multiverses every chat interaction spawns more and more difficult to distinguish from the real world. The worst nightmares in multiversal fiction arrive when an alternate reality is exactly the same save for one thing, which might not matter, or which might change everything entirely.

Why Hotel-Room Cancellations Disappeared

2025-11-13T13:31:33-05:00

Delayed two hours, hunched over my laptop in the Dallas Fort Worth C-terminal Admirals Club, I was frantically rearranging my plans. The government shutdown, still ongoing at the time, had caused major disruptions at U.S. airports. If my flight were canceled, the airline would refund me for my ticket. But my hotel room in Charlotte, North Carolina, appeared to be another matter. I clicked around the booking website on my screen. Its policy on cancellation was austere: You could void your reservation only if you did so three days in advance. If your plans happened to fall through unexpectedly the night before (because, let’s say, your nation’s legislature had failed to pass a budget), then you’d be out of luck.

This felt new. In the past, a hotel booking had been an easy thing to cancel. Up until the day before check-in, you could generally modify your plan without incident, and absent any fees. But this no longer seems to be the case. The age of travel flexibility is over. Hotel cancellation has been canceled.

The sad story of this change begins in about 2018; its villain is—surprise—the internet. Around that time, third-party travel-booking sites began to use a novel method of securing deals known to industry insiders as “cancel-rebook,” Christopher Anderson, a professor at Cornell University’s Nolan School of Hotel Administration, told me. It worked like this: The sites would let you book a room at the best available price, and then they would keep watching that hotel in the days and weeks that followed, to see if its posted rates would ever dip. A hotel might, for instance, drop its prices for last-minute bookings so that fewer rooms were left unfilled. If and when that happened, the travel websites’ cancel-rebooking scheme kicked in: Your reservation would be swapped for the cheaper one.

Cancel-rebook was great for consumers but terrible for hotels. The properties could no longer reduce their rates to manage unsold inventory without losing already-booked revenue to the online travel services. As a result, they started offering a bunch of different rates for the same room with varying degrees of flexibility. Travelers might find that they could book a room at a discounted, prepaid rate with no cancellation allowed, or at a mid-range rate with a two- or three-day cancellation deadline. In some cases, the old, until-the-night-before cancellation option would be on offer for a higher rate, too. This didn’t fully solve the problem of the cancel-rebook sites, because they could still swap reservations until just before the deadline. But it attenuated the worst effects.

[Read: The carry-on-baggage bubble is about to pop]

From the travelers’ perspective, the stakes of such restrictive policies are higher for hotels than they are for airlines. In most cases, if you cancel an airline flight on a major carrier, you can at the very least apply the value of your ticket to a future fare. But canceled-too-late and no-show hotel bookings are more likely a total loss. Hotels are not inclined to offer you a credit for your booking, even if they represent a sprawling chain with many thousands of properties. That’s because, unlike airlines, most hotels are not centrally owned. If you book the Ithaca Marriott, that would be owned by a franchisee, Anderson said. If a local owner has essentially licensed a hotel-chain brand for access to its customers, they may have no incentive to provide you with a credit that could be used some other time for a room at, say, the New York Marriott Marquis.

In other words, travelers and the online booking services have exploited—and unwittingly depleted—the shared resource of hotels’ flexibility in the hunt for the cheapest possible rooms. It’s a tragedy of the commons: Now all of us are left to hedge our travel plans against the hotels’ more restrictive policies, which themselves were hedges against the cancel-rebooking schemes.

What can travelers do to mitigate the situation? Hotel rates are less prone to drifting up and down than they used to be, Anderson told me, so there may be little cost in waiting to book your room until you’re sure you need it (and then choosing the best rate at that time). But even then, an unexpected delay or cancellation can still put you out of pocket. In that case, Anderson recommends a personal plea: Call the hotel, be nice, and explain your circumstances. “They want to make you happy,” he said. A Hilton spokesperson told me that exceptions to its properties’ cancellation policies are made on a “case-by-case basis, with broad waivers often extended” in the case of natural disasters or other events, and noted that cancellation charges may be waived for plans affected by the flight reductions. (Marriott and IHG, two other major hotel chains, did not respond to requests for comment.)

[Read: Why is airplane WiFi still so bad?]

This extended grace certainly sounds delightful! But the entire industry—perhaps the entire world—has been steered away from human interactions of this kind. Automated systems for e-commerce, such as cancel-rebook, have turned every commercial transaction—ordering a pizza, hiring a babysitter, hailing a car, whatever—into an opportunity to insert some technological middleman. Even if I did try to contact a hotel in the event of travel disruption, I’d expect to be funneled into a labyrinth of computerized customer-service menus or AI doomchats before anyone could even try to help me out of my predicament.

This is what it’s like to be a traveler today: You’re moving on a sea of internet-enabled processes, never really sure where the machines of arbitrage are pushing you, or why. If you don’t end up where you meant to go, then your options may be limited. You didn’t choose these terms for travel, but you now bear the risk they entail.

Why Students Are Obsessed With ‘Points Taken Off’

2025-11-04T09:50:39-05:00

This article was featured in the One Story to Read Today newsletter. Sign up for it here.

Harvard is worried about going soft. Specifically, about grade inflation, the name for giving ever higher marks to ever more students. According to an “Update on Grading and Workload” from the school’s office of undergraduate education, released last week to faculty and students, this trend has reached a catastrophic threshold. Twenty years ago, 25 percent of the grades given to Harvard undergrads were A’s. Now it’s more than 60 percent.

For all those students, though, the mere release of this document could be taken as its own catastrophe. “The whole entire day, I was crying,” one freshman told The Harvard Crimson. “It just felt soul-crushing.” One of her classmates warned that stricter standards would take a toll on students’ mental health—“I was looking forward to being fulfilled by my studies,” she said, “rather than being killed by them”—even as the report itself observed that deference to mental-health concerns has made the problem worse. A member of the men’s lacrosse team lamented that the findings failed to account for “how many hours we’re putting into our team, our bodies, and then also school.”

As a professor at another elite private university, who has been teaching undergraduates for more than 20 years, I have surely been guilty of inflating grades. I have also endured the confusing wrath of students who seem to think we professors are ruining their lives by awarding only 60 percent of each class with A’s. The spectacle unfolding at Harvard is more visible, but the condition that underlies it is widespread and chronic.

[Read: The perverse consequences of the easy A]

On the surface, grade inflation might seem simple to address: Just reestablish, in clear terms, that the baseline mark for showing up is not an A, but something lower; then give special credit only to the students who demonstrate their mastery and achievement. But it’s not so easy. Grade inflation has become a strange and wicked problem on campus—and it’s one without a single cause or an obvious solution.

If the culture of grading has eroded, it has done so over years and decades. Not all of the reasons are bad. Lower standards help first-generation college students and others who might arrive on campus with less traditional academic preparation. They also accommodate more modern forms of teaching, such as the “creative assignments and group projects” mentioned in the Harvard report. In total, the change has been so slow and steady that even faculty can barely feel it. We’ve simply been adjusting the expectations of our students, year after year.

Back in the mid-aughts, I was teaching at Georgia Institute of Technology, a highly selective technical university. One of my classes involved many different types of student work, including software projects, essays, and formal exams. The exams were hard, but everything they tested had been covered in my lectures, and the answers were definitive; grading was straightforward. Assessing the projects and essays was more subjective, as I was looking for creativity and insight. The students found this difficult, because they were unsure of what I wanted.

The thing is, figuring out what I might have wanted was supposed to be part of the assignment! I was asking my students to interpret my instructions in unexpected ways that exercised their own interests, abilities, and perspectives. For example, if I’d asked for a nontraditional computer paint program, and a student made one that let a user toss virtual pebbles into an on-screen pond to simulate water ripples that swelled and vanished, that would have surprised and delighted me. If the student really pulled it off, they’d get an A.

But students were complaining, so I tried to be responsive: I started giving unambiguous requirements. I told the students that simply meeting those requirements on a written or creative assignment, and doing nothing more, would earn them a C. To get a B or even an A, they would have to go further—not just by doing more, but by demonstrating a synthetic grasp of the material, carrying out their creative vision, completing the work with special polish, and so on. As ever, students were invited to my office hours to discuss the details.

[Read: College students have already changed forever]

The students hated this. They raised complaints with me or my teaching assistants: Why was mine a B and hers an A? What more could I have done? Appeals to “effort” were also common—as if exertion were a stand-in for achievement. What I was asking for was well outside the established norm. Students complained. A few even posted anonymous threats against my family in an online forum. (Such threats are not entirely uncommon in academia.) In other classes, a grade of A might have been earned by having met requirements. To get anything less would represent having “points taken off”—a concept that for students had by then become a gross obsession. Sometime since then, it became an ideology.

Over the past 25 years, while grades were going up, college was also getting more expensive and harder to get into. In 2001, Harvard accepted 10.7 percent of its applicants—an all-time low at the time. Last year it took in 3.6 percent. As a result, today’s average student may be of higher quality, and more deserving of an A, than ever before. But even if so, that’s not the whole story. Over the same period, college administrators institutionalized a concept called “student success.” Originally intended to reduce churn and increase graduation rates, student success expanded into something much broader—a blend of traditional academic achievement, personal satisfaction, and even wellness. These and other factors helped transform students from scholars into customers.

College in America has always been confused, a combination of a coming-of-age facility and a credentialing service. But the customer-centric, professionalizing function of undergraduate life muddled matters even further. College wasn’t just for discovering who you are or even meeting a future spouse, but for getting you into a career. The Harvard report notes this phenomenon: Many undergraduates see clubs, internships, and other extracurricular activities as necessary for getting jobs.

Meanwhile, the job of being a professor became more tenuous and provisional—some 75 percent of faculty are nontenured, many working term to term, with the fate of their employment determined, in part, by student course evaluations. Those surveys are, in turn, notoriously unsound as a measure of learning, but they do exert pressure to make students happy. And you know what makes a student happy? Giving them an A.

During the same period, due to changes in the university-accreditation process, a milkshake of new bureaucratic demands on classroom management was also served to the faculty. Failing to meet these requirements could put a school’s federal funding, including Pell Grants, at risk. Administrators started urging professors and departments to connect classroom work directly to “measurable learning outcomes” through “evaluative rubrics,” as the lingo of the process calls them. These are the elements that would satisfy the accreditors, and thus help the school maintain its student-aid support and ability to award degrees.

[Read: The most disrespected document in higher education]

But isn’t that what the grades are for? professors asked. Not anymore. We got the sense that as far as the accreditors were concerned, grades could not be trusted, because they sometimes varied by instructor, lacked diagnostic detail, or failed in other ways to provide sufficiently granular or reliable evidence of specific learning outcomes. In other words, the faculty was told—and has been reminded ever since—that grades do not prove mastery or achievement.

Amid and around this Kafkaesque affair, costs kept rising, students became even more like customers, and faculty came to accept that state of affairs. Constant pressure to perform and compete produced students so wound up with anxiety, they often came to office hours not for academic help but for therapy, despite our profound lack of qualifications for that role.

In the meantime, worries over students’ mental health, and a reasonable desire to accommodate disabilities or disadvantages, made us ever more inclined to yield to the rising tide of grade-inflationary demands. Administrators, attuned to student gripes and terrified of the accreditors, were pressing us to avoid any ambiguity in what we asked of students. I can say this from experience: Even the faculty who resisted these changes would endure year after year of pressure to conform.

And let’s not forget the computers. They’ve made it easier for undergrads to cheat on their assignments. But networked software services have also changed how classrooms work, and how students and professors relate to grades. By virtue of those changes, digitizing college life has led to grade inflation, too.

In the 1990s, when I was in college, your final grade in any class would be something of a mystery until you got it in the mail, or saw it taped to the professor’s office door. Until then, you’d do assignments and take exams. You’d get your scores, and if you were obsessive or concerned, you could calculate how you were doing by referring back to the syllabus: If I get an A on the second paper and the final, I can still eke out an A for the semester.

These days, thanks to the “courseware” that has become ubiquitous in higher ed, students can see exactly how they are performing in every course all the time. The software can even project their final grade based on how they’ve done so far, in a data-dashboard sort of way. Students love this, or think they do, because they don’t want to be surprised. But the courseware data dashboards have another clear effect: Like so many other aspects of the current college experience, they orient students’ attention toward their grades above all else.

At some colleges and universities, courseware has been mandated. Schools have done this, in part, because students are accustomed to using the software and prefer to see all their work in one place. But the mandates also help colleges shovel heaps of bureaucratic muck—validating data for accreditation, carrying out enrollment, flagging troubled students, aggregating metrics of all kinds. Whether this IT-ification of university life makes teaching and learning any better is not important for the topic at hand. The point is, all of it together has reinforced the focus on graded performance, offering students and faculty more opportunities for anxiety and conflict.

I have tried to find ways to return to the old ethos of grading, in which I would judge a student as a whole person rather than as a series of assignment transactions. But in the age of courseware, I must give this holism a name and a value and a slot in the gradebook. Fine. I call it “Slush”—a grading category that I put into the system to account for whatever the rubric, the outcomes, and their computerized rigidity cannot. Slush is my gesture at an overall assessment of student performance and growth.

But, alas, my Slush is making students anxious, too. “What’s Slush?” they sometimes ask, halfway through the course, because they didn’t read my explanation in the syllabus. Some complain, “Yours is the only class where I don’t know my grade.” Assessing overall performance and growth, it seems, might not be worth my trouble.

More students get A’s, yet students are unhappy with their grades. Professors, too, have been worn out by the grading nightmare. We now plan with dread for all the ways our students might misconstrue our feedback or petition for the “additional points” that they are sure we have stolen from them. Grading was never fun, but now it is odious. The easiest answer is just to give the students what they expect, at least some of the time, so you can get on with the rest of your job—which has been made immeasurably harder in lots of other ways.

The knots tighten and multiply. The courseware grading system enforces the use of an unambiguous grading rubric, which furthers the bureaucratization of classroom life that helped to amplify grade inflation in the first place. Just as the students seek out easier classes for a certain A, the professors pursue simpler course designs that de-escalate the fighting over grades.

Everybody understands that, in the current state of things, grades say little about what students know or learn. But the machinery of grading churns on. The same students who scrabbled for achievement to gain entry into colleges like mine, where they clamber for the A’s they believe they deserve, know that grades still matter—for medical- or professional-school admission, or to compete with their peers for limited slots at management-consulting internships, or even just to appease their parents, who may be just as prone as they are to mistaking assessment for achievement.

To demand a fix for grade inflation is to put multiple-choice answers on an essay question. It asks for something that cannot be marked as right or wrong. This, as it happens, is the type of lesson most professors yearn to impart in our classes: that process, not its product, is the goal. Many students learn that lesson much later, after they graduate, when they look back and wonder at their former obsession with grades. By then, however, they are no longer customers of higher ed. As for us professors, we never leave. Each year the grades rise a little more; each year we feel it less. And the bureaucratic strangle that leads to this inflation continues its creep.

You’re Getting ‘Screen Time’ Wrong

2025-10-23T07:00:00-04:00

This article was featured in the One Story to Read Today newsletter. Sign up for it here.

“That’s enough screen time for today,” you tell your kid, urging them to turn off the video-game console or iPad. As for what they should do instead, you are not quite sure. And what about you? If only you could put down your phone and listen to your spouse, or read a book, or embrace the sensation of your own existence, then surely you would be a happier, better person.

But this is wrong. Screen time is not a metric to optimize downward, but a name for the frenzy of existence in an age defined by screens. You may try to limit the time that you or your children spend with screens, and this may bring you minor triumphs. But you cannot rein in screen time itself, for screen time is the speed of life today. To recognize that fact—and to understand how it happened—is a small, important step toward salvation.

Long before screen time was a brand name for self-loathing—long before it had given rise to smartphone apps that were supposed to cleanse your soul of backlit sin—the notion had to be invented. This happened in the summer of 1991, when Mother Jones published an issue called “We Hate Kids.” Its cover featured Bart and Lisa Simpson, characters then but two years old; tucked away inside was an essay by the writer Tom Engelhardt called “Primal Screen.” “The screen offers only itself as an organizing principle for children’s experience,” it said. Television shows didn’t just tell stories; they showed characters such as Garfield watching television themselves, sometimes obsessively. MTV, then scarcely more than a decade old, famously put literal televisions on-screen and on set. Kids were watching “screens within screens within screens,” Engelhardt wrote, and they were doing it a lot: Even six-month-old babies were getting “an average hour and a half of screen time a day; the typical older child, about four hours.”

Televisions had already been around for decades, and people had lamented their existence from the start. The nickname “boob tube” first appeared in the 1950s. Screen panics of various kinds arose and subsided every decade thereafter. In 1984, the American Academy of Pediatrics warned parents that television might have ill effects on childhood development. Reaffirming the idea in 1990, the AAP reported that American children were spending more time watching television than any other activity, apart from sleeping. Time spent with TV was concerning, it said, because passive viewing of the screen “may displace more active experience of the world.”

But any measure of time spent in front of screens did not—and still does not—explain the changing nature of this experience. For Engelhardt, the problem screen time named was not merely one of duration but one of pace. The endless, frenzied display of screened images proceeds “as if chased by some implacable force,” he wrote, and “it is that pace that drives the child.” Time in the age of screens was rapidly changing, brazenly commercializing, exploding into bits and then getting distributed across multiple venues and devices. By the early ’90s, many American households had multiple televisions, along with VCRs and video-game consoles, and kids carried Game Boys when they went outside. Engelhardt observed that different sorts of screens, including cinema, television, and games, were getting linked together in what would later be described as “franchise media.” Merchandising was a side effect of this infectious spread, an invitation and a demand to consume the trappings of the images that screens emitted.

[Read: I tried to limit my screen time]

In the 34 years since Engelhardt’s essay, the outbreak of screens has increased in scale many times over. So has the pace of activity on those screens as the inescapability of screen life became entrenched. Engelhardt probably used a word processor to write his essay in 1991, but the experience would have been quiet and solitary. Today, I write this one on a different sort of screen: a windowed computer operating system. Even as I work, I will receive reminders about my appointments, various requests from friends and family, constant emails, and dozens of notifications from the other apps and services I use. Today’s screens within screens within screens are shredding my attention into bits.

Television foreshadowed this situation. Over the past few decades, all TV has become more like MTV. Commercial-free streamers offer no natural breaks for a viewer’s attention. Even standard news programming is as noisy and disjointed as the output of a Bloomberg terminal. But just using a television now demands a faster pace. Channels have been replaced by streaming services, each of which has its own menu screen with an individual visual language and interaction paradigm. Selecting a show may result in the involuntary viewing of multiple, in-line trailers. And then, once you’ve settled on a program, you probably also tap at and scroll on the smartphone in your hand while watching, whether to respond to work messages or shop for home goods or argue on social media or crush candy. If you go to the gym, you might watch a trainer or a YouTube video on the screen atop your stationary bike. You might even operate your motor vehicle (at your peril) by touching screens. The car I bought recently asks drivers to sign in to its screen with a profile, as if the vehicle were just another Netflix. Its hybrid engine adds yet one more screen to my instrument panel for monitoring electrification. When I go to fill up the tank, the gas pump has a screen, too, hawking services that might also take place on screens.

In this context of screen omnipresence, to measure the amount of time you spend looking at a screen is simply to ask how much time you spend awake and cogent, for almost everything you do now requires a screen to do it. (If you wear a watch or a ring at night for sleep tracking, your slumber will be reclaimed by a screen, too.) The fallacy of screen time holds that measuring a ubiquitous phenomenon provides information that allows for control of that phenomenon—that keeping records of a chronic state will give rise to certain habits of self-healing.

This won’t really help you to adopt new habits, for two reasons. First, because general-purpose computing made screens so widespread that many worthwhile activities take place on them. You are almost certainly reading this article on a screen! My Washington University in St. Louis colleague Phillip Maciak, who wrote a book called Avidly Reads Screen Time, points to the obviously beneficial practice of using screens to text or video-call friends and family. The raw quantity of time one spends in the thrall of screens says almost nothing about the value of the time spent.

[Read: What did people do before smartphones?]

Measuring screen time for self-improvement also fails because few, if any, activities even exist outside of screen time. Nature has been conquered by cellular coverage. Bars and restaurants were already riddled with screens before everyone at the table also clutched one in their hand. Psychologists and educators are calling for bans on smartphones at schools, but many schools have already replaced chalkboards with computer-controlled smartboards, or distributed Chromebooks or iPads to every student. Even bowling alleys—the fantasy site for mid-century, prosocial communion, thanks to the political scientist Robert Putnam’s Bowling Alone—require programming and then operating screens to keep score.

What to do, then, about screen time? The fact is, you cannot participate fully in contemporary life without devoting a substantial amount of time to the screen. Even if you try to pare back your screen time to some bare minimum for engagement with the world today, whatever quantity remains will still be chaotic and attention-shattering by nature. You might rationally take steps to protect your children from that situation while their identity (and brain) is still forming, but those efforts will only delay the inevitable. Every kid will be thrust into the frenzy of screen life at some point during their adolescence, or else they will fail to enter contemporary adulthood.

Screen time is a systemic issue, so an individual response—your screen-time monitoring, your screen-time mitigation—will likely be of little use. Past experience suggests that this problem will resolve itself at historical scale instead. After all, in the early days of literacy, reading—now perhaps the paradigmatic example of a non-screen-time activity—was considered ominous; people reading silently to themselves might have seemed demented. Even in the 19th century, the novel was considered a dangerous medium, one that would trap people—especially impressionable young women—in the thrall of isolation and fantasy. (Today, a couple of centuries later, people instead complain that young adults no longer have the attention span for isolationist fantasy.)

It’s hard to fathom now, but Marshall McLuhan, who became famous in the 1960s for the idea that media forms shape perception, saw the screen as an antidote to the poison of the book. McLuhan appreciated television for its lo-fi ambience that activated many senses all at once. As such, he thought that screens would help usher in a new age, a “global village” in which multisensory media would connect people in small scale, ad hoc ways, replacing the top-down, authoritative forms of media that preceded them, such as books.

The world ended up getting the global village McLuhan had predicted, albeit not in exactly the way he had predicted it. In particular, screens mated to computers, the most flexible machines ever invented. Together, the two amended all previous media forms. The computer-with-a-screen subsumed those media, and it did so at the pace of screen time, that is, with increasing speed and swelling fragmentation.

Will screen time ever slow? Can it ever be controlled? Tom Engelhardt thought such an end was inevitable—if for no other reason than sheer exhaustion. His object lesson was Pee-wee Herman, a “bizarrely hyperactive” screen-time-accelerated counterpoint to Mr. Rogers. Surely, Engelhardt suggested, the limits of the human body and brain could not sustain such extreme energy, not for long. His conclusion was wrong back then, half a lifetime ago: The hyperactive energy of the television age has persisted—and then spread into every corner of contemporary life. Perhaps someday the age of screens will end, at the hand of some unthinkable novelty or civilization-ending calamity. But until that happens, tracking use of screens—let alone trying to curtail it—will have little meaning. For now, at least, you are doomed to live at screen time.

Job Interviews Are Broken

2025-10-15T15:59:44-04:00

“Interviews are NOT real anymore.” So reads the opening caption of a TikTok posted in September, punctuated by the skull-and-crossbones emoji. In the video, a young woman interviews for a job on a video call. She has a smartphone propped up against her laptop screen, so she can read off the responses that an AI app has composed for her: “Um, yeah, so, one of my key strengths is my adaptability.” She’s got a point. Getting generative artificial intelligence to whisper into your ear during a job interview certainly counts as adaptable.

More clips from the same alleged job interview give the app a further showcase. “I prioritize clear communication and actively listen,” a woman says in one, as she reads from a phone instead of actively listening. Another such post, which has racked up 5.3 million views, is subtitled “My interviewer thought he caught me using Ai in our LIVE interview.” It shows the same potential boss from all the other videos asking her to share her screen and click through her browser tabs. After doing this, she resumes reading off her phone. “Little did he know,” the subtitle says.

AI-job-search anxiety has been growing for some time. In the past few years, employers started using AI to “read” and screen the thousands of résumés they may receive for each job posting; job searchers began to deluge HR departments (or at least their automated filters) with AI-generated applications; and companies began employing AI agents—fake people—to conduct their first-round interviews. Imagine eating a hearty breakfast, donning your best blazer, and discovering that you’ll be judged by a robo-recruiter.

By this spring, the arms race had advanced to the point where, apparently, applicants were using AI assistants to supply them with material for computer-programming interviews on Zoom. In August, The Wall Street Journal reported that AI is “forcing the return of the in-person job interview,” and that big firms such as Cisco and McKinsey have been urging hiring managers to meet with candidates in person at least once on account of the technology.

The letter of these reports suggest a simple story of innocence and malfeasance. Some HR companies have even described the phenomenon as “interview fraud,” attributing something akin to criminal intent to the job seekers who might pursue it. But the more I investigated and considered the circumstances, the less that label seemed appropriate. Something weirder is taking place. In the context of a tightening economy, employers have turned a powerful technology against their prospective employees. Who could blame the job seekers for retaliating?

Is a lot of “interview fraud” even really happening? TikTok seems to show a rising trend; posts depict job candidates—especially young ones afflicted by a difficult, AI-degraded job market—who have started using AI to game the interviews themselves. But on closer look, many of these videos are not documenting a scandal so much as wishing one into existence—and monetizing the result. For instance, the ones described above, with the woman who had her phone propped up against her laptop, were posted by an account called @applicationintel, which displays a bio that urges viewers to download an AI app called “AiApply.”

I found many others of this kind. An AI-interview-software company called LockedIn AI posts on TikTok about how to “Crush Any Job Interview” with its tools. Kazuyoshi Fujimoto, a young professional engineer with a side hustle as a “career expert,” has a series of posts with subtitles like “My brother is interviewing for a $469k engineer job using AI.” Fujimoto answered my request to talk for this story, but stopped responding when I followed up to ask whether any of his posts were staged.

The fact that AI-interviewing services are being pushed by TikTok influencers suggests that there is money to be made from this idea, and that these products’ customers are real. I wanted to see whether those customers were buying something useful. In one of his TikTok posts, Fujimoto recommends a tool he likes called Final Round AI, which “listens in real time” and “suggests killer responses.” I decided to sign up to see how it worked. (A basic subscription is free; one that allows unlimited live interviews and hides the app during screen sharing costs $96 a month.)

After opening the Final Round “Interview Copilot,” you have to tell it about the role for which you will be interviewing. By default, there are a few dozen options—and almost all of them are in software development or its orbit. I settled on “content writer” (ugh) as the closest match to what I’m doing here and started on a practice interview. I asked Final Round AI to supply me with an answer to this potential question: “If I assigned you a story on people using AI to cheat on job interviews, how would you approach that topic?”

It returned a lengthy, milquetoast answer that began, “First, I’d want to really understand the scope of the issue. How widespread is this? Are we talking about a few isolated incidents, or a growing trend? Also, I’d immediately flag the ethical considerations. This isn’t just about tech; it’s about fairness, integrity, and the future of work.” The entire thing was plausible in the way LLM responses often are; if an aspiring writer provided this response during a genuine interview, it wouldn’t be wrong so much as uninspired. It is the sound of a person performing the role of a job candidate, rather than one actually pursuing a job. (Final Round AI did not respond to my request to discuss its software for this story.)

Reading the app’s suggested interview response, and imagining myself actually delivering it with a straight face on a Zoom, brought to mind the opening scene from the 1990 film Joe Versus the Volcano, in which the title character arrives at work while his boss, Mr. Waturi, takes a phone call in the background. “I know he can get the job,” Mr. Waturi says into the handset. “But can he do the job?” Mr. Waturi repeats that sentence, varying his emphasis, over and over.

On its surface, Mr. Waturi’s question is a good one: A person can carry out the rituals of employability—assembling a good résumé, performing effectively at an interview, even carrying out a satisfactory test-case work assignment—and still be unable to produce useful results in the workplace. Today’s AI-interviewing tools would seem to make this problem worse: Now almost anyone can get the job, with automated help. Whether they can really do it is irrelevant. Just as students can now fake their way through school and college, what’s to stop them from cutting corners on their way into Meta or McKinsey?

[Read: AI cheating is getting worse]

But the film also makes clear that Mr. Waturi’s concern with job performance is vacuous. Joe’s dreary, squalid workplace, called American Panascope, is described as “Home of the Rectal Probe.” Given this backdrop of hostility toward the firm’s workers and its customers alike, Mr. Waturi’s incantation, I know he can get the job, but can he do the job?, comes across as bureaucratic nonsense, an exercise in the mere appearance of high standards. Joe, the defeated salaryman, takes all this in as he hangs his coat and hat: What would it even mean to do the job when the job is so meaningless?

This question reemerges in a twisted form today, when the same companies that worry over being duped by AI-assisted applicants would love to have a workforce that makes use of AI in lots of other ways. The people who use Final Round AI to get their software-engineering jobs might be superbly qualified, in fact, to do those jobs in just the way their bosses would prefer. And if consulting is an industry that steals your watch to tell you the time (as the classic line goes), then a junior consultant who used AI to fake his way into the role might well be on the road to make partner.

For some time now, workers—and especially young ones—have become ever more detached from their work lives. David Graeber called the roles they end up taking for lack of any better option “bullshit jobs.” Internet culture has more recently nicknamed them “email jobs”: work whose purpose is so cryptic, its effort detaches from motivations and outcomes, personal or professional. The Millennials who graduated into the Great Recession talked about LARPing their own jobs in order to reconcile this divide. Cheating on a job interview with AI feels like a realization of that vision: You are no longer a job candidate, but a person playing the role of one.

But wait, isn’t a junior-associate position at McKinsey or “a $469k engineer job” distinct from the sort of dead-end, bullshit job that produces so much workplace alienation? Yes and no. If you can land a role like that, certainly it may pay you well, and confer a degree of social status. But the pursuit of nearly every form of office job, even those that demand a particular credential and specific experience, has become a hellish ordeal. Candidates submit forms and résumés into LinkedIn or Workday, where they may be chewed up by AI processors and then consumed without response, or else advanced to interviews (which may also be conducted by AI). No matter who you are, the process of being considered for a job may be so terrible by now that any hidden edge in getting through it would be welcome.

Rewatching the AI-interview TikToks with new empathy for the young professionals who seek employment in today’s chaotic marketplace, I noticed a pattern I had previously overlooked: a realpolitik of resigned desperation. Some presented themselves as using AI to generate bespoke résumés in response to specific job postings, an act that now seems necessary to get around the AI filters that may perform first-round culling. One young woman offered tips for using AI to prepare for job interviews: Instead of buying an app that listens in and tells you what to say, she suggested using the technology to generate sample questions that you might be asked, so you can practice answering them. She titled the video, “How to use AI to pass ANY interview.”

This language struck me as both incisive and honest. Passing is a contemporary life philosophy, one adopted by habit rather than duplicity. Ironic detachment has moved well beyond LARPing a career. Now one simply attempts, against the odds, to luck into a career, or at least the appearance of one. Today, students might use AI to write college-entrance essays so that they can get into college, where they use AI to complete assignments on their way to degrees, so they can use AI to cash out those degrees in jobs, so they can use AI to carry out the duties of those jobs. The best one can do—the best one can hope for—is to get to the successive stage of the process by whatever means necessary and, once there, to figure out a way to progress to the next one. Fake it ’til you make it has given way to Fake it ’til you fake it.

Nobody has time to question, nor the power to change, this situation. You need to pay rent, and buy slop bowls, and stumble forward into the murk of tomorrow. So you read what the computer tells you to say when asked why you are passionate about enterprise B2B SaaS sales or social-media marketing. This is not an earnest question, but a gate erected between one thing and the next. Using whatever mechanisms you can to get ahead is not ignoble; it’s compulsory. If you can’t even get the job, how can you pretend to do it?

Online Shopping May Never Be the Same

2025-08-29T13:16:00-04:00

A few years ago, I found the perfect rug for my daughter’s room. It had pink unicorns and flowers. But I scoffed at the price tag on Anthropologie’s website: more than $1,000, plus an additional fee for “white glove delivery.” Then I fired up Etsy. I found a similar product made by a workshop in India that shipped directly from there. It took weeks to arrive, but it was half the price.

Online shopping is a miracle: You can find items of any kind, fit for any purpose, for affordable prices—and shipped from all over the world to your door. But as of today, buying from international sellers has become more expensive for Americans. That’s because President Donald Trump ended the de minimis exemption on imported goods, a loophole that allowed millions of daily packages to enter the country without paying duties. The exemption has been around for a long time—nearly a century—but it took on new import (get it?) in 2016, when the maximum value for untaxed goods rose from $200 to $800. In that moment, the social-media-driven rise of direct-to-consumer e-commerce, drop -shipping, and online-marketplace sales were also accelerating. Ever since, American ports, mailboxes, and homes have been flooded with cheap clothing, electronics, accessories, skin-care products, toys, and a host of other consumer goods.

The de minimis loophole is a big reason e-commerce sites including Shein and Temu could sell you things for so cheap: They shipped straight from China, skirting any tariffs. The White House ended the exemption for goods from China earlier this year, and now de minimis is ending for all countries. That means that many things you might import could become more expensive (on account of the additional taxes) or harder to buy (because sellers won’t bother shipping to the U.S.), or take longer to arrive (because of customs backlogs), or any combination of those. The rug I bought a few years ago would now be subject to a 50 percent import duty, when you factor in tariffs on India. Presuming that cost is passed down to consumers, it’s enough to give a buyer like me pause.

You might not realize how much of the stuff you buy online comes directly from overseas. I didn’t, until I looked closely at my buying habits over the past few years. After all, sites such as Etsy and eBay offer seamless global commerce: A handmade craft object could come from Maine or Myanmar, straight to you. Even Amazon has benefited from de minimis. Various strategies have allowed the retailer’s marketplace suppliers to take advantage of de minimis when they import goods; at other times, when you buy from the big platforms’ sites, those vendors might ship what you ordered directly from abroad, tax free.

[Read: Amazon decides speed isn’t everything]

Buying cheap imported goods has become the best part of online shopping: Not only can you find the best deals from international sellers, but also you can source items to satiate specific hobbies and interests—say, drafting pens from Japan or instrument reeds from Belgium. I found that I had bought a host of stuff, on Etsy and beyond, that took advantage of de minimis, including rubber-tree hippo figurines from Denmark (naturally) and a surprise mandolin from Ireland for my daughter. Those goods would now be subject to an additional tariff. I’ve bought incredibly cheap Chinese- and Japanese-manufactured camera lenses that have fueled a resurgence of photography hobbyism for me and my son; I also bought a detailed and shockingly high-quality Paul Revere costume to help a neighbor’s kid beat her classmates in a school costume contest—a small thing, but one we’ll all remember.

Ah, and then the British faucet doodad. This was a big deal. When I tried to repurpose an old, turn-of-the century washbasin with separate hot and cold water spigots, I couldn’t find a faucet that fit the sink. Sure enough, some vendor in the United Kingdom had a $30 plastic tube that did the trick. International sellers sometimes are the only ones that have what you need, and you don’t need to be a particularly adept shopper to find them. A simple Google search will suffice.

Of course, being able to seamlessly import cheap stuff has also encouraged mindless consumerism. Some imported goods are crap that nobody ever needed, produced at unconscionable labor and environmental costs. My family has a bit of a LEGO habit, and my son took to buying the cheaper Chinese knockoff sets to maximize our, well, brick-building value, I suppose. It felt a little suspect to do this—the sets are direct copies of LEGO designs—and many of them remain in bags in a closet, unbuilt. Surely we didn’t need to import those. Nor the piles of cables, chargers, head lamps, and other low-cost electronic goods that broke after a few uses.

Whether it’s junk or not, Americans have become acclimated to buying a prodigious variety of wares from all over the world. When de minimis fused with global online commerce a decade ago, ordinary buyers like you and me started to see behind the curtain of domestic retailers. Anthropologie’s website touted that the unicorn rug was “exclusive” to its store. But that was never entirely true: Sellers offering the same style with similar materials found a way to reach buyers like me directly, thanks to online commerce and its associated marketplaces. That’s not going to change anytime soon. Instead, buying things will just become more painful. Someone will bear the burden of the new duties, and that someone is likely to be you.

College Students Have Already Changed Forever

2025-08-17T07:00:00-04:00

A college senior returning to classes this fall has spent nearly their entire undergraduate career under the shadow—or in the embrace—of generative AI. ChatGPT first launched in November 2022, when that student was a freshman. As a department chair at Washington University in St. Louis, I witnessed the chaos it unleashed on campus. Students weren’t sure what AI could do, or which uses were appropriate. Faculty were blindsided by how effectively ChatGPT could write papers and do homework. College, it seemed to those of us who teach it, was about to be transformed.

But nobody thought it would happen this quickly. Three years later, the AI transformation is just about complete. By the spring of 2024, almost two-thirds of Harvard undergrads were drawing on the tool at least once a week. In a British survey of full-time undergraduates from December, 92 percent reported using AI in some fashion. Forty percent agreed that “content created by generative AI would get a good grade in my subject,” and nearly one in five admitted that they’ve tested that idea directly, by using AI to complete their assignments. Such numbers will only rise in the year ahead.

“I cannot think that in this day and age that there is a student who is not using it,” Vasilis Theoharakis, a strategic-marketing professor at the Cranfield School of Management who has done research on AI in the classroom, told me. That’s what I’m seeing in the classes that I teach and hearing from the students at my school: The technology is no longer just a curiosity or a way to cheat; it is a habit, as ubiquitous on campus as eating processed foods or scrolling social media. In the coming fall semester, this new reality will be undeniable. Higher education has been changed forever in the span of a single undergraduate career.

“It can pretty much do everything,” says Harrison Lieber, a WashU senior majoring in economics and computer science (who took a class I taught on AI last term). As a college student, he told me, he has mostly inhabited a world with ChatGPT. For those in his position, the many moral questions that AI provokes—for example, whether it is exploitative, or anti-intellectual, or ecologically unsound—take a back seat to the simple truth of its utility. Lieber characterized the matter as pragmatic above all else: Students don’t want to cheat; they certainly don’t want to erode the value of an education that may be costing them or their family a small fortune. But if you have seven assignments due in five days, and AI could speed up the work by tenfold for the cost of a large pizza, what are you meant to do?

In spring 2023, I spoke with a WashU student whose paper had been flagged by one of the generally unreliable AI detectors that universities have used to stem the tide of cheating. He told me that he’d run his text through grammar-checking software and asked ChatGPT to improve some sentences, and that he’d done this to make time for other activities that he preferred. “Sometimes I want to play basketball,” he said. “Sometimes I want to work out.”

[Read: The first year of AI college ends in ruin]

His attitude might have been common among large-language-model users during that first, explosive year of AI college: If a computer helps me with my paper, then I’ll have more time for other stuff. That appeal persists in 2025, but as these tools have taken over in the dorms, the motivations of their users have diversified. For Lieber, AI’s allure seems more about the promise of achievement than efficiency. As with most students who are accepted to and graduate from an elite university, he and his classmates have been striving their whole life. As Lieber put it, if a course won’t have “a tangible impact on my ability to get a good job,” then “it’s not worth putting a lot of my time into.” This approach to education, coupled with a “dismal” outlook for postgraduate employment, justifies an ever more ferocious focus on accomplishment. Lieber is pursuing a minor in film and media studies. He has also started a profitable business while in school. Still, he had to network hard to land a good job after graduation. (He is working in risk management.)

Da’Juantay Wynter, another rising senior at WashU who has never seen a full semester without AI, told me he always writes his own essays but feels okay about using ChatGPT to summarize readings, especially if he is in a rush. And like the other students I spoke with, he’s often in a rush. Wynter is a double major in educational studies and American-culture studies; he has also served as president of the Association of Black Students, and been a member of a student union and various other campus committees. Those roles sometimes feel more urgent than his classwork, he explained. If he does not attend to them, events won’t take place. “I really want to polish up all my skills and intellect during college,” he said. Even as he knows that AI can’t do the work as well, or in a way that will help him learn, “it’s always in the back of my mind: Well, AI can get this done in five seconds.”

Another member of his class, Omar Abdelmoity, serves on the university’s Academic Integrity Board, the body that adjudicates cases of cheating, with AI or otherwise. In almost every case of AI cheating he’s seen, Abdelmoity told me, students really did have the time to write the paper in question—they just got stressed or preoccupied by other things, and turned to AI because it works and it is available. Students also feel the strain of soaring expectations. For those who want to go to medical school, as Abdelmoity does, even getting a 4.0 GPA and solid MCAT scores can seem insufficient for admission to the best programs. Whether or not this is realistic, students have internalized the message that they should be racking up more achievements and experience: putting in clinical hours, publishing research papers, and leading clubs, for example. In response, they seek ways to “time shift,” Abdelmoity said, so they can fit more in. And that’s at an elite private university, he continued, where the pressure is high but so is the privilege. At a state school, a student might be more likely to work multiple jobs and take care of their family. Those ordinary demands may encourage AI use even more.

In the end, Abdelmoity said, academic-integrity boards such as the one he sits on can only do so much. For students who have access to AI, an education is what you make of it.

If the AI takeover of higher ed is nearly complete, plenty of professors are oblivious. It isn’t that they fail to understand the nature of the threat to classroom practice. But my recent interviews with colleagues have led me to believe that, on the whole, faculty simply fail to grasp the immediacy of the problem. Many seem unaware of how utterly normal AI has become for students. For them, the coming year could provide a painful revelation.

Some professors I spoke with have been taking modest steps in self-defense: They’re abandoning online and take-home assignments, hoping to retain the purity of their coursework. Kerri Tobin, an associate professor of education at Louisiana State University, told me that she is making undergrads do a lot more handwritten, in-class writing—a sentiment I heard many times this summer. The in-class exam, and its associated blue book, is also on the rise. And Abdelmoity reported that the grading in his natural-science courses has already been rejiggered, deemphasizing homework and making tests count for more. These adjustments might be helpful, but they also risk alienating students. Being forced to write out essays in longhand could make college feel even more old-fashioned than it did before, and less connected to contemporary life.

Other professors believe that moral appeals may still have teeth. Annabel Rothschild, an assistant professor of computer science at Bard College, said she’s found that blanket rules and prohibitions have been less effective than a personal address and appeal to social responsibility. Rothschild is particularly concerned about the “environmental harms” of AI, and she reports that students have responded to discussions that vein. (AI data centers consume a large amount of water and electricity.) The fact that she’s a scientist who understands the technology gives her message greater credibility. It also helps that she teaches at a small college with a focus on the arts.

Today’s seniors entered college at the tail end of the coronavirus pandemic, a crisis that once seemed likely to produce its own transformation of higher ed. The sudden switch to Zoom classes in 2020 revealed, over time, just how outmoded the standard lecture had become; it also showed that, if forced by circumstance, colleges could turn on a dime. But COVID led to little lasting change in the college classroom. Some of the students I spoke with said the response to AI has been meager too. They wondered why faculty weren’t doing more to adjust teaching practices to match the fundamental changes wrought by new technologies—and potentially improve the learning experience in the process.

Lieber said that he wants to learn to make arguments and communicate complex ideas, as he does in his film minor. But he also wonders why more courses can’t assess those skills through classroom discussion (which is hard to fake) instead of written essays or research papers (which may be completed with AI). “People go to a discussion-based class, and 80 percent of the class doesn’t participate in discussion,” he said.

The truth is that many professors would like to make this change but simply can’t. A lot of us might want to judge students on the merits of their participation in class, but we’ve been discouraged from doing so out of fear that such evaluations will be deemed arbitrary and inequitable—and that students and their parents might complain. When professors take class participation into account, they do so carefully: Students tend to be graded on whether they show up or on the number of times they speak in class, rather than the quality of what they say. Erin McGlothlin, the vice dean of undergraduate affairs in WashU’s College of Arts & Sciences, told me this stems from the belief that grading rubrics should be crystal clear in spelling out how class discussion is evaluated.

For professors, this approach avoids the risk of any conflicts related to accommodating students’ mental health or politics, or to bureaucratic matters. But it also makes the modern classroom more vulnerable to the incursion of AI. If what a student says in person can’t be assessed rigorously, then what they type on their computer—perhaps with automated help—will matter all the more.

Like the other members of his class, Lieber did experience a bit of college life before ChatGPT appeared. Even then, he said, at the very start of his freshman year, he felt alienated from some of his introductory classes. “I would think to myself, What the hell am I doing, sitting watching this professor give the same lecture that he has given every year for the last 30 years?” But he knew the answer even then: He was there to subsidize that professor’s research. At America’s research universities, teaching is a secondary job activity, at times neglected by faculty who want to devote as much time as possible to writing grants, running labs, and publishing academic papers. The classroom experience was suffering even before AI came onto the scene.

Now professors face their own temptations from AI, which can enable them to get more work done, and faster, just as it does for students. I’ve heard from colleagues who admit to using AI-generated recommendation letters and course syllabi. Others clearly use AI to write up their research. And still more are eager to discuss the wholesome-seeming ways they have been putting the technology to use—by simulating interactions with historical authors, for example, or launching minors in applied AI.

But students seem to want a deeper sort of classroom innovation. They’re not looking for gimmicks—such as courses that use AI only to make boring topics seem more current. Students like Lieber, who sees his college education as a means of setting himself up for his career, are demanding something more. Instead of being required to take tests and write in-class essays, they want to do more project-based learning—with assignments that “emulate the real world,” as Lieber put it.

But designing courses of this kind, which resist AI shortcuts, would require professors to undertake new and time-consuming labor themselves. That assignment comes at the worst possible time. Universities have been under systematic attack since President Donald Trump took office in January. Funding for research has been cut, canceled, disrupted, or stymied for months. Labs have laid off workers. Degree programs have cut doctoral admissions. Multi-center research projects have been put on hold. The “college experience” that Americans have pursued for generations may soon be over.

[Read: The end of college life]

The existence of these stressors puts higher ed at greater risk from AI. Now professors find themselves with even more demands than they anticipated and fewer ways to get them done. The best, and perhaps the only, way out of AI’s college takeover would be to embark on a redesign of classroom practice. But with so many other things to worry about, who has the time? In this way, professors face the same challenge as their students in the year ahead: A college education will be what they make of it too. At some point, everyone on campus will have to do the work.

Why Is Airplane Wi-Fi Still So Bad?

2025-07-25T14:07:00-04:00

“Wi-Fi is available on this flight,” the flight attendant announced on a recent trip I took from New York City to St. Louis. She recited her routine by rote, and Wi-Fi is among the details that now need to be conveyed, along with explaining how to use a seatbelt and enjoining passengers not to smoke e-cigarettes on board.

But when the time came to use the Wi-Fi, the service didn’t work. Eventually, enough people noticed this that the crew “rebooted” it, after which it still didn’t work. A new announcement acknowledged that Wi-Fi was, in fact, not available on this flight (and offered an apology). This was the can’t even access the portal kind of failure, but I’ve frequently encountered others, including can log in but not connect and so slow as to be worse than nothing. And then, at other times, the internet works great—as reliably as it does in an office building.

For two decades now, in-flight Wi-Fi has occupied this limbo between miracle and catastrophe. Way back in 2008, on Conan O’Brien’s late-night show, Louis C.K. told the story of a man who was complaining about the in-flight Wi-Fi not working mere moments after learning of its existence. “Everything is amazing right now and nobody’s happy,” the comedian joked. The bit was never quite right—nobody was happy, because services such as in-flight Wi-Fi were not yet amazing, actually. A chasm separated the service’s promise and its reality.

Today, 17 years later, I sense that same distance when I try to go online in the air. The matter feels more urgent now that more airlines, including JetBlue, Delta, and soon American and United, are offering free, purportedly better in-flight Wi-Fi (mainly to loyalty members so far). Air travel is neither a haven for offline delight nor a reliable place to carry out normal online life. Either option would be welcome, because each would be definitive. Instead, one is left to wonder if the hours about to be spent in flight can be filled with scrolling, shopping, Slacking, and tapping at Google Docs—or not.

I set out to learn why. Is the issue technological? Are the airlines promising more service than they can deliver? Most of all, I wanted to know if this situation will ever be fixed, making airplane Wi-Fi feel as brisk and reliable as it does elsewhere. The answer, it turns out, is familiar: soon, any day now, probably next year.

* * *

Because it’s the thing they use most often and turn on directly, people use Wi-Fi as a nickname for internet access in general. (“The Wi-Fi is down,” your spouse or child might say.) But the Wi-Fi part of airplane Wi-Fi—the access points in the plane that appear as “Delta Wi-Fi” or whatever on your computer or smartphone—is almost never part of the problem. Instead, the problem is the pipe to which the Wi-Fi connects—the in-flight equivalent of the cable or fiber that delivers internet service to your house.

An airplane flies in the air, and there are two ways to get the internet to connect to such a place: from above or below. At first, the only option was down. If you’re old enough to remember the September 11 attacks, you might also recall the Airfone service on some airlines—a phone handset stuffed into the seatback. These phones used air-to-ground communication, meaning that the signal was sent from the plane to a relay on the ground. Airfone (and its competitors) were expensive, didn’t work well, and few people used them. But that technology would be repurposed for early in-flight internet, offered via providers such as Gogo Inflight.

Jack Mandala, the CEO of Seamless Air Alliance, a standards organization for in-flight connectivity, told me that air-to-ground works like your cellphone—the bottom of the plane needs a view (metaphorically speaking) of base stations from the air. That’s why, for a time, you could use in-flight internet only over 10,000 feet. It’s also why the service is unreliable. Just like your cellphone might hit a dead spot, so can your airplane. Air-to-ground bandwidth was limited, meaning that the service would get worse as more people on a plane used it. And finally, air-to-ground service operates extremely slowly when it sends data down to the ground—this is why sending an email attachment or texting an image from a plane can take an eternity, before possibly failing completely.

Going up instead of down mostly solved these issues. Around the time of Louis C.K.’s Conan bit, airlines began offering internet service to planes via satellite communication. The improved speed and reliability allowed JetBlue to provide the industry’s first free in-flight internet to commercial passengers, in 2013. According to Mandala, satellite services are easier to scale as more planes adopt them and more passengers use them. Satellite also has the benefit of being usable over water, in bad weather, and on the ground.

The problem is that having viable technology is different from rolling it out seamlessly everywhere. Doing so requires investing in the equipment and service, and that requires time and money. In 2019, Delta, for instance, made a commitment to roll out free Wi-Fi across its entire fleet. Joseph Eddy, the airline’s director of cabin and in-flight entertainment and connectivity, told me that Delta’s effort is still ongoing. Unlike hotels or convention centers, Eddy reminded me, aircraft are highly regulated. Each type of aircraft needs to be configured differently, and a big airline such as Delta—or American, which told me it will also soon have 1,500 aircraft of its own with Wi-Fi service—requires some planning. “We need to make software upgrades. We need to make sure we have all the satellite coverage that we need to ensure that we have enough capacity and the experience is as good as possible,” Heather Garboden, American Airlines’ chief customer officer, told me.

But, hold up: American is the carrier I fly most these days, and I keep finding myself unable to use the internet. Garboden confirmed that American is still transitioning its regional jets to satellite service—many are still using air-to-ground. And that’s exactly the kind of plane I was on from New York. Delta’s Eddy told me that its regional jets and some short-haul planes, including the Boeing 717, are also still operating on air-to-ground service.

In both cases, the airlines made a deliberate choice to invest first in the routes and planes that carry the most passengers—big, mainline jets. That means that if you’re flying on a long flight across or between continents, or on an airline with fewer types of planes, such as JetBlue or Southwest, you might have a better shot at reliable internet. And if you’re on a small or regional jet, chances are greater that the Wi-Fi won’t work, or won’t work well. Eddy told me that Bombardier CRJ regional jets have proved more troubling to certify for the satellite antennas that sit on top of the fuselage, because of the aircraft’s rear-mounted engines. “You can’t allow any form of debris to fly off the antenna at all,” he said. If you board a plane and Wi-Fi isn’t available on the ground, that’s a sign that your aircraft is still using air-to-ground service. Good luck.

* * *

Beyond the technology itself, the expectation of always being connected is also driving flier perceptions of in-flight internet performance. Fliers are only now starting to take in-flight internet access as a given, rather than viewing it as a surcharged luxury.

Eddy thinks the tide started to turn during COVID. Even though people weren’t flying as much, everyone became more familiar with digital tools—Zoom, but also Slack, Teams, Google Docs—that might once have been lesser known. When travel resumed, those expectations made in-flight Wi-Fi “significantly more important,” Eddy said. American Airlines’ Garboden added that a younger, always-online generation is buying tickets now—26 percent of the airline’s customers are Gen Z and younger, she told me. For both airlines, the evolution of in-flight entertainment has reinforced the need for internet service. American delivers its movies and shows directly to its passengers’ devices; once those people are already staring at their phones, habit makes them expect to be able to switch to email or a social-media app. But Delta, which offers seatback screens on most of its planes, believes that having a television in front of you also now implies the need for internet. “If you look at the younger generations, they’re at home watching Netflix and they’re playing on their phone. They’re doing both almost constantly,” Eddy said, adding that 20 percent of Delta’s Wi-Fi customers use more than one device at a time.

Competition and passenger expectations may be the key to making in-flight internet work for good. After 9/11, the domestic airline industry devolved into pure carriage, stripping away all comforts in the name of safety—and profit. That appears to be changing. Nomadix, the company that invented the enter-your-name-and-room-number hotel-internet service more than 25 years ago, told me that the quality of Wi-Fi is one of the top three factors in customer satisfaction at every hotel property. That’s because hotels are in the hospitality business, and catering to customer comfort (not to mention facilitating work for business travelers) is core to their success. Airlines haven’t been as concerned with making fliers content in the cabin, but both Delta and American admitted that in-flight internet service is transitioning from an amenity into part of the hard product. “You would expect that your seat is there, right? Wi-Fi has become that for us,” Eddy said. Almost overnight, he told me, Wi-Fi went from having no impact on people choosing Delta to being “more important than flight times and airports.”

For now, consistency is the missing ingredient. This is what Louis C.K. failed to grasp: The issue has never been the flying public’s unwillingness to marvel at the miracles of human invention, but rather, the fact that carriers appear to make promises and then fail to deliver on them. Now that customer expectations, technological feasibility, and airline investments all align, it should just be a matter of time before the air is as well connected as the ground. But how much time? Delta initially promised “fast, free Wi-Fi” across its global fleet by the end of 2024, but now the airline thinks reaching that milestone will take until the first half of 2026. Garboden said American is on track for early 2026. United also plans to offer free satellite Wi-Fi across its entire fleet, but offered no projected date for full rollout.

Like cabin safety or timely arrival, until every passenger on every flight feels confident that the internet will take off along with their bodies and their luggage, the service doesn’t really exist, because it can’t be relied upon. Internet in the air is both a concrete advancement that’s mature and widespread, and a conceptual one frequently deferred into the future. That future may come, and perhaps even soon. Or it might not. Just like the Wi-Fi on your next flight.

The AI Mirage

2025-07-14T11:41:59-04:00

“I’m not going to respond to that,” Siri responded. I had just cursed at it, and this was my passive-aggressive chastisement.

The cursing was, in my view, warranted. I was in my car, running errands, and had found myself in an unfamiliar part of town. I requested “directions to Lowe’s,” hoping to get routed to the big-box hardware store without taking my eyes off the road. But apparently Siri didn’t understand. “Which Lowe?” it asked, before displaying a list of people with the surname Lowe in my address book.

Are you kidding me? Not only was the response incoherent in context, but also, only one of the Lowe entries in my contacts included an address anyway, and it was 800 miles away—an unlikely match compared with the store’s address. AI may not ever accomplish all of the things the tech companies say it will—but it seems that, at the very least, computers should be smarter now than they were 10 or 15 years ago.

It turns out that I would have needed an entirely new phone for Siri to have surmised that I wanted to go to the store. Craig Federighi, Apple’s senior vice president of software engineering, said in an interview last month that the latest version of Siri has “better conversational context”—the sort of thing that should help the software know when I’m asking to be guided to the home-improvement store rather than to a guy called Lowe. But my iPhone apparently isn’t new enough for this update. I would need cutting-edge artificial intelligence to get directions to Lowe’s.

[Read: The entire internet is reverting to beta]

This is effectively Apple’s entire pitch for AI. When it launched Apple Intelligence (the company’s name for the AI stuff in its operating systems) last year, the world’s third-most-valuable company promised a rich, contextual understanding of all your data, and the capacity to interact with it through ordinary phrases on your iPhone, iPad, or Mac. For example, according to Apple, you would be able to ask Siri to “send the photos from the barbecue on Saturday to Malia.”

But in my experience, you cannot ask even the souped-up Siri to do things like this. I embarked on a modest test of Apple Intelligence on my Mac, which can handle the feature. It failed to search my email, no matter how I phrased my command. When I tried to use Siri to locate a PDF of a property-survey report that I had saved onto my computer, it attempted to delegate the task to ChatGPT. Fine. But ChatGPT provided only a guide to finding a survey of a property in San Francisco, a city in which I do not live. Perhaps I could go more general. I typed into Siri: “Can you help me find files on my computer?” It directed me to open Finder (the Mac’s file manager) and look there. The AI was telling me to do the work myself. Finally, I thought I would try something like Apple’s own example. I told Siri to “show me photos I have taken of barbecue,” which resulted in a grid of images—all of which were stock photos from the internet, not pictures from my library.

These limitations are different from ChatGPT’s tendency to confidently make up stories and pass them off as fact. At least that error yields an answer to the question posed, albeit an inaccurate one. Apple Intelligence doesn’t even appear to understand the question. This might not seem like a problem if you don’t use Apple products or are content to rawdog your way to Lowe’s. But it does reveal a sad state of affairs for computing. For years, we’ve been told that frictionless interactions with our devices will eventually be commonplace. Now we’re seeing how little progress has been made toward this goal.

I asked Apple about the problems I’m having with Apple Intelligence, and it more or less confirmed that the product doesn’t work—yet. Apple’s position is that the 2024 announcement, featuring Malia and the cookout, represents a vision for what Siri can and should do. The company expects that work on functionality of this kind will continue into 2026, and it showed me a host of other forthcoming AI tools, including one with the ability to recognize an event in a screenshot of a text message and add the info to a calendar, or to highlight an object in a photo and search for similar ones on Google or Etsy. I also saw a demo of live language translation on a phone call, updated AI-created emoji, and tools to refine what you’ve written inside emails and in Apple software. Interesting, but in my mind, all of these features change how you can use a computer; they don’t improve the existing ways.

After rolling around in my head the idea that Apple Intelligence represents a vision for how a computer should work, I remembered that Apple first expressed this vision back in 1987, in a concept video for a product called Knowledge Navigator. The short film depicts a university professor carrying out various actions of daily and professional life by speaking directly to a personified software assistant on a tablet-like computer—all of the things I long to do with my computer 38 years hence. Knowledge Navigator, per the video, could synthesize information from various sources, responding to a user’s requests to pull up various papers and data. “Let me see the lecture notes from last semester,” the professor said, and the computer carried out the task. While the professor perused articles, the computer was able to identify one by a colleague, find her contact info, and call her upon his request.

Although obscure outside computer-history circles, Knowledge Navigator is legendary in Silicon Valley. It built on previous, equally fabled visions for computing, including Alan Kay’s 1972 proposal for a tablet computer he called DynaBook. Apple would eventually realize the form of that idea in the iPad. But the vision of Knowledge Navigator wasn’t really about how a device would look or feel. It was about what it would do: allow one to integrate all the aspects of a (then-still-theoretical) digital life by speaking to a virtual agent, Star Trek style. Today, this dream feels technologically feasible, yet it is still, apparently, just out of reach. (Federighi promised in the June interview that a better Siri was right around the corner, with “much higher quality and much better capability.”)

Apple Intelligence—really, generative AI overall—emphasizes a sad reality. The history of personal-computer interfaces is also a history of disappointments. At first, users had to type to do things with files and programs, using esoteric commands to navigate up and down the directory structures that contained them. The graphical user interface, which Apple popularized, adapted that file-and-folder paradigm into an abstraction of a desktop, where users would click and move those files around. But progress produced confusion. Eventually, as hard disks swelled and email collected, we ended up with so much digital stuff that finding it through virtualized rummaging became difficult. Text commands returned via features such as Apple’s Spotlight, which allows a user to type the name of a file or program, just as they might have done 50 years ago.

But now the entire information space is a part of the computer interface. The location and route to Lowe’s gets intermixed with people named Lowe in my personal address book. A cookout might be a particular event I attended, or it might be an abstraction tagged in online images. This is nothing new, of course; for decades now, using a computer has meant being online, and the conglomeration of digital materials in your head, on your hard disk, and on the internet often cause trouble. When you’re searching the web, Google asks if you’re perhaps really looking for the thing it deems more common based on other people’s behavior, rather than the thing you typed. And iCloud Drive helpfully uploads your files to the cloud to save disk space, but then you can’t access them on an airplane without Wi-Fi service. We are drowning in data but somehow unable to drink from its wellspring.

In principle, AI should solve this. Services such as ChatGPT, built on large language models that are trained on vast quantities of online and offline data, promised to domesticate the internet’s wilds. And for all their risk of fabrication and hallucination, LLMs really do deliver on that front. If you want to know if there exists a lens with specific properties compatible with a particular model of camera, or seek advice on how to carry out a plumbing repair, ChatGPT can probably be of use. But ChatGPT is much less likely to help you make sense of your inbox or your files, partly because it hasn’t been trained on them—and partly because it aspires to become a god rather than a servant.

[Read: The AI birthday letter that blew me away]

Apple Intelligence was supposed to fill that gap, and to do so distinctively. Knowledge Navigator never got built, but it was massively influential within the tech industry as a vision of a computing experience; it shows that Apple has expressed this goal for decades, if under different technological conditions and executive leadership. Other companies, including Google, are now making progress toward that aim too. But Apple is in a unique position to carry out the vision. It is primarily a personal-computer-hardware business focused on the relationship between the user and the device (and their own data) instead of the relationship between the user and the internet, which is how nearly every other Big Tech company operates. Apple Intelligence would make sense of all your personal information and grant new-and-improved access to it via Siri, which would finally realize its purpose as an AI-driven, natural-language interface to all that data. As the company has already done for decades, Apple would leave the messy internet mostly to others and focus instead on the device itself.

That idea is still a good one. Using a computer to navigate my work or home life remains strangely difficult. Calendars don’t synchronize properly. Email search still doesn’t work right, for some reason. Files are all over the place, in various apps and services, and who can remember where? If computationalists can’t even make AI run computing machines effectively, no one will ever believe that they can do so for anything—let alone everything—else.

The End of Airport Shoe-Screening Is Populism Theater

2025-07-10T16:33:20-04:00

Air travelers in America shall no more doff their chukkas, their wedges, their wingtips, their espadrilles, or their Mary Janes, according to a rule-change announced by Department of Homeland Security Secretary Kristi Noem on Tuesday. It’s been more than two decades since the Transportation Security Administration started putting people’s footwear through its scanners, after a man named Richard Reid tried and failed to detonate his high-top sneakers on a flight to Miami in December 2001. Indeed, the requirement has been in place so long that my adult children, who were born just before and after the September 11 attacks, didn’t even know its rationale. Feeling the cold airline-terminal floor through socks has been, for them, a lifelong ritual—as fundamental to the experience of flight as narrow seats and insufficient overhead bins.

The TSA’s mandate to go shoeless, like the volume limit on toiletry items (to thwart the assembly of explosives from liquids) and the need to remove laptops from carry-on bags (to better examine them for hidden threats), came to give the mere appearance of vigilance: not security but security theater. From the start, it provided newly federalized and uniformed TSA agents with stuff to do at every moment, and government officials with the chance to embrace “an abundance of caution,” a stock idea that can transform almost any inconvenience into leadership. Now, by closing the curtain on the shoe requirements, Noem has indulged in a rival form of spectacle: populism theater. Her new policy gives citizens something they actually want, and something that has until this point been reserved for upscale travelers who pay for premium airport-security-hopping services. But with this week’s change, the system hasn’t really been democratized so much as made indifferent. In this case, the fact of the TSA’s doing less—and caring less—just happens to be helpful.

[Read: The carry-on-baggage bubble is about to pop]

In its earliest phase, the shoe-removal policy was applied haphazardly, showing up from time to time and terminal to terminal in response to ever-shifting, secret intelligence on terrorist threats tracked by the Department of Homeland Security. Where the new form of screening was in place, it served not only to avert future shoe bombs but also to speed up the queue. Metal detectors had been tuned to be more sensitive, and the metal shank inside the soles of many shoes, installed to provide support, often set them off. (In response, some major footwear brands, including Rockport and Timberland, rushed out lines of shoes with plastic shanks that were marketed as being “security friendly.”)

By the summer of 2003, the policy had become more formalized; the TSA started “strongly” recommending that all passengers everywhere remove their shoes, or else risk being subject to a secondary screening. Speaking to The New York Times, a TSA representative said this new approach would “ensure that the experience you have in one airport is similar to the experience you have in another airport coast to coast.” Three years later, the policy of universal urging was made into a hard rule: Now your shoes had to come off, no matter what.

Although footwear checks applied to all in principle, some individuals—especially those deemed suspicious on the basis of their looks, or who evinced anxiety—were getting more aggressive treatment from the screeners. The system seemed unfair for some, and also far too burdensome for everyone. Why couldn’t some new and better form of scanner be invented, one that could spot a shoe explosive even as the wearer stood there? Would Americans be padding across the gross airport floors forever, just because of Richard Reid?

Better technology should have been the answer. In the decade after 9/11, tech firms completely reinvented everyday life: Web search, broadband, mobile telephony, e-commerce, smartphones, social networking, and real-time document collaboration all became routine. Back in 2002, many travelers would not have had so much as a flip phone in their carry-ons; 10 years later, most were toting handheld supercomputers. Yet when it came to building new devices for screening shoes, very little was accomplished. DHS spent millions of dollars in an effort to buy or subcontract the development of next-generation scanners that could avert sole-borne risks in airports, to no avail. (During this time, airport screening’s most significant innovation was the gray plastic bin into which you might hurl your pumps, boots, or loafers.) Shoe removal would “be a part of air travel for the foreseeable future,” a TSA spokesperson somberly announced in 2012, after another four experimental scanners had failed in real-world testing.

But a different way to solve the problem also started to emerge that summer: It turned out just to be money. The privately operated Clear service was launched in airports, giving travelers willing to pay a couple of hundred dollars a year and hand over their biometrics the ability to shortcut the screening line. And when the government’s own pay-for-comfort airport-security service, TSA PreCheck, rolled out widely in 2013, enrollees could finally forgo the lingering inconvenience of taking off their shoes. PreCheck also let them keep their laptops packed and their toiletries inside their bags. For a time, airline flyers with elite status got special access to both PreCheck and Clear.

This would be right in line with other trends of the early 2010s, when the VIP experience was being sold in a thousand different ways. Pay-to-play became a way of life. It’s hard to remember anymore, but before ride-hailing apps were available for nearly everyone, private cars were associated with late-night talk-show guests and people being shuttled to airports directly after giving conference keynotes. The precursors to the modern smartphone, such as the BlackBerry, were originally made for important executives before everyone adopted the air of importance. Since then, the whole economy has shifted upmarket. Those with money can now buy online memberships that get them tables at restaurants or tickets to shows whenever they want. Even Disneyland lets you pay to skip ahead in line.

Trading cash for the right to get through airport security with your shoes on prefigured all this and made it visible for everyone to see. Being in the TSA PreCheck queue not only gave you quick, shod access to the terminal; it also offered a perch from which to look down on the rabble nearby, stripped down to their socks and belt loops, presenting their shampoos and ointments, and unsheathing their electronics. What a bunch of losers, frequent fliers might think, before ascending to the airline club in their Lobbs or Louboutins.

It’s surely long past time to broaden out this special privilege and to stop demanding that every other person among the 1 billion annual air passengers in the United States take off their shoes because one guy tried to hide a bomb in his sneakers a quarter century ago. But the termination of the policy does not feel justified by any new development in science, technology, intelligence, or geopolitics. In announcing the change, Noem gave no satisfying explanation. She said only that it was enabled by the presence of “multi-layers of screening,” new scanners, more personnel, and Real ID—a nationwide identification system that was ginned up by Congress 20 years ago and somehow still has not been fully implemented.

By all appearances, the rule on shoes was not rescinded just because rescinding it happens to make sense. Rather, the change was made because the terror-hardened discipline of the millennium’s beginning has finally, fully been replaced by nihilism. These days, you board a plane that might or might not be flight-worthy, regulated by a shrunken-down Federal Aviation Administration, routed by an air-traffic-control system undermined by neglect and disdain. The president blamed a fatal plane collision on diversity programs, while selling access to the White House in plain view. No one seems to care. But at least you’ll be able to keep your shoes on before lifting off into America’s sunset.

Nobody Cares If Music Is Real Anymore

2025-07-04T07:00:00-04:00

The traffic receded as Chicago withdrew into the distance behind me on Interstate 90. Barns and trees dotted the horizon. The speakers in my rental car, playing Spotify from my smartphone, put out the opening riff of a laid-back psychedelic-rock song. When the lyrics came, delivered in a folksy vibrato, they matched my mood: “Smoke in the sky / No peace found,” the band’s vocalist sang.

Except perhaps he didn’t really sing, because he doesn’t exist. By all appearances, neither does the band, called the Velvet Sundown. Its music, lyrics, and album art may be AI inventions. Same goes for the photos of the band. Social-media accounts associated with the band have been coy on the subject: “They said we’re not real. Maybe you aren’t either,” one Velvet Sundown post declares. (That account did not respond to a request for comment via direct message.) Whatever its provenance, the Velvet Sundown seems to be successful: It released two albums last month alone, with a third on its way. And with more than 850,000 monthly listeners on Spotify, its reach exceeds that of the late-’80s MTV staple Martika or the hard-bop jazz saxophonist Cannonball Adderley. As for the music: You know, it’s not bad.

It’s not good either. It’s more like nothing—not good or bad, aesthetically or morally. Having listened to both of the Velvet Sundown’s albums as I drove from Chicago to Madison, Wisconsin, earlier this week, I discovered that what may now be the most successful AI group on Spotify is merely, profoundly, and disturbingly innocuous. In that sense, it signifies the fate of music that is streamed online and then imbibed while one drives, cooks, cleans, works, exercises, or does any other prosaic act. Long before generative AI began its takeover of the internet, streaming music had turned anodyne—a vehicle for vibes, not for active listening. A single road trip with the Velvet Sundown was enough to prove this point: A major subset of the music that we listen to today might as well have been made by a machine.

The technical quilt that was necessary to produce an AI album has been assembling for some time. Large language models such as ChatGPT can produce plausible song lyrics, liner notes, and other textual material. Software such as Suno can, based on text prompts, create songs with both instrumentation and vocals. Image generators can be directed to create illustrated compositions for album art and realistic images of a band and its members, and then maintain the appearance of those people across multiple images. When I got to Madison, I signed up for Suno’s service. Mere moments later, I had created my own psychedelic-rock, road-trip-themed jam, a bit more amplified and less sitar-adjacent than the Velvet Sundown’s. I didn’t even have to name the track; Suno dubbed it “Endless Highway” on my behalf. “Rubber burns, the map fades away / Chasing the ghosts of yesterday,” its fake male vocalist intoned. Sure, fine.

But cultural circumstances have also made AI music tolerable, and even welcome to some listeners. At the turn of the century, Napster made digital music free, and the iPod made it legitimate. You could carry a whole record store in your pocket. Soon after, Spotify, which became the biggest music-streaming service, started curating and then algorithmically generating playlists, which gave listeners recommendations for new music and offered easy clicks into hours of sound in any subgenre, real or invented—acid jazz, holiday bossa nova, whatever. Even just the phrase lazy Sunday could be turned into a playlist. So could lawn mowing or baking. Whatever Spotify put into your queue was good enough, because you could always skip ahead or plug in a new prompt.

Real or not, the Velvet Sundown feels more like a playlist than a band. Its “Verified Artist” description on Spotify used to read, “Their sound mixes textures of ’70s psychedelic alt-rock and folk rock, yet it blends effortlessly with modern alt-pop and indie structures.” That assembly of influences, stretching across half a century, appears with greater and lesser prevalence in each of the band’s numbers. “As the Silence Falls” feels indie folk, with washed-out guitars and soft vocals; “Smoke and Silence” is more bluesy, with stronger vocals and a classic-rock feel. From track to track, the singer’s voice seems to change in tone too—perhaps a quirk of generativity—making the collection feel less like a purposeful LP and more like a blind-bag gamble.

Music used to define someone’s identity: punk, rock, country, alternative, and so forth. Asking “What music do you like?” could elicit a person’s taste, values, and fashion sense. The rockers might hang out behind the gym and smoke cigarettes; they were a clique just like the jocks and the nerds. Finding, joining, and deepening a connection to a music subculture required effort; you had to find the right venues, records, zines, or crowd. In that era, music was tribal. A relationship with the Sisters of Mercy, Guns N’ Roses, or Bauhaus represented a commitment.

Not so much today. The internet has fragmented and flattened subcultures. The Velvet Sundown’s puppeteers present the band’s soft pastiche of genres—psychedelic, folk, indie—as sophisticated fusion, but of course it’s nothing more than a careless smear of stylistic averages. Psychedelic, folk, and indie rock each in their own way have something to say, musically and lyrically—about musical convention, spirituality, introspection, or social and political circumstances. The Velvet Sundown doesn’t seem to care about any of those things.

This approach appears to be serving the band or its creators very well. The Velvet Sundown may actually appeal to people. None of its tracks go hard; instead, each one offers something slightly different—a sitar lick, a blues-guitar solo, a folk-adjacent country twang—that might prove palatable to any given listener. Perhaps no human artist could tolerate producing such soulless lackluster, but an AI is unburdened by shame.

The lyrics’ milquetoast moodiness may also contribute to the band’s listener numbers. Each line is short, and the phrases barely connect to one another, making it easy for listeners to hear whatever they might want to hear: “Dust on the wind / Boots on the ground / Smoke in the sky / No peace found.” Really makes you think, until you realize that, no, it doesn’t at all. Where the music engages with the political commitments that often characterize its influences, it does so in a way that could mean anything. Take the chorus of “End the Pain,” one of the band’s top songs on Spotify. Singing with folk-rock urgency, the alleged “frontman and mellowtron sorcerer” Gabe Farrow pleads, “No more guns, no more graves / Send no heroes, just the brave.” These words convey the sensibility of an anti-war anthem, but they offer so little detail that the song could adequately service supporters or detractors of any conflict, past or present.

Anonymous and mild sensibilities have currency because today’s music—whether created and curated by humans or machines—is so often used to make people feel nothing instead of something. In open-plan offices, people started donning headphones to gain some semblance of privacy. At home, they do the same to mask the sound of traffic or their roommates’ Zoom calls. Internet-connected, whole-house audio systems can turn any room into a souped-up, algorithmic white-noise machine that sounds like Italo disco or chillhop in the way that LaCroix tastes like lime. The music that is best adapted for these settings is that which descends from what Brian Eno dubbed, on his 1979 album, Music for Airports, “ambient.” This music is not meant to be listened to directly; it’s used to drown out everything else.

As I drove amid the cornfields on I-90, the Velvet Sundown did just that. The band’s tracks were not satisfying in any way, but they were apt. I was on the road, but I could be anywhere—awaiting a Pilates class, paying for deli meat, scrolling through internet memes—and the sound would hit the mark.

And the worst part was that it was fine. It was fine! To my great embarrassment, the Velvet Sundown’s songs even managed to worm their way into my brain. Did I like their music? No, but my aesthetic judgment had given over to its vibes, that contemporary euphemism for ultra-processed atmosphere.

How far could I push this feeling? Returning to the car after a refreshment stop, I tried to make Spotify go meta on the band: I asked the app to generate a playlist made from songs that are similar to the Velvet Sundown’s. A list appeared of bands I didn’t recognize. Many seemed a little off: Appalachian White Lightning and Flaherty Brotherhood sounded like they might be AI acts as well. (A little Googling revealed that others suspect the same.) I suppose this makes sense; I was asking the algorithm to give me a channel of sanitized, inauthentic-seeming psychedelic-folk-indie rock, and it delivered. I pondered for a moment whether any of the other artists on my custom playlist (the South Carolina folk-rock singer-songwriter Johnny Delaware? The Belgian folk-pop quartet Lemon Straw?) might be fake—and how one might try to suss that out.

The question felt exhausting, so I switched back to the Velvet Sundown. As I drove and the music played, I felt nothing—but I felt that nothing with increasing acuteness. I was neither moved nor sad nor pensive, just aware of the fact that my body and mind exist in a tenuous zizz somewhere between life, death, and computers. This is second-order music listening, in which you experience the idea of listening to music. What better band to provide that service than one that doesn’t even exist?

But looking toward the blushing sky ahead of me, I realized that I didn’t even want this music to be art, or to feel that I was communing with its makers. I simply hoped to think and feel as little as possible while piloting my big car through the empty evening of America. This music—perhaps most music now—is not for dancing or even for airports; it’s for the void. I pressed “Play” and gripped the wheel and accelerated back onto the tollway, as the machines lulled me into oblivion.

A Computer Wrote My Mother’s Obituary

2025-06-10T14:04:45-04:00

The funeral director said “AI” as if it were a normal element of memorial services, like caskets or flowers. Of all places, I had not expected artificial intelligence to follow me into the small, windowless room of the mortuary. But here it was, ready to assist me in the task of making sense of death.

It was already Wednesday, and I’d just learned that I had to write an obituary for my mother by Thursday afternoon if I wanted it to run in Sunday’s paper. AI could help me do this. The software would compose the notice for me.

As a professional writer, my first thought was that this would be unnecessary, at best. At worst, it would be an outrage. The philosopher Martin Heidegger held that someone’s death is a thing that is truly their own. Now I should ask a computer to announce my mother’s, by way of a statistical model?

“Did you say AI?” I asked the funeral director, thinking I must have been dissociating. But yes, she did. As we talked some more, my skepticism faded. The obituary is a specialized form. When a person of note dies, many newspapers will run a piece that was commissioned and produced years in advance: a profile of the deceased. But when a normal person dies—and this applies to most of us—the obituary is something else: not a standard piece of journalistic writing, but a formal notice, composed in brief, that also serves to celebrate the person’s life. I had no experience in producing anything like the latter. The option to use AI was welcome news.

After all, there were lots of other things to do. The obituary was one of dozens of details I would have to address on short notice. A family in grief must choose a disposition method for their loved one, and perhaps arrange a viewing. They must plan for services, choose floral arrangements or other accessories, select proper clothing for the deceased, and process a large amount of paperwork. Amid these and other tasks, I found that I was grateful for the possibility of any help at all, even from a computer that cannot know a mother’s love or mourn her passing.

The funeral director told me I would be given access to this AI tool in the funeral-planning online account that she had already created for me. I still had a few misgivings. Would I be sullying Mom’s memory by doing this? I glanced over at an advertisement for another high-tech service—one that could make lab-grown diamonds from my mother’s ashes or her hair. Having an AI write her obituary seemed pretty tame in comparison. “Show me how to do it,” I said.

Actually getting a computer to do the work proved unexpectedly difficult. Over the next 24 hours, the funeral director and I exchanged the kind of emails you might swap with office tech support while trying to connect to the shared printer. I was able to log in to the funeral portal (the funeral portal!) and click into the obituary section, but no AI option appeared. The funeral director sent over a screenshot of her display. “It may look slightly different on your end,” she wrote. I sent a screenshot back: “That interface is not visible to me.” Web-browser compatibility was discussed, then dismissed. The back-and-forth made me realize that Mom’s memorial would be no more sullied by AI than it was by the very fact of using this software—a kind of Workday app for death and burial.

In the end, the software failed us. My funeral director couldn’t figure out how to give me access to the AI obituary writer, so I had to write one myself, using my brain and fingertips. I did what AI is best at: copying a formula. I opened up my dad’s obituary, which Mom had written a couple of years earlier, and mirrored its format and structure. Dates and locations of birth and death, surviving family, professional life, interests. I was the computer now, entering data into a pre-provided template.

[Read: A secret history of the obituary page]

When I finally did get the chance to try the AI obituary writer a few weeks later—after reaching out to Passare, the company behind it—I found its output more creative than mine, and somehow more personal. Like everything else, the funeral-services industry is now operated by cloud-based software-as-a-service companies. Passare is among them, and offers back-office software for funeral-home management along with family-facing funeral-planning tools.

Josh McQueen, the company’s vice president of marketing and product, explained why my earlier attempt to use the obituary-writing tool had failed: The funeral home must have had that feature set for staff-only access, which some businesses prefer. Then he gave me access to a mock funeral for the fictional departed John Smith so I could finally give it a go.

I couldn’t change John Smith’s name, but I pretended I was writing the obituary for my mother instead. Using simple web forms, I put in her education and employment information, some life events that corresponded to her “passions” and “achievements,” and a few facts about relevant family members who had survived her or preceded her in death. These had to be entered one by one, choosing the type of relation from a drop-down and then checking a box to indicate whether the person in question was deceased. I felt like I was cataloging livestock.

From there, Passare’s software, which is built on top of ChatGPT technology, generated an obituary. And you know what—it was pretty good. Most of all, it was done, and with minimal effort from me. Here’s an excerpt, with John Smith’s name and pronouns swapped out for my mother’s, and a couple of other very small alterations to smooth out the language:

Sheila earned her bachelor’s degree and dedicated her career to managing her late husband David’s psychology private practice for decades. She was not only devoted to his work but also a dedicated caregiver for Dave in his later years. Throughout her life, Sheila nurtured his passions, which included playing music—especially the piano—and a deep appreciation for Native American art. She found joy in teaching skiing to children and sharing the vibrant personalities of her many pet birds.

The AI obituary can also be tuned by length and tone—formal, casual, poetic, celebratory. (The poetic version added flourishes such as “she found joy in the gentle keys of her piano, filling her home with music that echoed her spirit.”) Because an obituary is already a schematic form of writing, the AI’s results were not just satisfactory but excellent, even. And, of course, once the draft was done, I could adjust it as I wished.

“When we first started testing this, ChatGPT would just make up stories,” McQueen told me. It might assert that someone named Billy was often called Skippy, for example, and then concoct an anecdote to explain the fake nickname. This tendency of large language models, sometimes called hallucination, is caused by the technology’s complex statistical underpinnings. But Passare found this problem relatively easy to tame by adjusting the prompts it fed to ChatGPT behind the scenes. He said he hasn’t heard complaints about the service from any families who have used it.

Obituaries do seem well suited for an AI’s help. They’re short and easy to review for accuracy. They’re supposed to convey real human emotion and character, but in a format that is buttoned-up and professional, for a public audience rather than a private one. Like cover letters or wedding toasts, they represent an important and uncommon form of writing that in many cases must be done by someone who isn’t used to writing, yet who will care enough to polish up the finished product. An AI tool can make that effort easier and better.

And for me, at least, the tool’s inhumanity was also, in its way, a boon. My experience with the elder-care and death industries—assisted living, hospice, funeral homes—had already done a fair amount to alienate me from the token empathy of human beings. As Mom declined and I navigated her care and then her death, industry professionals were always offering me emotional support. They shared kind words in quiet rooms that sometimes had flowers on a table and refreshments. They truly wanted to help, but they were strangers, and I didn’t need their intimacy. I was only seeking guidance on logistics: How does all this work? What am I supposed to do? What choices must I make?

A person should not pretend to be a friend, and a computer should not pretend to be a person. In the narrow context of my mom’s obituary, the AI provided me with middle ground. It neither feigned connection nor replaced my human agency. It only helped—and it did so at a time when a little help was all I really wanted.

First My Mother Died. Then My Home Got Hit by a Tornado.

2025-05-22T10:38:42-04:00

The wind was whipping up, but I ignored it. I was at my house in St. Louis, on the phone with the rabbi who would officiate my mother’s funeral, a thousand miles away. We spoke about her life, her family, the service, and other matters both material and spiritual. Mom had been sick for well over a year, but she started declining rapidly in December. Late last month, she was admitted to hospice. Along with her nurses and aides, I helped tend to her frail form as she slowly ceased to be able to eat, to speak, to breathe. Finally relieved of pain, she allowed comfort to overtake her.

When the emergency alert blared on my smartphone, I told the rabbi that we should probably finish talking later. My wife had just raced down the stairs to the basement, calling for me to follow. I did, but also I lingered: The sky was so dark. I had never seen a storm like this before. Later I’d realize that’s because I had never been inside an EF-3 category tornado with 150-plus-mph winds, like the one that tore across metro St. Louis on Friday. But on my way to the basement, I didn’t know that. I took in the surreal, terrifying sight of a full-grown shingle oak scraping the ground. The storm seemed gentle to me in that moment, as it laid the tree to rest inside my yard. I saw it cradling the oak to its now-certain end, as I had done for my mother the week before.

My feeling of repose was gone by the time I reached the basement and heard windows shattering. Glass is a human invention, and its breakage is inevitably associated with human violence or a human accident: a burglar’s incursion, a child’s wayward baseball, a pogrom. I knew in my head that nature, too, can impose itself on the built environment, but still I was unprepared for the sensation of its happening.

As a midwesterner in the age of anthropogenic climate change, I have spent many hours in the basement waiting out tornado warnings. Normally, it’s boring to be down there in storm isolation, even though we all bring phones and tablets, and the power usually stays on. We might express frustration at the fact that official warnings rarely come to much. The tornadoes never pass through here, we say. They always move west of the city. As of Friday morning, I understood that tornadoes were unlikely; baseball-size hail was the greater concern. But when a tornado has begun to whirl around your home, a sense of smallness overtakes you. Who are you to think you know how any of this works?

[Read: The hybrid system that spots tornadoes]

In the basement, my wife held my daughter tightly, begging me to stop wandering toward the walls and windows. I didn’t do so out of bravado or even apprehension. I was enrapt. To watch the storm was to be a party to a power much greater than myself. As one gets older and more experienced, novel encounters become more precious. This one, embossed by the force of the powerful winds, was new to me. The philosopher Immanuel Kant thought that appreciating the sublime requires the safety of distance. Now I wondered whether he was wrong. Perhaps the sublime has to be confronted viscerally to be made complete, just like one cannot truly appreciate vertigo by watching roller coasters from the ground.

People lament and worry about the loss of human life. “I’m sorry for your loss,” they say when I tell them my mother died. “Is everyone okay?” they ask after the storm passes. At least five people were killed and dozens injured in St. Louis on Friday. But when we emerged from our homes to assess the outcome—which included a splay of tar roofing, air-conditioning condensers, and insulation hurled from neighboring buildings—it still didn’t feel right to relay the news that no one on our street had been hurt.

That’s because of the trees. The tornado appears to have begun in Clayton, a well-to-do municipality just west of St. Louis. It crossed the edge of Forest Park, site of the 1904 World’s Fair, and tore through residential neighborhoods as it moved northeast. Within them are residential streets planned in the late 19th century and built up in part by industrialists of the Gilded Age and progressive era. At the park and in the neighborhoods, the tree canopy has grown since then to some 80 feet in height. After a long and dreary winter, the pin oaks on my block, planted in tidy rows, had finally leafed out a few weeks earlier, casting an arch of shade over the whole street.

Almost all of them are gone now, felled whole or disfigured into shrapnel. To say they can’t be replaced isn’t quite right; it just takes decades to grow new ones. And yet, even this arboreal tragedy felt sublime, in its way: more than a century of slow progress wiped out in seconds. I will never see those trees again, not like that—but then again, neither would the people who first planted them in the early 1900s, when the saplings were too young to offer shade.

Trees are no less mortal than human beings. The pin oaks, by any measure, had already exceeded their typical lifespan of 100 to 120 years, and many had already suffered the ills of poorly drained soil and compaction. They’d been dying by the pair every year, but enough remained to give me and my neighbors the false impression that their shade was eternal, that we were owed it, that it was ours. The tornado ended that delusion.

At 75, my mother was young to die, by contemporary standards, but ancient by historical ones. Friends and family keep asking “What did she have?,” hoping for a simple answer. But what she had was something more amorphous, a set of interconnected but distinct ailments that, when blended together and seasoned by accident, led to a slow decline and then a quick one. To yearn for a tidy word—cancer, stroke—to name misfortune is to make a category error, like trying to lasso the ocean. It betrays the mystery of life and death, fortune and accident. It is no more or less unfair that this fate would befall her than that a tornado would careen across my fancy street. If such things happen to someone, why not us?

[Read: What the tornadoes in Nashville revealed]

Mom and Dad were married for 52 years before he died two years ago. They worked together and did everything else together, too, a feat that would make me crazy but that my mother embraced. My father had a disability—I wrote about it for The Atlantic—stemming from a terrible auto accident in his teens, which he always tried to mask. Sometimes, especially late in his life, my mother would say that she remained so attached to him in order to take care of him, which is true. But she also maintained that close connection by choice. Seeing her confined to the same hospital bed that he had used, in the same room, taking the same narcotics prescriptions, felt somehow apt. This, too, they would do together, if slightly apart.

Mom kept close tabs on the weather wherever I lived, which was always too far away, by her judgment. She would text or call when she saw storms in the forecast. Are you okay? she might ask. And I would play the role of churlish son, answering We’re fine mom, don’t worry, or The tornadoes always pass to the west, as if I had a say in the matter. But the one time she was finally right to be concerned, she couldn’t express the worry anymore. I am tempted to call this irony, but it is better named indifference.

What a shame that indifference is seen only in a negative light. The storm’s disregard was terrifying and awesome. I felt it in the basement as the gale whipped around my house, and then in the street, amid the fallen oaks and the hurtled air-conditioning condensers. And I’d felt the same sense of the sublime at Mom’s bedside earlier that week as her fever became terminal. Neither Mom nor I was targeted for calamity, but it found us nevertheless. The universe is indifferent, and that is terrifying, and that is beautiful.

American Panopticon

2025-04-27T08:00:00-04:00

If you have tips about DOGE and its data collection, you can contact Ian and Charlie on Signal at @ibogost.47 and @cwarzel.92.

If you were tasked with building a panopticon, your design might look a lot like the information stores of the U.S. federal government—a collection of large, complex agencies, each making use of enormous volumes of data provided by or collected from citizens.

The federal government is a veritable cosmos of information, made up of constellations of databases: The IRS gathers comprehensive financial and employment information from every taxpayer; the Department of Labor maintains the National Farmworker Jobs Program (NFJP) system, which collects the personal information of many workers; the Department of Homeland Security amasses data about the movements of every person who travels by air commercially or crosses the nation’s borders; the Drug Enforcement Administration tracks license plates scanned on American roads. And that’s only a minuscule sampling. More obscure agencies, such as the recently gutted Consumer Financial Protection Bureau, keep records of corporate trade secrets, credit reports, mortgage information, and other sensitive data, including lists of people who have fallen on financial hardship.

A fragile combination of decades-old laws, norms, and jungly bureaucracy has so far prevented repositories such as these from assembling into a centralized American surveillance state. But that appears to be changing. Since Donald Trump’s second inauguration, Elon Musk and the Department of Government Efficiency have systematically gained access to sensitive data across the federal government, and in ways that people in several agencies have described to us as both dangerous and disturbing. Despite DOGE’s stated mission, little efficiency seems to have been achieved. Now a new phase of Trump’s project is under way: Not only are individual agencies being breached, but the information they hold is being pooled together. The question is Why? And what does the administration intend to do with it?

In March, President Trump issued an executive order aiming to eliminate the data silos that keep everything separate. Historically, much of the data collected by the government had been heavily compartmentalized and secured; even for those legally authorized to see sensitive data, requesting access for use by another government agency is typically a painful process that requires justifying what you need, why you need it, and proving that it is used for those purposes only. Not so under Trump.

This is a perilous moment. Rapid technological advances over the past two decades have made data shedding ubiquitous—whether it comes from the devices everyone carries or the platforms we use to communicate with the world. As a society, we produce unfathomable quantities of information, and that information is easier to collect than ever before.

Illustration by Anson Chan

The government has tons of it, some of which is obvious—names, addresses, and census data—and much of which may surprise you. Consider, say, a limited tattoo database, created in 2014 by the National Institute of Standards and Technology, and distributed to multiple institutions for the purpose of training software systems to recognize common tattoos associated with gangs and criminal organizations. The FBI has its own “Next Generation Identification” biometric and criminal-history database program; the agency also has a facial-recognition apparatus capable of matching people against more than 640 million photos—a database made up of driver’s license and passport photos, as well as mug shots. The Social Security Administration keeps a master earnings file, which contains the “individual earnings histories for each of the 350+ million Social Security numbers that have been assigned to workers.” Other government databases contain secret whistleblower data. At the Department of Veterans Affairs, you’ll find granular mental-health information on former service members, including notes from therapy sessions, details about medication, and accounts of substance abuse. Government agencies including the IRS, the FBI, DHS, and the Department of Defense have all purchased cellphone-location data, and possibly collected them too, via secretive groups such as the National Geospatial-Intelligence Agency. That means the government has at least some ability to map or re-create the past everyday movements of some American citizens. This is hardly even a cursory list of what is publicly known.

Advancements in artificial intelligence promise to turn this unwieldy mass of data and metadata into something easily searchable, politically weaponizable, and maybe even profitable. DOGE is reportedly attempting to build a “master database” of immigrant data to aid in deportations; NIH Director Jay Bhattacharya has floated the possibility of an autism registry (though the administration quickly walked it back). America already has all the technology it needs to build a draconian surveillance society—the conditions for such a dystopia have been falling into place slowly over time, waiting for the right authoritarian to come along and use it to crack down on American privacy and freedom.

But what can an American authoritarian, or his private-sector accomplices, do with all the government’s data, both alone and combined with data from the private sector? To answer this question, we spoke with former government officials who have spent time in these systems and who know what information these agencies collect and how it is stored.

To a person, these experts are alarmed about the possibilities for harm, graft, and abuse. Today, they argued, Trump is targeting law firms, but DOGE data could allow him to target individual Americans at scale. For instance, they described how the government, aside from providing benefits, is also a debt collector on all kinds of federal loans. Those who struggle to repay, they said, could be punished beyond what’s possible now, by having professional licenses revoked or having their wages or bank accounts frozen.

Musk has long dreamed of an “everything app” that would combine banking, shopping, communication, and all other human affairs. Such a project would entail holding and connecting all the information those activities produce. Even if Musk were to step back from DOGE, he or his agents may still possess data they collected or gained access to in the organization’s ongoing federal-data heist. (Musk did not respond to emailed questions about this, nor any others we posed for this story.)

These data could also allow the government or, should they be shared, its private-sector allies to target big swaths of the population based on a supposed attribute or trait. Maybe you have information from background checks or health studies that allows you to punish people who have seen a therapist for mental illness. Or to terminate certain public benefits to anybody who has ever shown income above a particular threshold, claiming that they obviously don’t need public benefits because they once made a high salary. A pool of government data is especially powerful when combined with private-sector data, such as extremely comprehensive mobile-phone geolocation data. These actors could make inferences about actions, activities, or associates of almost anybody perceived as a government critic or dissident. These instances are hypothetical, but the government’s current use of combined data in service of deportations—and its refusal to offer credible evidence of wrongdoing for some of those deported—suggests that the administration is willing to use these data for its political aims.

Harrison Fields, a spokesperson for the White House, confirmed that DOGE is combining data that it has collected across agencies, but he did not respond to individual questions about which data it has or how it plans to safeguard citizens’ private information. “DOGE has been instrumental in enhancing data accuracy and streamlining internal processes across the federal government,” Fields told us in an emailed statement. “Through data sharing between agencies, departments are collaborating to identify fraud and prevent criminals from exploiting hardworking American taxpayers.”

For decades, government data have been both an asset and a liability, used and occasionally abused in service of its citizens or national security. Under Trump and DOGE, the proposition for the data’s use has been flipped. The sensitive and extensive collective store of information may still benefit some American citizens, but it is also being exploited to satisfy the whims and grievances of the president of the United States.

Trump and DOGE are not just undoing decades of privacy measures. They appear to be ignoring that they were ever written. Over and over, the federal experts we spoke with insisted that the very idea of connecting federal data is anathema. An employee in senior leadership at USAID told us that the systems operate on their own platforms with no interconnectivity by design. “There’s almost no data sharing between agencies,” said one former senior government technologist. That’s a good thing for privacy, but it makes it harder for agencies to work together for citizens’ benefit.

On occasions when sharing must happen, the Privacy Act of 1974 requires what’s called a Computer Matching Agreement, a written contract that establishes the terms of such sharing and to protect personal information in the process. A CMA is “a real pain in the ass,” according to the official, just one of the ways the government discourages information swapping as a default mode of operation. According to the USAID employee, workers in one agency do not and cannot even hold badges that grant them access to another agency—in part to prevent them from having access to an outside location where they might happen upon and exfiltrate information. So you can understand why someone with a stated mission to improve government efficiency might train their attention on centralizing government data—but you can also understand why there are rigorous rules that prevent that from happening. (The Privacy Act was passed to curtail abuses of power such as those exhibited in the Watergate and COINTELPRO scandals, in which the government conducted illegal surveillance against its citizens.)

The former technologist, who worked for the Biden administration, described a system he had tried to facilitate building at the General Services Administration that would provide agencies with income information in order to verify eligibility for various benefits, such as SNAP, Medicaid, and Pell Grants. A simple, basic service to verify income, available only to federal and state agencies that really needed it, seemed like it would be an easy success.

Illustration by Anson Chan

It never happened. (The former federal technologist blamed “enormous legal obstacles,” including the Privacy Act itself, policies at the Office of Management and Budget, and various court rulings.) The IRS even maintains an API—a way for computers to talk to one another—built to give the banking industry a way to verify someone’s income, for example to underwrite a mortgage application. But using that service inside the government—even though it was made by the federal government—was forbidden. The best option for agencies who wanted to do this was to ask citizens to prove their eligibility, or to pay a private vendor such as Equifax, which can leverage the full power of data brokering and other commercial means of acquiring information, to confirm it.

Even without regulatory hurdles, intermingling data may not be as straightforward as it seems. “Data isn’t what you’d imagine,” Erie Meyer, a founder of the U.S. Digital Service and the chief technologist for multiple agencies, including the CFPB, told us. “Sometimes it’s hard-paper information. It’s a mess.” Just because a federal agency holds certain information in documents, files, or records doesn’t mean that information is easily accessed, retrieved, or used. Your tax returns contain lots of information, including the charities to which you might have contributed and the companies that might have paid you as an employee or contractor. But in their normal state—as fields in the various schedules of your tax return, say—those data are not designed to be easily isolated and queried as if they were posts on social media.

An American surveillance society that fully stitched together the data the government already possesses would require officials to upend the existing rules, policies, and laws that protect sensitive information about Americans.

To this end, DOGE has strong-armed its way into federal agencies; intimidated, steamrolled, and fired many of their workers; entered their IT systems; and accessed some unknown quantity of the data they store. DOGE removes the safeguards that have protected controls for access, logs for activity, and of course the information itself. Borrowing language from IT management, the senior USAID employee called DOGE a kind of permission structure for privacy abuse.

But the federal technologist added something else: “We worship at the altar of tech.” Many Americans have at least a grudging respect for the private tech industry, which has changed the world, and quickly—a sharp contrast to the careful, if slow-moving, government. Booting out the bureaucrats in favor of technologists may look to some like liberation from mediocrity, even if it may lead to repression.

Musk has said that his goal with DOGE is to serve his country. He says he wants to “end the tyranny of bureaucracy.” But around Washington, people are asking one another what he really wants with all those data. Keys to the federal dataverse could, for example, be extremely useful to a highly ambitious man who is aggressively trying to win the AI race.

We already know that Musk’s people have access to large swaths of information from federal agencies—what we don’t know is what they’ve copied, exfiltrated, or otherwise taken with them. In theory, this material, whether usable together or not, could be recombined with other identifying information from private companies for all kinds of purposes. There has been speculation already that it could be fed into third-party large language models to train them or make the information more usable (Musk’s xAI has its own model, Grok); outside firms could use their own technologies to make sense of disparate sets of data, as well. Such approaches, the federal workers told us, could make it easier to turn previously obfuscated information, such as the individual elements of a tax return, into something to be mined.

Tech companies already collect as much information as possible not because they know exactly what it’s good for, but because they believe and assume—correctly—that it can provide value for them. They can and do use the data to target advertising, segment customers, perform customer-behavior analysis, carry out predictive analytics or forecasting, optimize resources or supply chains, assess security or fraud risk, make real-time business decisions and, these days, train AI models. The central concept of the so-called Big Data era is that data are an asset; they can be licensed, sold, and combined with other data for further use. In this sense, DOGE is the logical end point of the Big Data movement.

Collecting and then assembling data in the industrial way—just to have them in case they might be useful—would represent a huge and disturbing shift for the government. So much so that the federal workers we spoke with struggled even to make sense of the idea. They insisted that the government has always tried to serve the people rather than exploit them. And yet, this reversal matches the Trump transactional ethos perfectly—turning How can we serve our fellow Americans? into What’s in it for us?

Us, in this case, isn’t even the government, let alone your fellow Americans. It’s Trump’s business concerns; the private-sector ones that have supplicated to him; the interests of his friends and allies, including Musk, and other loyalists who enter their orbits. Once the laws, rules, and other safeguards that have prevented federal data from comingling fall away—and many of them already have in practice—previously firewalled federal data can be combined with private data sets, such as those held by Trump allies or associates, tech companies who want to get on the administration’s good side, or anyone else the administration can coerce.

Illustration by Anson Chan

Many Americans have felt resigned to the Big Data accrual of their information for years already. (Plenty of others simply don’t understand the scope of what they’ve given up, or don’t care.) Data breaches became banal—including at Equifax and even inside the government at the Office of Personnel Management. Some private firms, such as Palantir, already hold lucrative government data-intelligence contracts. As Wired recently reported, ICE cannot track “self-deportations” in near-real time—but Palantir can. Lisa Gordon, Palantir’s head of global communications, told us that the company does not “own, collect, sell or provide any data to our customers—government or commercial,” and that clients are ultimately in control of their information. However, she also added that Palantir “is accredited to secure a customer’s data to the highest standards of data privacy and classification.” Theoretically, even if federal data are stored by a third-party contractor, they are protected legally and contractually. But such guarantees might no longer matter if the government deems its own privacy laws irrelevant. Public data sets could become a gold mine if sold to private parties, though there is no evidence this is taking place.

The thought that the government would centralize or even give away citizen data for private use is scandalous. But it’s also, in a way, expected. The Vietnam War and Watergate gave Americans reasons to believe that the government can’t be trusted. The Cold War issued a constant, decades-long threat of annihilation and the necessary surveillance to avoid it. The War on Terror extended the logic into the 21st century. Optical, recording, and then computer technologies arose, offering new ways to watch the public. During the 2010s, Edward Snowden’s NSA surveillance leaks took place, and the Facebook–Cambridge Analytica scandal was brewing. By then, the 20th-century assumption that U.S. intelligence agencies were running mind-control experiments, infiltrating and disrupting civil-rights groups, or carrying out surreptitious missions at home like they do abroad had been fully internalized, and fused with the suspicion that Google, Facebook, Amazon, and Walmart were—in their own ways—following suit.

Earlier this month, The Washington Post reported that government agencies are combining data that are normally siloed so that identifying undocumented immigrants would be easier. At the Department of Labor, DOGE has gained access to sensitive data about immigrants and farmworkers, Wired reported. This and other reporting shows that DOGE seems to be particularly interested in finding ways to “cross-reference datasets and leverage access to sensitive SSA systems to effectively cut immigrants off from participating in the economy,” according to Wired.

A worst-case scenario is easy to imagine. Some of this information could be useful simply for blackmail—medical diagnoses and notes, federal taxes paid, cancellation of debt. In a kleptocracy, such data could be used against members of Congress and governors, or anyone disfavored by the state. Think of it as a domesticated, systemetized version of kompromat—like opposition research on steroids: Hey, Wisconsin is considering legislation that would be harmful to us. There are four legislators on the fence. Query the database; tell me what we’ve got on them.

Say you want to arrest or detain somebody—activists, journalists, anyone seen as a political enemy—even if just to intimidate them. An endless data set is an excellent way to find some retroactive justification. Meyer told us that the CFPB keeps detailed data on consumer complaints—which could also double as a fantastic list of the citizens already successfully targeted for scams, or people whose financial problems could help bad actors compromise them or recruit them for dirty work. Similarly, FTC, SEC, or CFPB data, which include subpoenaed trade secrets gathered during long investigations, could offer the ability for motivated actors to conduct insider trading at previously unthinkable scale. The world’s richest man may now have access to that information.

An authoritarian, surveillance-control state could be supercharged by mating exfiltrated, cleaned, and correlated government information with data from private stores, corporations who share their own data willingly or by force, data brokers, or other sources. What kind of actions could the government perform if it could combine, say, license plates seen at specific locations, airline passenger records, purchase histories from supermarket or drug-store loyalty cards, health-care patient records, DNS-lookup histories showing a person’s online activities, and tax-return data?

It could, for example, target for harassment people who deducted charitable contributions to the Palestine Children’s Relief Fund, drove or parked near mosques, and bought Halal-certified shampoos. It could intimidate citizens who reported income from Trump-antagonistic competitors or visited queer pornography websites. It could identify people who have traveled to Ukraine and also rely on prescription insulin, and then lean on insurance companies to deny their claims. These examples are all speculative and hypothetical, but they help demonstrate why Americans should care deeply about how the government intends to manage their private data.

A future, American version of the Chinese panopticon is not unimaginable, either: If the government could stop protests or dissent from happening in the first place by carrying out occasional crackdowns and arrests using available data, it could create a chilling effect. But even worse than a mirror of this particular flavor of authoritarianism is the possibility that it might never even need to be well built or accurate. These systems do not need to work properly to cause harm. Poorly combined data or hasty analysis by AI systems could upend the lives of people the government didn’t even mean to target.

“Americans are required to give lots of sensitive data to the government—like information about someone’s divorce to ensure child support is paid, or detailed records about their disability to receive Social Security Disability Insurance payments,” Sarah Esty, a former senior adviser for technology and delivery at the U.S. Department of Health and Human Services, told us. “They have done so based on faith that the government will protect that data, and confidence that only the people who are authorized and absolutely need the information to deliver the services will have access. If those safeguards are violated, even once, people will lose trust in the government, eroding its ability to run those services forever.” All of us have left huge, prominent data trails across the government and the private sector. Soon, and perhaps already, someone may pick up the scent.

A New Kart-Racing Appliance

2025-04-02T15:20:07-04:00

If your virtual kart-racing life was missing something, you’re in luck. Nintendo, the Japanese electronics manufacturer, announced its new Mario Kart appliance today. The Switch 2, which can be used handheld or connected to a television, allows players to race go-karts piloted by characters from the company’s entertainment franchises: Mario, Yoshi, Princess Peach.

The karting games that ran on previous appliances allowed racers to compete on only a series of discrete tracks. But the updated hardware allows for something else: Mario Kart World, as the new software is called, presents its users with the tantalizing prospect of a digital commute. Racers may now convey from one track to the next through a large and continuous simulated world. This new capacity will unlock other new ways to kart, among them 1980s-style arcade racing and more contemporary, open-world kart tourism.

Longtime kart racers will surely celebrate the opportunity to kart anew. Someone who might have played Mario Kart 8—the previous fully original home release in the franchise—in 2014, when they were 12, has now graduated college. In the gaps between soul-crushing weeks at an investment bank or a management consultancy, karting sons and daughters who became karting adults might sneak in a nostalgic trip or race with their aging parents or once-baby siblings, now adolescents.

To facilitate the process, Nintendo has finally improved its online kart-racing infrastructure. Its competitors Sony and Microsoft, whose entertainment appliances mostly facilitate simulated sports or ritualistic arena murder, have allowed players to connect by voice or even video while playing, both to coordinate matches and to issue racist or homophobic taunts. The Switch 2 finally adds this capacity to kart racing, deployed via a “C” (“Cart”? No, “Chat”) button on its controllers.

[Read: Video games are better without stories]

All of this kart racing comes at a hefty price: $450 for the appliance itself, or $500 for the device bundled with the Mario Kart software. Those who would choose to forgo the bundle in favor of purchasing inscrutably updated rehashes of previous works, such as embarrassing fantasy-adventure games and insipid party titles, will have to hand over $80 for Mario Kart World if they choose to add it later. That might put kart-based home entertainment out of reach for many Americans. But others will surely see the value in the Switch 2, given the appeal and frequency of these karting delights.

Games such as Mario Kart World will be delivered on cartridges matching the size and shape of those from the previous appliance. Those carts may not contain software, instead acting as dummy keys that will unlock a probably time-consuming download. In exchange for this inconvenience, players will be able to “gameshare” some software titles with up to four friends, allowing the games to trickle down, Reaganomics-style, from the wealthy to the aspirant underclasses (though even these paupers will apparently still have to pay for a separate Nintendo Switch Online subscription to voice- or videochat with their game-giving overlords). But not Mario Kart World, which is ineligible for gameshare. All citizens must purchase their own access to karting.

[Read: The quiet revolution of Animal Crossing]

Nintendo has also updated the guts of the Switch 2 kart appliance. It will finally be capable of using the entire 4K resolution of the televisions that were being sold back when your college graduate was still 12. Note that the appliance itself features an LCD screen rather than the rich OLED displays that have been commonplace in smartphones for the past decade or so.

Nintendo has also failed to heed the lessons from its previous Mario Kart appliance. That device, the Switch, featured finicky, removable Joy-Con kart-racing controllers. Inevitably, kart-racing fanciers elected to pay exorbitant prices for traditional, add-on controllers instead. A new version of those controllers is also on offer for the Switch 2, requiring a new investment of $80 each for a racing tether that features the new “C” button.

Will this new kart chaos be worthwhile? Emphatically yes. I spent $1,600 on a new washer-dryer this year, and I use it only once a week, whereas I kart (or long to kart) far more often. Similarly, a good countertop air fryer might cost hundreds of dollars; why not a karting appliance too? And like an air fryer, which can toast and roast in addition to convection bake, the Switch 2 Mario Kart appliance is also capable of supporting other Nintendo-crafted experiences, such as an ape-oriented romp game announced today and, perhaps eventually, attempts to rehabilitate the non-karting titles from which the Mario character and his kindred had been mercifully liberated.

The End of College Life

2025-03-30T08:00:00-04:00

The start of spring semester is a hopeful time on college campuses. Students fill the quads and walkways, wearing salmon shorts or strappy tank tops. Music plays; Frisbees fly. As a career academic, I have been a party to this catalog-cover scene for more than 30 years running. It looks made-up, but it is real. Every year in the United States, almost 20 million people go to college, representing every race, ethnicity, and social class. This is college in America—or it has been for a long time.

But college life as we know it may soon come to an end. Since January, the Trump administration has frozen, canceled, or substantially cut billions of dollars in federal grants to universities. Johns Hopkins has had to fire more than 2,000 workers. The University of California has frozen staff hiring across all 10 of its campuses. Many other schools have cut back on graduate admissions. And international students and faculty have been placed at such high risk of detainment, deportation, or imprisonment that Brown University advised its own to avoid any travel outside the country for the foreseeable future.

Higher education is in chaos, and professors and administrators are sounding the alarm. The targeting of Columbia University, where $400 million in federal grants and contracts have been canceled in retribution for its failure to address campus anti-Semitism and unruly protests against the war in Gaza, has inspired particular distress. Such blunt coercion, Princeton University President Christopher Eisgruber wrote in The Atlantic earlier this month, amounts to “the greatest threat to American universities since the Red Scare.” In The New York Times, the Yale English professor Meghan O’Rourke called it and related policies “an attack on the conditions that allow free thought to exist.”

Those assessments are correct, but they’re also incomplete. So are the many paeans to the social and economic benefits of university research that schools have posted in the past two months. Yes, academic freedom is at stake, along with scientific progress. But the government’s attacks also threaten something far more tangible to future college students and their parents. The entire undergraduate experience at residential four-year schools—the brochure-ready college life that you may once have experienced yourself, and to which your children may aspire—is itself at risk of ruination.

[Read: A new kind of crisis for American universities]

Few administrators have talked about this risk in public, but they take a different tone in private as they try to figure out how broken budgets can be fixed. I’ve spent the past month discussing the government’s campaign to weaken higher learning with current and former college presidents, provosts, deans, faculty, and staff. And in the course of these informal, sometimes panicked text exchanges, emails, and phone calls, I’ve come to understand that the damage to our educational system could be worse than the public comprehends—and that calamity could arrive sooner than people expect.

Any one of the Trump administration’s attacks on research universities, let alone all of them together, could upend the college experience for millions of Americans. What’s at stake is far from trivial: Forget the Frisbees on the quad; think of what it means to go to college in this country. Think of the middle-class ideal that has persisted for most of a century: earning a degree and starting a career, yes, but also moving away from home, testing limits, joining new communities, becoming an adult.

This might all be changing for fancy private schools and giant public universities alike. If you, or your son, or your daughter, are in college now, or are planning to enroll in the years ahead, you should be worried.

The familiar image of campus life was well established by the early 20th century. Whether a student was enrolled at Yale or Ohio State, they could expect to find the same mix of fraternities, coursework, and college sports; the same commingling of parties and protests.

At first that experience was limited to the wealthy. The American approach to higher ed had arisen from a melting pot of influences, some homegrown and others pulled from overseas. U.S. schools combined the practicality of land-grant institutions with the research focus of German universities. They borrowed also from British residential colleges, adopting both pastoral campuses and an aristocratic temperament. But after World War II, U.S. higher education was transformed. It quickly spread into the middle class, and helped expand it. The GI Bill, among many other efforts by the government, made college more accessible. Federal funding for research turned the college campus into a source of valuable expertise, and the locus for its own, growing set of jobs. And the rise of the knowledge economy made the skills one learned in college more desirable for all. In 1940, only 5 percent of American adults had earned a bachelor’s degree. By 1960, 45 percent of high-school graduates were enrolled in college; by 1997, that figure reached 67 percent.

All of this seems natural to Americans, but it was, and is, an unusual way to handle higher learning. Elsewhere in the world, colleges are not defined as mediators between childhood and adulthood. They may not be specific places where a student “goes” (as in the common formulation “I went to college”) but rather sets of nondescript buildings interwoven with cities. Only in America could “college” refer to the amalgamation of a coming-of-age experience and a credentialing service, based in a planned community that was mainly built to facilitate scientific research but also provides diversion, dining, and professional-quality sporting events.

As this peculiar form of college life expanded, Americans became attached to it. Many meet their future spouses while in college, then settle down not far away from where they went to school. Many cheer for their alma mater’s teams on television. Many send their alma mater money every year. College life stretches far beyond the years they spend on campus. In a way, it even stretches beyond the people who take part in it directly. It imbues the culture.

But creating this experience has always been just one of many, varied roles that schools must play. Tens of thousands of people—undergraduates, but also faculty researchers, administrative staff, and residential-life personnel—might go about their business on a given campus unaware of the purposes of those with whom they cross paths. Amid this web of private factions and concerns, the day-to-day routines of students—their “college experience”—can be a fragile thing. One part of that experience is learning: going to class, pursuing a major, studying, doing research, and completing a thesis. But for the schools, the cost of providing these activities doesn’t balance with tuition revenue. In 2024, Columbia spent about $3 billion on instructional expenses, facilities costs, and operations. It took in about $1.75 billion in tuition and fees (such as room and board). The balance is made up by combining revenue from all across the campus, including money that comes in from its investments, for example, or from taking care of patients at its hospital.

That’s why even just the Trump administration’s first strike against its targets—a mass curtailment of science-research funding—could end up being felt by students right away. At research universities, federal grant dollars may represent 15 to 25 percent of overall budgets. Even schools with huge endowments—Columbia’s is about $15 billion, for example—lack an easy way to fill the gaps, because that money may be spread across thousands of accounts, each of which may have rules for how it can be spent. Cutting grant dollars so substantially and unexpectedly cannot be addressed in a way that limits the impacts of those cuts to research alone. Something else will have to give.

[Read: America will sacrifice anything for the college experience]

If a school’s athletics program is unable to cover its own costs, it may need to scale back spending on its sports teams. I’ve also heard from colleagues at schools across the country that construction projects have been scrapped, that study-abroad programs are getting canceled, and that career services face cuts. In the meantime, faculty hiring freezes such as those adopted at Harvard, the University of Pittsburgh, and the University of Vermont could mean that fewer classes will be taught next year.

And all of that could be just the beginning. About one-third of U.S. college students rely on federal Pell grants, a financial-aid program for low-income families. That program was already facing a shortfall, and now the Trump administration is dismantling the Department of Education and its Office of Federal Student Aid. The White House claims that financial-aid programs will be unaffected, but personnel who manage those grants, other loans, and federal work-study programs now say they are unable to carry out their duties effectively. For students who rely on these funds, even short disruptions could end their college aspirations.

In the meantime, the need-blind admissions and no-loan financial-aid packages that many elite private universities have adopted over the past decade may start to seem untenable amid a budget crunch; top schools could focus on enrolling students whose families can pay top dollar, at the expense of everybody else. Public universities could respond to their duress by bringing in more, higher-paying students from out of state. (That trend is preexisting, but it could get much worse.) And those who can’t afford the rising costs would be relegated to community colleges and online degrees—useful options, but far removed from the cherished image of American higher education.

The point is this: It’s not just research; all of the trappings of college life are potentially at risk, and soon. If the campus experience is able to survive, it may revert to a more elitist, century-old version of itself, where the sons and daughters of oligarchs are among the only students on its lawns.

Those who plan to study the humanities may have the most to lose. The speed and scale of funding cuts demand immediate action, one top administrator at a state flagship university told me, and that could put programs in English, natural languages, art history, and the like into the crosshairs.

Doctoral students in such fields are mostly paid out of general university funds, which means that their salaries could easily be reclaimed and used as filler for any budgetary holes. (In contrast, money tied to research grants must be spent in certain ways.) But these same grad students also teach large numbers of undergraduates, so losing them might cause a university’s course catalog to shrink. When fewer people are on campus to teach humanities classes, fewer humanities classes will be taught on campus.

It doesn’t help that the programs the White House seems most interested in targeting are found in the humanities. At Columbia, for example, the Department of Middle Eastern, South Asian, and African Studies has been stripped of its independence in accordance with one of the Trump administration’s specific demands. Another, related problem is that the humanities faculty are already alienated from the grant-funded research mission of the postwar university. Professors of cultural studies, or history, or the arts, have long seen themselves as critics of institutions, including the universities that employ them. At Columbia and elsewhere, the attack on research funding is only worsening intrafaculty resentments that have been felt for decades.

The present crisis also carries existential risks for general education, the part of college that requires students to take courses from across the curriculum. Some humanities departments with shrinking majors have been able to persevere in part by fostering humanistic education and scholarship on campus. But the state-flagship-university administrator said that just this sort of work may start to seem expendable, at a time when STEM investments are imperiled.

[Read: The first year of AI college ends in ruin]

For the coming crop of undergraduates, cuts to the humanities, paired with the collapse of general education, would alter the foundations of their college experience. The long-standing idea that you go to college to discover your interests and abilities—to “find yourself”—would be replaced by a full embrace of professional training. Admittedly, this change has already been in the works for years, even at prestigious schools: The rising cost of college makes experimentation more difficult to justify, and students have grown more market-focused over time. Up until this point, even the most practical-minded among our undergraduates could still balance their vocational majors with breadth-giving minors, or see the benefits of merely a passing exposure to other fields and forms of inquiry. But now the university—a home for all knowledge, universally—could be forced to downsize its ambition.

“Our problem in part is a failure of imagination,” Lee Bollinger, a former Columbia president, told The Chronicle of Higher Education earlier this month. “We cannot get ourselves to see how this is going to unfold in its most frightening versions.” Now the most frightening versions are taking shape, as specters in the hallways of the ivory tower. Jobs are being cut, labs closed, building projects canceled. Faculty members may decide that their devotion to the pursuit of knowledge may not be worth the occupational risk. Those in science, medicine, and engineering whose grants have been withheld may start exiting for jobs abroad or in industry while they still can.

The college experience could very soon be one that bears little semblance to the classic picture. Your kid could end up on a campus with reduced student services and activities, aging rec centers, shrunken-down humanities departments, less prestigious faculty, and a class cohort that has been stripped of foreign students, and also thinned of anyone who happens not to be well-off. It could be a dreary and degraded version of the life at school that you may have once enjoyed yourself.

College students’ prospects would have been slightly better if the schools had seen this coming. Attacks on higher education are not new, but they have never been waged at this level and this speed. The government has at other times threatened to withhold funds from universities in order to enforce federal law, but such examples have been “rare and marginal,” David Labaree, a Stanford historian who studies higher education, told me. That may be why no university in America was prepared for this assault, and none seems ready to respond.

Neither the White House nor the Department of Education responded to a request for comment for this story. But even the most enthusiastic advocates for going after higher ed may have been surprised by the intensity of the Trump administration’s actions. Max Eden, a senior fellow at the American Enterprise Institute, suggested in an essay for the Washington Examiner in December that the new regime, and its then-presumptive education secretary, Linda McMahon, should increase the endowment tax on universities and threaten to withhold federal funds to bring them into line on matters of DEI and anti-Semitism. “To scare universities straight, McMahon should start by taking a prize scalp,” he wrote. “She should simply destroy Columbia University.”

Eden, who did not respond to a request to comment for this story, seemed to expect at least some kind of formal process for this scalping. In the bloodthirsty scenario laid out in his op-ed, McMahon would threaten Columbia before cutting off its funds. In fact, the Trump administration cut off those funds just days after McMahon had been confirmed, and before demands were even made. Then it pressed for multiple concessions before the school would have any opportunity at all to negotiate for the funding to restart. The Justice Department attorney Leo Terrell insisted that the school was still “not even close” to having its funds unfrozen; the next day, Columbia gave in.

For two months now, universities have had no idea what, if anything, might stave off further punishments. The state-flagship-university official I spoke with admitted to hoping that the problem will just go away. Perhaps one scalp will be enough, a message sent. For the moment, though, the demands themselves (whatever those might be for different schools) seem less destructive than the sudden, chaotic application of extreme financial leverage: “It’s all arson and no architecture,” the official said. Universities might be amenable to adjusting the terms of their relationship with the federal government, but they cannot do so quickly and under such duress. The Trump administration appears to want them not to talk, but to die.

As for future college students and their parents, the campus experience they expect—the one that generations of Americans took for granted—is no longer guaranteed. Here on campus, the undergraduates seem unaware of this alarming fact. The crisis for universities may be existential, but another spring is blooming on the quad. College has persisted for a whole lifetime in its present form, in a pastoral setting, underwritten by federally funded research, with football crowds cheering in the distance. Last week, a student in my course on artificial intelligence bounded into the lecture hall, full of energy and optimism. “How was your break?” she asked. I’d designed the class to give undergraduates from across the university insights into the changes AI might wreak on their future professions, but just then I found myself wondering more about the future of my own. Not because technology may disrupt it, but because my own government seems intent on destroying it. “It was good,” I said, faking a smile, unsure of what to say. “It was good.”

Grad School Is in Trouble

2025-02-27T07:00:00-05:00

Jennie Bromberg was somehow still exuberant last weekend about her future career in public health. In January, she interviewed for a competitive Ph.D. program in epidemiology at the University of Washington, one of several to which she has applied. “I loved them. It was amazing,” she told me by phone while on a walk with her Australian shepherd. But the email that arrived from UW shortly after she got home was not the acceptance letter that she’d hoped for. Nor was it even a rejection. Instead, it said that she’d been placed in grad-school purgatory. All new offers of admission were being put on hold “in response to the uncertainty we are facing because of the rapidly changing financial landscape.” The email finished: “We appreciate your patience as we navigate through these uncertainties and disruption.”

Those words euphemize a cascade of traumas that have befallen higher education since Inauguration Day. The Trump administration has frozen, slashed, threatened, and otherwise obstructed the tens of billions of dollars in funding that universities receive from the government, and then found ways around the court orders that were meant to stop or delay such efforts. In the meantime, new proposals to raise the tax on endowment income could further eat away at annual budgets. And all of this is happening at just the time when graduate admissions are in progress. Future researchers such as Jennie Bromberg are caught in the middle.

The University of Washington is not alone in putting things on hold. The University of Pennsylvania, the University of Pittsburgh, and the University of Southern California have also paused or cut their graduate admissions, at least temporarily. Ilya Levental, a biophysicist at the University of Virginia, told me that his program in biomedical sciences reduced the size of its incoming class by 30 percent. In other words, grad school is in trouble. And because grad school trains the next generation of academics—those who will be teaching students, discovering knowledge, and translating science into practice—this means the future of the university itself is in trouble too.

Doctoral students typically do not pay for their advanced degrees. Instead, they work in research groups or labs, or sometimes as classroom instructors. In exchange for this work, universities usually pay them a modest salary and waive or cover their tuition. In engineering, the sciences, and medicine, the cost of that support comes mostly from faculty research that is in turn paid for by grants received from the federal government.

[Read: A new kind of crisis for American universities]

Once it became clear, in recent weeks, that this grant money was in jeopardy, schools began gaming out contingencies. Reducing the number of graduate students they will have to pay next year is one way to lower near-term risk. It’s also an act that universities would want to take right now, before their offers of admission are sent out. “People are trying to be conservative, because the worst outcome is very bad here,” Aaron Meyer, an associate bioengineering professor at UCLA, told me. “A commitment to a Ph.D. student in the sciences is easily half a million dollars, over many years.”

Administrators’ choices on admissions are made even more complicated by a weird dynamic in play across higher ed. No one wants to overreact and cut new students without good reason, but they also have to hedge against the risk of others’ cuts. The situation is structured like a prisoner’s dilemma: If lots of programs start reducing their admissions, that means fewer total spots for applicants, which in turn could lead to greater “yields”—that is, a higher proportion of each school’s offers gets accepted. No school wants to end up with too many students, so if one expects a growing yield, it may decide to cut admissions offers on that basis—and thus exacerbate the larger trend.

The administration has also called for tightened scrutiny on visas of all kinds, including student visas. This could further muddy grad-school yields by making some applicants unable to accept their offers of admissions or enroll. Graduate-student unions, which now represent more than 150,000 students nationwide, add another layer of uncertainty. Organizing has allowed grad students, who can barely afford to live in many cities, to advocate for better pay and labor practices. But it also increased the cost of graduate education in a way that worried administrators even before the grant and overhead cuts arrived. Schools sometimes take graduate tuition, and normally pay student stipends, from the same grants that are now at risk. And some grants have already been canceled, leading to a scramble for money to cover current students. The whole system has been thrown out of whack.

Choosing to take fewer students forestalls or even ends the careers of future scientists. It also makes research harder. In most science, engineering, and medicine programs, students get accepted into specific labs or groups led by the faculty whose grants also fund those students. These faculty members take on students to help them carry out their research. “Ph.D. students make up the bulk of the academic-research workforce,” Levental told me. Without their labor, work on already awarded grants can’t be done—assuming the funds to carry out those grants continue flowing in the first place.

[Read: The chaos in higher ed is only getting started]

The situation could deteriorate if current doctoral students start jumping ship. A Ph.D. student might make $35,000 a year, a sum they tolerate because “they are investing in themselves and are dedicated to the cause,” as Levental put it. But that investment might start to look foolish. Dallas McCulloch, a doctoral student who studies health and illness at Wayne State University with four years of supposedly guaranteed funding, told me that he is thinking of quitting and moving abroad to pursue his degree, because of “the grim prospects of any future funding, including for my dissertation.” McCulloch, an American who also holds a German passport, said he is worried that if he doesn’t act soon, he’ll end up competing with a “mass exodus” of researchers seeking to leave the United States.

Universities could decide to cover shortfalls in science and engineering by reallocating funds for graduate education from elsewhere. Some faculty and administrators I spoke with are worried that the humanities might become a casualty of such reapportionment. There, graduate students are typically paid for teaching, not research. Knock-on cuts to their admissions could follow, the effects of which might then reverberate into undergraduate education. If grad school in the sciences falters, the effects will not be contained.

For the moment, though, the whole system is in limbo. UW’s “pause” on graduate admissions was set to last at least two weeks, according to the email that was sent to Bromberg two weeks ago. No news was promised either way—and no news is what Bromberg has received so far. Given the chaotic and aggressive rush of new directives from the federal government, universities have no idea whether their financial outlooks will improve or worsen in the coming months. They don’t even know when they’re likely to find out. Over the weekend, Carolyn Ibberson, a microbiologist at the University of Tennessee at Knoxville, created a shared spreadsheet to track the latest news. Its title sounds definitive, “Graduate Reductions Across Biomedical Sciences (2025),” but much of the information there is cited to private conversations and internal emails. In other words, academics face uncertainty about how universities are handling uncertainty.

Bromberg can only take things as they come. She lives in Columbus, Ohio, but plans to attend, at her own expense, Washington’s on-campus open house for prospective graduate students and is still waiting to hear back from other schools. She told me that she understands the pressures that administrators are feeling at the moment: “I just feel so bad for people who have to make these decisions.” And if Bromberg doesn’t get into a doctoral program—or if the research career she hopes the degree will unlock becomes unviable—she’ll just have to think of something else. Like McCulloch, she has wondered if she could flee to Europe. Even before the Irish journalist Fintan O’Toole urged his government to steal American scientists, Bromberg had already researched the cost of moving Gatsby, her 70-pound dog, from Columbus to Dublin: $8,000, or about one-quarter of a typical annual graduate salary. “I’ll be devastated if this is the end of everything I’ve worked for in my career,” Bromberg said. “But what am I going to do? I have to start looking into these things.”

DOGE Has ‘God Mode’ Access to Government Data

2025-02-19T15:59:19-05:00

If you have tips about the remaking of the federal government, you can contact Charlie, Ian, and Matteo on Signal at @cwarzel.92, @ibogost.47, and @matteowong.52.

DOGE has achieved “God mode.” That’s according to an employee in senior leadership at USAID, who told us that Elon Musk’s Department of Government Efficiency now has full, unrestricted access to the agency’s digital infrastructure—including total control over systems that Americans working in conflict zones rely on, the ability to see and manipulate financial systems that have historically awarded tens of billions of dollars, and perhaps much more.

The employee’s account, along with the accounts of several others across federal agencies, offers the clearest portrait yet of just how deep DOGE has burrowed into the systems of the federal government—and the sensitive information of countless Americans.

In the coming weeks, the team is expected to enter IT systems at the CDC and Federal Aviation Administration, and it already has done so at NASA, according to sources we’ve spoken with at each of those agencies. At least one DOGE ally appears to be working to open back doors into systems used throughout the federal government. Thomas Shedd, a former Tesla engineer who was recently appointed director of the Technology Transformation Services, requested privileged access to 19 different IT systems administered by teams within TTS, according to two federal workers we spoke with who are familiar with his request. With this level of control, Shedd would be able to not only view and modify federal data, but also grant and revoke access to other people. (In a written statement, Will Powell, the acting press secretary for the General Services Administration, of which TTS is a part, said Shedd needs this level of access to rapidly identify “areas for optimization and efficiencies” and insisted that he is working with “appropriate GSA officials” to follow established protocols.)

[Read: The government’s computing experts say they are terrified]

Over the past few days, we’ve talked with civil servants working for numerous agencies, all of whom requested anonymity because they fear what will happen if they lose their job—not just to themselves, but to the functioning of the federal government. Their observations reveal the abnormal degree of power that DOGE has already achieved. Federal agencies are subject to various forms of administrative and legal oversight, but they operate separately from one another for good reasons: to support a specialized purpose and to insulate them from undue outside influence. Now they effectively roll up to Elon Musk. (Neither the White House nor DOGE responded to requests for comment for this story. Earlier this week, a White House official claimed that Musk is not the head of DOGE. He is clearly the group’s functional leader.)

Among the federal agencies we reported on, USAID is the only one where we could confirm that DOGE has acquired God-mode access across the entire digital system. (The Trump administration has sought to effectively shut down USAID since the inauguration.) But as Musk and his acolytes enter a growing number of federal databases and IT systems, their unfettered access at USAID offers a sense of what they might be able to do elsewhere. At NASA, for example, it could mean access to knowledge about sensitive government technologies used for defense. At the CDC, such ability could expose millions of Americans’ health data and allow DOGE to access labs that store deadly pathogens. At Treasury, such access would allow Musk’s employees to view Americans’ names, Social Security numbers, and financial information. “It is not ridiculous to think they’d have bank-account and routing numbers for every single person in the United States,” the senior USAID source said. “What do you do with this information? I had to ask myself, Do I file my taxes this year or not? I had to sit and debate that.”

The federal government does not typically grant such wide-ranging access to a single entity, let alone one that is effectively under the control of an unelected, erratic, and politically extreme actor such as Musk. The group is working on behalf of Donald Trump, but sources we spoke with emphasized that the level of access DOGE possesses means that the organization may already be able to siphon data that Musk or his agents could hold on to forever, long after his time as a government liaison, or even after a potential falling-out with the president.

One experienced government information-security contractor offered a blunt response to the God-mode situation at USAID: “That sounds like our worst fears come true.” The purpose of DOGE’s incursions remains unclear to employees at these agencies. Musk was supposed to help improve the workings of the government—that is DOGE’s stated purpose. But in the offices where the team is reaching internal IT systems, some are beginning to worry that he might prefer to destroy it, to take it over, or just to loot its vaults for himself.

“Once they’re in, they’re in,” the USAID employee told us. And this is a big part of the problem in a nutshell: Access is everything, and in many cases, DOGE has it.

At USAID and other agencies whose employees we spoke with, leaders explained that Musk’s team could copy and remove information from government servers without anybody knowing. The team could then feed this classified information into AI tools, either for training purposes or to mine the data for insights. (Members of DOGE already reportedly have put sensitive data from the Education Department into AI software.) Within USAID, DOGE has full access to human-resources information—Social Security numbers, addresses, reputational data such as performance reviews, plus classified information and disciplinary information. The USAID source noted that DOGE can also control USAID systems that help with disbursement of funds, building-access tools, and payroll: “If they wanted to change how much a person is making, they could modify that, given their access in the system.” According to the employee, DOGE is also inside of an internal system for managing contracts and grants, which functions like a high-security online marketplace where USAID plans and approves billions in government spending.

Inside NASA, according to one agency employee we spoke with, DOGE workers already have access to contracts, partnerships, performance reviews, classified national-security information, and satellite data, among other materials. The NASA worker told us that such knowledge could erase generations of advantage in aerospace and defense capabilities if it falls into the wrong hands. Agency technologies such as propulsion systems, novel materials, and satellites overlap with Department of Defense projects. Someone with information about NASA’s thermal-protection or encryption technologies could take advantage of vulnerabilities in aerospace vehicles, for example.

[Derek Thompson: DOGE’s reign of ineptitude ]

USAID employees have felt more acute effects of DOGE’s operations. Employees there say they have been rattled by the demands of DOGE engineers: “They have walked in and said to senior staff, You have 15 minutes to do this or you’re fired,” the USAID senior leader told us. Now USAID staff are “operating in a zero-trust environment.” With its God-mode IT control at the agency, DOGE can read emails and chats, plus see who’s attending which meetings. The source described employees in a recent meeting growing alarmed when transcription services seemed to turn on without warning. An employee at NASA reported similar concerns, after unfamiliar messages appeared on workstations. “We’re operating believing our systems are completely bugged,” one person told us.

The senior USAID official fears that DOGE could terminate somebody working in a conflict zone like Ukraine, Sudan, or Ethiopia from an agency system. “If they lose access to their USAID laptop, phone, and accounts, for a lot of them that’s their only means of communication. We are putting their lives on the line,” one said.

For those who have watched DOGE storm into their workplace, what is perhaps most terrifying is its attempts to scale. If DOGE were to acquire God-mode administrative access across many systems, several sources told us, that level of control could affect every citizen at home, and many American interests abroad: personal financial data, defense secrets, and more, all in the palm of Musk’s hand.

There’s reason to believe that health information may be next. The Trump administration fired roughly 700 people at the CDC last weekend. As in other agencies, the firings will hollow out expertise but also remove obstacles to further changes. A CDC employee told us that the agency’s Office of the Chief Information Officer is expecting DOGE, but “no one has seen anyone yet.”

The individual, who has knowledge of how CDC information systems work, fears that DOGE could gain access to an abundant store of sensitive information about health and disease. This year, the CDC is supposed to roll out a central data platform for public-health surveillance and emergency response to better address new threats such as H5N1 bird flu and old ones such as measles. The new system, called the One CDC Data Platform, promises to aggregate all of the CDC’s public-health data, including hundreds of thousands of daily anonymized lab tests, data from emergency-room visits, and measurements from wastewater disease-reporting sites.

The design and rollout of this system were already controversial inside the agency, our source said, even before Trump and Musk came on the scene. Putting everybody’s health data in one place carries risks. Although the health data the CDC houses are usually de-identified or aggregated, “people with very stigmatizing illnesses could be identified by certain characteristics” if the data are exposed or misused, the CDC worker said. What’s more, plenty of health data contain information that, when correlated with other data outside the system, could pinpoint specific individuals. Given all of the data that DOGE appears to be capable of siphoning from all over the government, such identification could become much easier. The CDC collects electronic health-record details from all over the country, meaning that this could affect just about everyone—including us, and you too.

CDC systems control more than mere information about disease. At the agency’s facility in Atlanta, the CDC stores the microbes that cause disease and can hold secrets to treatment. Some are relatively benign, such as strains of E. coli. Others are intrinsically dangerous, including the Ebola and Marburg viruses, and the bacterium that causes tuberculosis. These materials are housed in labs with associated biosafety levels. The highest level, BSL-4, applies to only a small number of labs around the world containing “dangerous and exotic” microbes, as the CDC describes them, that pose a high risk of spread.

Access to such labs is managed by computers, and management of those computers is local to the CDC. If DOGE got the same kind of access to CDC IT systems as it has elsewhere, would that give the group direct access to CDC facilities? “Yes, those are all out of CDC level,” our source at the agency told us. Does that mean that DOGE could gain direct access to BSL-4 labs? we asked. “It’s definitely possible,” the employee said.

[Read: If DOGE goes nuclear]

Our source hoped that such a prospect would be averted. (To repeat, DOGE hasn’t taken control of CDC IT systems yet, at least as far as our source knew.) But the employee also explained that the recent layoffs will reduce the agency’s ability to defend against IT or security errors, on top of diminishing morale. The CDC did not respond to requests for comment about whether someone with full, local IT control could indeed grant entry to, or control over, BSL-4 labs and their contents.

The risk of harm, abuse, or political revenge is clear. But simple, brazen corruption is also a concern among the federal workers we spoke with. The CDC staffer wondered if DOGE’s unelected and seemingly unaccountable leadership, including Musk, might simply want to sell the public-health data the CDC collects. Democratic leaders have also expressed the worry that Musk’s interest in SpaceX, which has received billions of dollars in contracts from NASA over the years, creates an untenable conflict of interest. The NASA employee worried that Musk would end up “reaping all of the profits of the investment that the American public put into NASA’s research, which was being shared with the country.” NASA holds technical specs and research data for SpaceX competitors, and insiders fear that such information will soon be compromised, too. They also worry that classified NASA R&D in areas such as quantum, biotech, and astrobiology could be stolen for private gain.

A number of lawsuits have been filed seeking to limit DOGE’s access, with mixed results. Meanwhile, Trump and Musk have both attacked judges who have ruled against their interests; Musk has said they should be impeached. Trump has also indicated that he might just ignore the courts—an act that would be challenging to counter, providing plenty of opportunity for the administration to get its way. Across agencies, leaders have started to step aside voluntarily. Jim Jones, head of the Food and Drug Administration’s food division, resigned this week. Michelle King, who ran the Social Security Administration, stepped down too rather than carry out DOGE’s wishes. The resignations may be principled, but they open the door for more compliant replacements.

The request from Shedd, the former Tesla engineer, in particular, illustrates the variety of avenues and back doors that DOGE and Musk’s allies may have available to enter the government’s systems. According to the two federal workers we spoke with about Shedd’s efforts, such access typically is not granted to TTS leadership and requires a specific reason and the permission of each system’s owner. Shedd initially issued a blanket request, the sources told us, and is now attempting to bypass the individual system owners by seeking permission from other officials, circumventing standard security procedures. He also had not completed a background check, which is usually required for such access, at least as of when he first made the request, according to our sources. How much access Shedd has been granted remains uncertain.

This is the DOGE playbook: There are no norms to be respected, and everything is up for grabs. Once the damage is done, it will be difficult to remedy, especially if DOGE staffers can themselves grant or remove access to others at their discretion.

Musk and DOGE’s first month has been so chaotic, their incursions so haphazard, that assessing what has even happened is difficult. DOGE claims to be improving the government, but the agency workers we spoke with feel that they are being hacked instead. So it is worth stepping back to note the most basic fact: No good reason or case can be made for one person or entity to have this scope of access to this many government agencies containing this much sensitive information. Even in one government office, full administrative access to all systems is the rarest privilege. In the aggregate, across the whole of the government, it would be unfathomable.