Inside Elon Musk’s ‘Digital Coup’

As America’s most decorated civil servants sipped cocktails in the presidential ballroom of the Capital Hilton, worrying about their table assignments and wondering where they fell in the pecking order between US senator and UAE ambassador, Elon Musk sat staring at his phone, laughing.

Few of the guests at the Alfalfa Club banquet in Washington, DC, on January 25 knew what he knew: that a crew of senior executives and young Musk loyalists was preparing to occupy the top offices of a nearby federal building. Under guard, they would sleep on mattresses lined with body temperature and breath rate sensors as they raced to refactor the nation’s code base—or, better yet, scrap it altogether.

Musk wasn’t big on formalities, but he’d dressed up for the occasion. The Alfalfa Club had been around since 1913 and existed solely to host a yearly banquet where the most important people in government could hobnob with the most important people in business. Membership was limited to around 200, and the Alfalfas admitted new “sprouts” only when existing members died. That evening, Supreme Court justice Elena Kagan and Nasdaq CEO Adena Friedman joined the likes of Microsoft CEO Satya Nadella and US senator Kirsten Gillibrand as members. Musk was attending as a guest.

The social chairs of the Alfalfa Club seemed to think that elections and constitutional norms should determine the seating chart in American political life. The head table was reserved for Alfalfas in government. Musk, the assumed leader of the so-called Department of Government Efficiency, sat on the opposite side of the room. He spent much of the dinner on his phone—talking to the president, if whispers were to be believed. Musk was closer than ever to Donald Trump. He told friends he was crashing in government buildings. He would soon move in next door to the White House, staying in the Eisenhower Executive Office Building’s Secretary of War Suite. He’d even had his video-gaming rig installed there.

As Musk sat in the Hilton ballroom, his operatives, working under a trusted lieutenant, had already gained access to systems at the Office of Personnel Management, the federal HR department for 2.2 million or so career civil servants. Many of these operatives would show up later at agencies across the federal government—people like Akash Bobba, a UC Berkeley graduate and former intern at Palantir, the defense contractor cofounded by Peter Thiel; Edward Coristine, a 19-year-old who has gone by the online nickname “Big Balls”; and Nikhil Rajpal, an engineer in his thirties who had worked at Twitter during Musk’s acquisition, where he’d once pitched the idea of auctioning off dormant usernames to the highest bidder. As an undergraduate, also at UC Berkeley, Rajpal had been president of a libertarian student group that was fond of the motto “Futuate cohortem urbanam”—Latin for something like “Fuck these city dwellers.”

In Musk’s mind, Washington needed to be debugged, hard-forked, sunset. His strike teams of young engineers would burrow into the government’s byzantine bureaucratic systems and delete what they saw fit. They’d help Trump slash the budget to the bone. Musk turned to those around the table at the Hilton: Can you believe we were spending taxpayer money on condoms? They shook their heads. Musk looked back at his phone. Then: What if we cut all federal grants to NGOs?

In the days and weeks that followed, DOGE hit one part of the federal government after another. The Departments of Agriculture, Defense, Education, Health and Human Services, Homeland Security, Housing and Urban Development, Justice, and Veterans Affairs; the Federal Aviation, General Services, Social Security, and National Oceanic and Atmospheric administrations; the Consumer Financial Protection Bureau and the Internal Revenue Service; the US Agency for International Development and the Centers for Disease Control and Prevention; the National Park Service and the National Science Foundation—all fell under Musk’s control. An estimated tens of thousands of federal employees were effectively fired or resigned. “This is a digital coup,” one USAID source told WIRED at the time.

Along the way, DOGE also gained access to untold terabytes of data. Trump had given Musk and his operatives carte blanche to tap any unclassified system they pleased. One of their first stops: a database previously breached more than a decade ago by alleged Chinese cyberspies that contained investigative files on tens of millions of US government employees. Other storehouses thrown open to DOGE may have included federal workers’ tax records, biometric data, and private medical histories, such as treatment for drug and alcohol abuse; the cryptographic keys for restricted areas at federal facilities across the country; the personal testimonies of low-income-housing recipients; and granular detail on the locations of particularly vulnerable children.

What did DOGE want with this kind of information? None of it seemed relevant to Musk’s stated aim of identifying waste and fraud, multiple government finance, IT, and security specialists told WIRED. But in treating the US government itself as a giant dataset, the experts said, DOGE could help the Trump administration accomplish another goal: to gather much of what the government knows about a given individual, whether a civil servant or an undocumented immigrant, in one easily searchable place.

WIRED spoke with more than 150 current and former federal employees, experts, and Musk supporters across more than 20 agencies to expose the inner workings of DOGE. Many of these sources requested anonymity to speak candidly about what DOGE has done—and what it might do next.

Musk and Trump’s relationship was cemented on July 13, 2024, when a would-be assassin came within inches of killing the former president in Butler, Pennsylvania. Musk was impressed by the photo of Trump, blood streaming down his face, raising his fist in the air and shouting “Fight, fight, fight” for the cameras. The image quickly became a meme—Musk’s love language. He endorsed Trump that day and pivoted his recently launched super PAC to get the former president reelected.

The following month, during a live discussion on X, Musk floated the idea of working for Trump on a “government efficiency commission.” Trump’s response was enthusiastic. “You’re the greatest cutter,” he said admiringly.

Two years earlier, after Musk purchased Twitter in a chaotic blitz of last-minute paperwork and hundred-million-dollar money transfers, he had cut roughly 80 percent of the company’s staff, closed at least a dozen international offices, and rolled back Twitter’s content moderation policies in the name of free speech. He demanded change at such speed that one of his lieutenants, Steve Davis, took to sleeping at Twitter’s San Francisco headquarters with his partner and their newborn baby.

In Washington, Musk estimated that his team could cut “nearly $2 trillion” from the federal budget. After you set aside politically sacrosanct military spending, nondiscretionary spending such as Medicare and Social Security benefits, and interest payments on the national debt, that number, $2 trillion, was a little more than what you had left over. In other words, Musk was functionally proposing to cut everything else, from foreign aid to housing subsidies, from the maintenance of national parks to the collection of basic weather data, from investigations into predatory lenders to the operation of air traffic control systems.

After Trump won, he announced that Musk, along with former Republican presidential candidate Vivek Ramaswamy, would colead DOGE. The announcement kicked off a stealthy recruiting process, led by Davis, the same executive who’d slept at Twitter HQ. Musk pictured a team of super-high-IQ individuals joining him in Washington for an 80-hour-a-week, 18-month hackathon on the US government.

The DOGE brain trust camped out on the eighth floor of the SpaceX office in Washington, DC, commandeering multiple conference rooms and conducting meetings and interviews with DOGE hopefuls, according to a person with knowledge of the events. One question for applicants: Who did you vote for in 2024?

Among Davis’ early recruits was Zsombor (Anthony) Jancso, a San Francisco–based engineer and former Palantir employee in his mid-twenties. After Palantir, Jancso had worked on a project called Accelerate X, which purported to offer “a modern OS for government” with solutions “delivered in days.” His cofounder, an MIT-educated engineer named Jordan Wick, joined DOGE too.

A few weeks after the 2024 election, an online handle associated with Jancso reached out to a group of people who had participated in an AI challenge put on by the US Space Force. The person said they were looking for “hardcore engineers” and instructed applicants to send their GitHub or LinkedIn to @DOGE on X and reply privately with their X handle. (To do all this, they’d need to pay for X premium.) Not long after that, the same handle posted in a group for Palantir alums: “This is a historic opportunity to build an efficient government, and to cut the federal budget by 1/3.”

Luke Farritor, a 23-year-old engineer, quickly joined in the DOGE recruitment effort. The son of an academic at the University of Nebraska-Lincoln and a physician, Farritor was homeschooled and filled his childhood bedroom with books by James Baldwin and Jordan Peterson. During college he got prestigious internships at SpaceX and took to wearing his SpaceX T-shirt all the time. Later, Farritor came to a certain nerdy prominence for his role in using machine learning to decipher an ancient papyrus charred by the same volcanic eruption that buried Pompeii. He said his work received support from a $2 million gift from the Musk Foundation, and Musk backed it on X. The papyrus project also helped earn Farritor a Thiel Fellowship last spring—a $100,000 grant funded by the oligarch and meant to encourage bright young people to drop out of college, which Farritor promptly did.

On December 5, Farritor posted in a Discord group for SpaceX interns, noting that DOGE was looking for “skilled software engineers (and ops people) at any career stage who are willing to work for ~6mo in person in DC. Paid.” He added: “We’re going to fix the government!”

Musk, meanwhile, was spending time at Mar-a-Lago and getting a crash course on American civics, as taught by an array of Washington bureaucrats, venture capitalists, and right-wing shitposters on X. One of Musk’s advisers was Antonio Gracias, a private equity investor and early Tesla backer, who later summarized what they’d learned on a podcast: “A department just basically asks for money from Treasury and they send it out.”

Of course, the truth was a lot more complicated. Before the Treasury cuts a check, the payment authorization has to pass through an array of bespoke technical systems built up over decades to ensure that the money Congress appropriates is properly spent by the executive branch. The system is inefficient by design, so as to provide backstop after backstop. To Musk, that meant it was ripe for disruption. (His adviser Gracias would go on to become a DOGE “IT specialist” at the Social Security Administration.)

Still, Musk seemed to grasp that he’d been cavalier in setting his goal at $2 trillion. Government spending factors into gross domestic product; that level of cutting, the economist Dean Baker told WIRED, would represent a hit to the economy akin to the 2008 financial crisis, with the potential to produce double-digit unemployment. By mid-January, Musk suggested that $1 trillion—a sum merely greater than the entirety of nondefense discretionary spending—was a reasonable aim. He started to home in on individual government agencies. He seemed especially eager to eviscerate USAID, the government’s primary instrument for foreign aid. The right-wing conspiracy theories he’d eventually repost on X portrayed it as a criminal enterprise, a tool of the deep state’s woke agenda.

As Trump’s second inauguration drew near, Musk’s and Ramaswamy’s visions for DOGE started to sharply diverge. Ramaswamy advocated an incremental approach: Push for changes in law that would eventually cut off spending at the source. Musk didn’t want to wait around for that. Increasingly, his plans seemed in line with a scheme that incoming vice president JD Vance had laid out in a podcast interview in 2021: Fire all midlevel civil servants and replace them with Trump loyalists. Just before Trump officially took office, Musk’s vision won out, and Ramaswamy left the organization.

Trump officially established DOGE on the afternoon of his inauguration. He effectively pasted it on top of the US Digital Service, an Obama-era agency set up to attract private-sector talents into a few years of civil work. Now the D in USDS would stand for “DOGE.” The order also established a another organization within it—the US DOGE Service Temporary Organization—that would expire on July 4, 2026. This would give DOGE the ability to bring in special government employees, people who would serve for a limited time before returning to the private sector and who, critically, would not be subject to the same transparency requirements as regular government workers.

As Trump’s presidency began, the DOGE brain trust broke camp at SpaceX headquarters in DC. An employee who was there says they left a men’s bathroom “trashed.” One of the urinals was “filled with gum and Zyn.”

Daisy Kid Henderson first heard from DOGE around 9:30 pm on Inauguration Day, when she received an email inviting her to a meeting with anonymous staffers the following afternoon. Henderson, a fiery 28-year-old software engineer at the USDS, describes herself as an “eternal optimist.” Still, she found the lack of names on the invite disconcerting.

Henderson, who is based in Denver, had been working for the USDS since January 2024. She landed there after a seven-year stint at Comcast, where her job involved everything from quantum computing to robotics to an audio algorithm for filtering sounds that people with PTSD or autism might find triggering. For Henderson, USDS was a chance to take a breath and reevaluate her career while doing some good. “It was my dream job,” she recalls. “When I was hired on, it was even better than I imagined.” Henderson had overseen three generative AI pilot programs at the Department of Homeland Security. One, created for the Federal Emergency Management Agency, helped local officials plan and apply for disaster aid funding. She also partnered with the Food and Nutrition Service to help a number of states more quickly deliver assistance to low-income families.

By the time Henderson’s appointment with DOGE rolled around, she recalls, “they had been doing these meetings all day nonstop.” DOGE “had four or five interviewers and worked in shifts,” Henderson says, likely in an attempt to churn through all 200 legacy USDS employees as quickly as possible. According to another USDS staffer, someone had tried asking the interviewers for their last names, and the reply had been: “This is a one-way conversation. We won’t be answering any questions.”

One question the DOGE people seemed to be asking everyone: What do you think about DOGE?

Henderson tried to stay positive as her call with the two young DOGE operatives began. They introduced themselves by their first names only, Cole and Ram. (That was Cole Killian, a 24-year-old who had attended McGill University, and Nikhil Rajpal, the former college libertarian and X employee.) As the men asked Henderson about her projects, they seemed particularly interested in her AI-related work for FEMA and her technical abilities. Then the questions took a strange turn—centering around “who the underperformers are at USDS,” Henderson recalls, and what skills she brought to her role. At one point, Killian simply got up and walked out of the room without a word. Rajpal carried on as though nothing had happened. Afterward, Henderson learned her colleagues had been asked whether they should keep their jobs.

Similarly strange meetings were going on at the General Services Administration, another early DOGE target. If OPM was the federal government’s HR department, GSA was the operations and IT departments rolled into one. It oversaw more than 1,000 federally owned buildings, from anonymous office parks to minimalist masterpieces, along with hundreds of thousands of government vehicles and tens of billions in annual government purchasing.

As would happen at agencies across the government, the GSA seizure took place in the shadows—a matter not of announcements but of calendar invitations from unknown people, of unfamiliar names appearing in internal directories. The sixth and seventh floors, which had offices and suites used by the administrator, the “A-suite,” were restricted and largely locked down. No longer could employees simply badge in through the turnstile. Now they had to pass through metal detectors and have their belongings x-rayed.

During that first week, GSA employees caught a glimpse of a whiteboard sitting in a large, vacant room with three items written on it:

Spending Cuts $585 m

Regulations Removed 15

Square feet sold/terminated 203,000 sf.

No one in the rank and file seemed to know who wrote it or what it meant.

As one of Musk’s top operatives at the GSA, DOGE installed Nicole Hollander—Steve Davis’ partner, the other parent of the newborn who slept at Twitter HQ. A former Tesla software engineer, Thomas Shedd, became the director of the GSA’s Technology Transformation Services, which operates dozens of crucial systems used across government agencies, including Login.gov, Cloud.gov, and the Federal Procurement Data System, a database that makes all unclassified government contracts above the micropayment level freely available to the public. DOGE’s strike force at GSA included Coristine, the young engineer known as “Big Balls,” and Farritor, the papyrus whiz kid. Another young DOGE recruit, Ethan Shaotran, also got a GSA email account and A-suite clearance. Shaotran had recently served as the president of Harvard’s mountaineering club. He’d drifted into DOGE’s orbit at a hackathon for xAI, another company Musk owns.

Soon it became clear that DOGE wanted GSA to adopt one product in particular: an AI chatbot that could plug into the agency’s main portal, the Enterprise Data Solution. Such a tool would allow a handful of DOGE technicians to ask questions in plain language and get answers from vast stores of government data. (How this would accord with the GSA’s Internal Data Sharing Policy, which mandates that requests for certain kinds of controlled unclassified information must be approved by supervisors, was unclear.) To DOGE operatives unfamiliar with GSA’s systems, this might have seemed like a quick build—particularly if the team used an off-the-shelf large language model, like Claude or Gemini or Llama, as a starting point.

But the engineers at GSA knew the project DOGE had in mind was far more complex than it seemed. The Enterprise Data Solution is a maze of disparate databases, analytical tools, and machine-learning systems, all with tightly controlled permissions. Creating even a quick chatbot that could tap into these datasets and produce useful answers was anything but trivial. During the Biden administration, employees at TTS had started exploring the possibility of building a simpler chatbot called GSAi, which they hoped would increase productivity by helping people write emails and eventually process contract and procurement data. By the end of Biden’s term, though, there was no GSA chatbot on the horizon.

“Anyone can build a chatbot today; it’s really not that interesting,” a data scientist remarked during a February meeting about GSAi. A version of it—one that wasn’t directly connected to EDS—was set to go live soon. “The interesting part is in the quality. Can we build a high-quality chatbot, one where our domain expertise is being applied?”

To bridge the gap, GSA engineers proposed building what they called a discovery layer, an intermediary designed to decode user queries, identify relevant data sources, and generate precise searches that returned data the AI could interpret. The proposal, pitched to the A-suite—the select group that sources say includes DOGE members like Davis and Hollander—would also give GSA the ability to audit queries and check the quality of the responses. But for that to work, every database would need to be mapped, its columns and metadata described and categorized, ensuring the system understood what data lived where. None of this would happen automatically. It would be a manual, painstaking process.

As the GSA engineers discussed the scope of what needed to happen, according to sources familiar with the events, they seemed deflated. DOGE’s timeline was unrealistic. “This is a multiyear play,” one employee said bluntly in a meeting about the project, “and they think in terms of days and weeks.”

A day after the Alfalfa Club dinner, rumors were swirling that the Department of Housing and Urban Development would soon have its funding frozen. Grantees who already had their projects approved—state and county governments, nonprofits—began aggressively drawing down their funding at an unprecedented scale. While HUD employees could have stopped the withdrawals, they didn’t. And so, in the space of just a few days, around 1,400 grantees withdrew $1.5 billion in federal funding—five times the normal rate, according to data compiled from the agency’s banking system and shared with WIRED.

For Musk to gain what he seemingly most wanted—a “delete” button he could wield against any agency by cutting off its funding at the source—he would need direct access to the US Treasury. DOGE dispatched operatives to the Treasury’s Bureau of Fiscal Service, which controls more than $5 trillion in disbursements, including Social Security and Medicare payments, tax refunds, and the salaries of federal workers. The operatives wanted access to two key systems: the Payment Automation Manager and the Secure Payment System. David Lebryk, the highest-ranking career official at the Treasury, retired rather than comply. The financial journalist and payments expert Nathan Tankus would later say that the news of Lebryk’s retirement gave him a “panic attack,” because everyone above Lebryk was a political appointee.

Meanwhile, Farritor and a DOGE colleague had started showing up at USAID, one of Musk’s political bêtes noires. They carried backpacks with six or seven laptops each, sources familiar with the events told WIRED, and were allegedly instructed to access “employee email accounts and all digital infrastructure,” according to a lawsuit later filed against DOGE on behalf of USAID employees. Initially turned away for lack of security clearance, they came back with a handwritten note on stationery from the Executive Office of the President saying they were suitable.

On X, Musk called USAID a “criminal organization” and said it was “time for it to die”; Trump alleged that the agency was “run by a bunch of radical lunatics.” According to The Washington Post, Farritor and the other DOGE operative would run manually through payments, clicking off lifesaving programs. On February 3, Musk bragged that he had spent the weekend “feeding USAID into the wood chipper.”

Soon afterward, as WIRED first reported on February 4, a 25-year-old former X engineer named Marko Elez was granted the ability not only to read the code in the Treasury systems but also to write—or change—it. With that level of access, he (or anyone he reported to) could potentially have cut off congressionally authorized payments, effectively allowing Trump or Musk to exercise a line-item veto. More immediately ominous to people familiar with the systems was the possibility that, by tampering with the code, Elez could cause the systems, in whole or in part, to simply stop working. “It’s like knowing you have hackers on your network, but nobody lets you do anything about it,” a Treasury employee told WIRED.

The fact of Elez’s read/write access to Treasury payment systems, confirmed later by Tankus, became a source of contention. Scott Bessent, the treasury secretary, denied that DOGE had read/write access. In a letter to Congress that same day, Bessent wrote that members of the Treasury staff, including DOGE operative and Cloud Software Group CEO Thomas Krause, would have “read-only access,” which they needed to “continue this operational efficiency assessment.” But the letter didn’t mention Elez, whom a Treasury employee described to WIRED as the “hands-on-keyboard” person. (White House officials have gone back and forth about DOGE’s access to Treasury payment systems. This issue ended up in the courts. The cases are ongoing.)

By early February, Elez came under fire after The Wall Street Journal found racist comments from accounts linked to him on social media. One suggested that “99%” of immigrant workers from India would be “replaced by slightly smarter” large language models. When the Journal asked whether Elez was connected to the account, he resigned. Later, with public support from Musk and Vice President Vance, DOGE rehired him. Before long, he was designated as an “IT specialist” at another target of Musk’s: the Social Security Administration.

DOGE had installed a handpicked chief information officer at the SSA—the former CTO of a payments company headed by Jared Isaacman, a billionaire who once commanded two trips to space using SpaceX rockets and is Trump’s current nominee to lead NASA. That new CIO, Michael Russo, had asked to bring on Akash Bobba, the former Palantir intern who had been working out of OPM, as an engineer.

But there were “challenges” with Bobba’s background check, Tiffany Flick, the acting chief of staff to the acting administrator, later stated in a sworn affidavit given in a suit against the SSA. Bobba wasn’t brought on immediately. By February 10, seven days after he was requested to be onboarded, phone calls and emails started to come in—from Russo, Steve Davis, and others—making clear that Bobba was to be given access to SSA systems and data by the end of the day. As Russo and Davis “grew increasingly impatient” that evening, Flick recalled, Bobba was sworn in over the phone at 9 pm.

Initially, Flick and officials from the CIO’s office determined that Bobba would be given anonymized, read-only access to records in the Numerical Identification System, which contains information on everyone who has ever applied for a Social Security number. On February 15, Bobba reported that there were issues with the dataset he’d been provided. Russo demanded that Bobba be given full access to “everything, including source code,” Flick recalled. This included the SSA’s Enterprise Data Warehouse, which contains the “names of spouses and dependents, work history, financial and banking information, immigration or citizenship status, and marital status,” according to Flick’s affidavit.

Later that day, the chief information officer for the whole federal government—a political appointee working out of the Office of Management and Budget—issued an opinion to Russo granting Bobba the access. Flick retired. In her affidavit, she expressed serious concerns about the potential for SSA records to be “inadvertently transferred to bad actors” and about “incredibly complex web of systems” being “broken by inadvertent user error.”

Over at the USDS, Daisy Kid Henderson had decided it was time for her to leave too. Since DOGE’s takeover, Henderson and her colleagues had had little contact with their new overlords. Each day, the legacy employees logged on to their government laptops to work on projects from the last administration, like generative AI for the IRS. One USDS worker said at the time that they did their best to “ignore the clown show.”

Then Henderson received another email from DOGE: They wanted to speak with her again the following morning. She was told they liked her enthusiasm and wanted to draw her closer into the fray.

Henderson did not return their admiration. From what she had seen of DOGE, she says, they showed a “blatant lack of regard for the American people’s private data.” Henderson wanted no part of it. Instead of accepting the meeting, she forced-shut down her laptop and phone and didn’t acknowledge the email until Monday, when she resigned. “I didn’t want to be seen as being locked in with what DOGE was doing,” she says. “On the flip side, if I was to stand up to them, or to say no, Elon Musk has shown that he is totally OK blasting and doxing employees and having that army descend upon me.” She responded to the email saying she had found a new opportunity in the private sector. She hadn’t, but with her skills, it wouldn’t be long before she did.

“I was put in a position where, whether I chose to engage or not engage, I might have to cross ethical lines. I would have to breach the oath to the Constitution that I swore. I would have to breach my morals,” she says. “It was clear I was going to be asked to join in on one of their tirades and work on a complete dismantlement of government systems.”

Later that month, DOGE imposed a $1 spending limit on federal employee credit cards. The move instantly roiled agencies from the National Park Service to the National Institutes of Health as employees scrambled to buy basic necessities to do their jobs. Yet again, DOGE’s move-fast ethos would put Americans’ personal data at risk.

At one SSA office, a manager confirmed to his staff that he could no longer pay the company that shreds sensitive documents. “We print a lot of shit daily,” one SSA employee told WIRED. “Stuff with people’s names, addresses, phone numbers, SSNs, bank accounts, you name it. We have giant locking trash bins we put it in, and we pay a shredding company to empty them every month.” With the new $1 limit, the employee said, a “stockpile” of sensitive data was growing, leaving workers with two options: “Shred it ourselves on regular office shredders—of which we have two, I think, and so it would take forever—or just sit on it while it piles up.”

Other employees at SSA told WIRED that rather than showing their anger at work where they could face retribution, they instead attended weekend protests at Tesla dealerships in DC, part of a nationwide protest effort to drive down the share price of Musk’s electric car company.

On March 7, DOGE got one of the things it seemed to want most from GSA: a chatbot that could automate work previously done by federal employees. The tool rolled out to some 1,500 employees at GSA, with an agencywide launch planned a week later. An internal memo about the tool touted the “endless” tasks it could help with: “draft emails, create talking points, summarize text, write code.” The memo hinted at the dangers of deploying chatbots at the federal level, warning workers not to “type or paste” internal or personally identifiable information as inputs.

People who used it weren’t impressed. “It’s about as good as an intern,” one GSA employee told WIRED. “Generic and guessable answers.” This version of GSAi almost certainly couldn’t interact with the EDS discovery layer first proposed by engineers. More likely it was just the first step in an iterative approach. As one official said in the February meeting about the project that the first goal could be to “deliver this sort of janky, doesn’t-work-all-the-time chatbot” to pave the way for a “turbo-charged” version down the line.

It’s unclear whether DOGE was on the same page.

Around that time, GSA employees learned there were cuts on the way. “I encourage each of you to consider your options as we move forward,” wrote Stephen Ehikian, GSA’s acting administrator, whose wife had recently worked for Musk at X. “The new GSA will be slimmer, more efficient, and laser-focused on efficiency and high-value outcomes.”

Similar messages were going out across DC. The federal government’s funding was due to lapse on March 14, potentially triggering a shutdown. While Trump was busy lobbying House Republicans to pass a continuing resolution to keep the government afloat until September, WIRED reported that Musk had expressed interest in a shutdown—in part because doing so would potentially make it easier to cut hundreds of thousands of federal workers.

But hours before the House convened to vote on the resolution, Musk and Trump had apparently put their differences aside. They convened on the White House’s South Lawn to admire a small fleet of Teslas. Trump had recently posted that he planned to buy one. Of the protesters at Tesla factories who were speaking out against Musk’s work with DOGE, Trump said, “We’re going to catch them” and that they would “go through hell.”

The president and his richest ally walked around the cars, admiring the different colors, before getting into one. “Everything’s computer!” Trump exclaimed, after sitting down. “That’s beautiful.”

It was Musk’s car, but Trump occupied the driver’s seat.

Additional reporting by Paresh Dave and Matt Giles

Let us know what you think about this article. Submit a letter to the editor at mail@wired.com.