We’re standing on the edge of a cliff on a remote Texas ranch, a long patch of rocky desert stretching out below to the verdant banks of the Rio Grande, a silver ribbon 2 miles distant. On the horizon, a light haze shrouds the mountains of northern Mexico. The whistle of a stiff and constant wind cuts through a silence that gives no hint of the hostilities, both physical and political, that animate these borderlands.

Palmer Luckey—yes, that Palmer Luckey, the 25-year-old entrepreneur who founded the virtual reality company Oculus, sold it to Facebook, and then left Facebook in a haze of political controversy—hands me a Samsung Gear VR headset. Slipping it over my eyes, I am instantly immersed in a digital world that simulates the exact view I had just been enjoying in real life. In the virtual valley below is a glowing green square with text that reads PERSON 98%. Luckey directs me to tilt my head downward, toward the box, and suddenly an image pops up over the VR rendering. A human is making his way through the rugged sagebrush, a scene captured by cameras on a tower behind me. To his right I see another green box, this one labeled ANIMAL 86%. Zooming in on it brings up a photo of a calf, grazing a bit outside its usual range.

The system I’m trying out is Luckey’s solution to how the US should detect unauthorized border crossings. It merges VR with surveillance tools to create a digital wall that is not a barrier so much as a web of all-seeing eyes, with intelligence to know what it sees. Luckey’s company, Anduril Industries, is pitching its technology to the Department of Homeland Security as a complement to—or substitute for—much of President Trump’s promised physical wall along the border with Mexico.

Anduril is barely a year old, and the trespassing I’d witnessed was part of an informal test on a rancher’s private land. The company has installed three portable, 32-foot towers packed with radar, communications antennae, and a laser-­enhanced camera—the first implementation of a system Anduril is calling Lattice. It can detect and identify motion within about a 2-mile radius. The person I saw in my headset was an Anduril technician dispatched to the valley via ATV to demonstrate how the system works; he was about a mile away.

As Luckey and his team see it, Lattice will become not just a system for securing the border but a general platform for geographic near-omniscience. With the aid of artificial intelligence, it aims to synthesize data from potentially thousands of sensors and local databases, displaying the most relevant data in phone apps, on laptop screens, and in mixed-reality headsets. Anduril’s goal is to become a major tech startup that builds hardware and software specifically for the defense industry, a venture-capital-infused outsider challenging the likes of Lockheed Martin and Northrop Grumman with their multibillion-dollar government contracts and strong establishment ties.

The idea of the nimble maverick overthrowing lead-footed incumbents is, of course, the favorite startup narrative. But the people behind Anduril are not untested newbies; they have significant experience in tech and politics. Besides Luckey, who gave money to an alt-right group and donated to Trump’s inaugural committee, the team includes former executives from the secretive data-crunching company Palantir, whose work for many government agencies has raised alarms about intrusive surveillance. And Anduril’s lead investor is Founders Fund, the VC firm headed by Peter Thiel, a prominent Trump supporter and the guy who shut down Gawker.

The politics of Anduril’s founders may not be popular in liberal Silicon Valley, but they need to please a different audience: members of Congress and government bureaucrats. To win big border contracts, Anduril must beat out other companies peddling visions of an electronic border wall, including an Israeli firm called Elbit Systems, as well as traditional defense giants. Its advantages are operating cheaply and moving quickly. In a little over a year the company not only built and deployed its prototype in Texas, it has also launched a government-funded evaluation project under way outside of San Diego. It promises a system that would cost a small fraction of a physical wall and is cheaper than its digital competitors.

Of course, Anduril still has to prove its technology works in a more extensive test. But early signs look good. According to US Customs and Border Protection, in a 10-week period, Lattice’s test in Texas helped customs agents catch 55 unauthorized border crossers, a notable figure for a system still in development. If Luckey has his way, the border wall of the future will be Anduril’s.

At Anduril’s headquarters in Orange County, Palmer Luckey has been hands-on in building the company’s technology.

Benjamin Rasmussen

Luckey grew up in Long Beach, California; his dad was a car salesman, and his mother homeschooled him and his three sisters. “I was a PC gamer,” he told me in 2015, “and I was always upgrading my PC, getting the best monitors, the newest graphics cards.” He wanted to feel as if he were “actually in the game, like the game is actually real.” By collecting and sublimely tweaking the technology available, Luckey created a homegrown VR system. He called his system Oculus and described it on a Kickstarter page in August 2012 as “designed by gamers, for gamers.” But when Mark Zuckerberg tried it out in 2014, the Facebook CEO saw it as the social computing platform of the future. Facebook bought Luckey’s company for $2 billion.

In June of that year, a newly enriched Luckey attended a retreat hosted by Founders Fund, which had been an early Oculus investor, on Sonora Island in British Columbia. There he met an employee at the fund named Trae Stephens, then age 30. Earlier in his career, Stephens had worked at a government intelligence agency that he will not publicly identify; in 2008, he joined Palantir. In 2014, Thiel convinced Stephens to join Founders Fund and specialize in investments involving the government. Stephens found it ridiculous that almost no venture-backed companies worked closely with the government, with its billions of dollars to spend. “After Palantir and SpaceX, there’s nothing,” he says. Founders Fund also was an early SpaceX investor, and Stephens’ goal was to fund a company to join that duo. He was coming up empty. The Valley, it seemed, didn’t do government.

Over meals at the Canadian eco-resort, Luckey and Stephens bonded over a shared passion for defense tech. Luckey had once worked on a program that used VR to treat PTSD, which led him to think about how military tech worked—and how it didn’t. During his Oculus years, he had read up on projects like the troubled F-35 fighter, which had a problematic head-up display, and realized that applying lessons from the consumer world could improve its design and lower costs.

After the Sonora Island trip, Luckey and Stephens kept in touch, and in 2016 the pair began speculating about starting a company together. They threw around a lot of ideas, some of them straight out of comic books—What if we built a force field? As that year ended, Stephens was making regular trips to Washington, DC, from San Francisco. Donald Trump was the president-elect, and Thiel, who was on the presidential transition team, brought Stephens on to focus on the Department of Defense. It was a useful post for someone thinking about a defense business.

Meanwhile, Luckey’s political activities had made him the object of tech-press scorn. News reports claimed that Luckey was involved in an alt-right group called Nimble America, paying for billboards ripping Hillary Clinton as “Too Big to Jail” and allegedly penning vicious Reddit posts for the group. On his public Facebook page, he denied many of the allegations but confirmed that he donated $10,000 to Nimble America because he “thought the organization had fresh ideas on how to communicate with young voters.” He apologized for “negatively impacting the perception of Oculus and its partners.” When asked about this now, the normally buoyant Luckey drops his smile and chooses his words carefully, claiming that his politics are misunderstood. “The alt-right, as it exists, as it’s defined, I do not support, never have,” he says. He describes himself as “fiscally conservative, pro-freedom, little-L libertarian, and big-R Republican.”

On the last day of March 2017, Luckey was ousted from Facebook. Neither party is sharing the details of his exit. (The issue even came up at Zuckerberg’s April 2018 Senate hearing, when Republican senator Ted Cruz, who has received $5,400 in political donations from Luckey, demanded, “Why was Palmer Luckey fired?” Zuckerberg said only that it wasn’t because of his politics.) And what did Luckey learn from his experience at Facebook and Oculus? “Be careful who you trust,” he says. “Be careful who has control.”

On his first day as a free agent, Luckey connected with Stephens, ready to start building the company they’d discussed. Stephens didn’t hesitate. Their guiding vision was something like Stark Industries—the mind-blowing font of matériel in the Iron Man movies. (Luckey is a voracious consumer of popcorn flicks; one of his favorites is Pacific Rim.) And it would probably involve VR.

A collector’s version of Anduril, the enchanted blade from Lord of the Rings, hangs on Luckey’s wall.

Benjamin Rasmussen

They began recruiting a team. Stephens suggested Matt Grimm, a former Palantir colleague. Luckey proposed a fourth cofounder, Joe Chen, an engineer who had worked at Oculus before joining a Hollywood VR startup. Chen had also served in the National Guard. Both men signed on. “I’d been an end user on some very, very bad VR military simulation systems,” Chen says. “Once Palmer said ‘Hey, we gotta fix this,’ I was like, ‘All right, cool.’ ”

On April 7, exactly a week after Luckey left Facebook, the four invited around half a dozen potential recruits to Luckey’s Orange County home. As the guests ate Chick-fil-A, the founders presented a pitch deck. By attracting “disruptive talent with a Silicon Valley vision, Anduril will be the next great defense company,” it promised. They would need “crazy mad scientists,” political connections, and lots of capital. “Almost every single person that was at that initial dinner is here right now,” Stephens says.

Luckey secured warehouse space in an industrial area of Orange County. When the team approached Founders Fund, Brian Singerman, a partner who was also the first Oculus investor, agreed to lead the fund’s $17.5 million seed round. “Palmer is an insanely brilliant technologist,” he says. “A little bit … out there. But most brilliant people are.” (This May, Founders Fund led a $41 million Series A round.)

Luckey, Stephens, and Grimm also made their pitch to Palantir’s directors. In attendance was Brian Schimpf, Palantir’s head of engineering. After the session, Schimpf told them he wanted in. He became the fifth cofounder and CEO, with Grimm as COO and Luckey as CTO. Stephens chairs the board (he never left Founders Fund).

The company’s name also has a Palantir connection. Middle-earth buffs will recognize Anduril as the enchanted blade that was Aragorn’s go-to lethal weapon; a palantir is a magical crystal ball from the same Tolkien universe. “All of us are Lord of the Rings fans, so it was a pretty fun name,” Luckey says. “Also, I have Anduril the sword hanging on my wall.” (Luckey procured a collector’s version, not the original movie prop.)

They had a name and an executive team. But what was the product? “The DOD has been asking for what some people describe as Call of Duty goggles,” Luckey says. “Like, you put on the glasses, and the headset display tells you where the good guys are, where the bad guys are, where your air support is, where you’re going, where you were.” (Pause to consider this Escher-esque scenario of soldiers clamoring for gear inspired by a game that mimics their combat experience.) But tiny Anduril—with no experience or history—couldn’t just barge into the Pentagon and demand to build battlefield tech. “We needed a quick win,” Schimpf says.

Anduril’s pitch deck offered a sci-fi fantasia, including autonomous long-range bombers, attack-drone swarms, and something they called “perimeter security on a pole.” The team zeroed in on this last notion. They figured they could build a surveillance tower using off-the-shelf sensors and cameras, connect them in a network, and make something in the spirit of Google Maps and Pokémon Go. By using AI, the system would identify what data was important. Stephens thought the Pentagon might see its value in securing forward operating bases—outposts in hostile territory. But Luckey had another idea: border security. A system to monitor America’s southern perimeter would require components similar to those in a combat awareness platform. What’s more, it was clear Mexico wasn’t paying for that big, beautiful wall that Trump had promised. The government, they realized, might be receptive to their budget-friendly pitch.

Silicon Valley, meet the US-Mexico border.

The border towers include radar, communications antennae­, and a camera enhanced by a laser from a hair-removal device.

Benjamin Rasmussen

Screens show information about Anduril’s towers.

Benjamin Rasmussen

To find their way to the border, Anduril executives started by approaching a California office of the Department of Homeland Security in June 2017. “They said they could provide broader border security for a lower cost. We were intrigued by that,” says Melissa Ho, managing director of Silicon Valley’s DHS office. The DHS introduced Anduril’s executives to border patrol officials, and a border patrol team near San Diego was happy to brief them. “They saw us as their own SpaceX,” Schimpf says—that is, a nimble private entity that could provide specialized technology. Later, when the San Diego office of Customs and Border Protection was setting up tests of new border systems, it selected Anduril for a pilot project.

Anduril is suggesting a new way to secure the border electronically, but it is far from the first. Hundreds of millions of dollars have been spent on comically ineffective systems (in one of them the radars would get activated by rain). In the mid-2000s, Homeland Security initiated a competition to create SBInet, a comprehensive virtual wall. In September 2006, Boeing won a contract to start building a system that was estimated to cost $7.6 billion. It began constructing 80-foot-high towers loaded with equipment. In January 2011, after a series of cost overruns, late deliveries, and a basic failure to catch people crossing the border, then Homeland Security chief Janet Napolitano pulled the plug. With massive understatement, a DHS report said that SBInet “does not meet current standards for viability and cost effectiveness.”

SBInet was a case of government contracting run amok. “You learn lessons from failure,” says CBP commissioner Kevin McAleenan. Anduril, like both Palantir and SpaceX, seeks to avoid some common pitfalls. Instead of selling technology to the government for a huge up-front fee, it plans to own the system and lease it, with the data it collects belonging to whatever agency issues the contract. This arrangement, Stephens says, creates an incentive to keep development costs low.

Part of SBInet’s failure was that it came too early. Sensors that cost a few bucks today were thousands of dollars a decade ago. Artificial intelligence is no longer an aspiration but a tool that delivers results. But technological attempts to secure the border have also tended to rely on complicated technologies, such as Predator drones, that aren’t cost-effective for long stretches of the border. A much simpler surveillance system could work fine, as long as agents received useful alerts from it. “The key was just finding a way to get information in the hands of agents,” Schimpf says.

The camera mounted on Anduril’s surveillance towers takes still images in quick succession.

Benjamin Rasmussen

Its competitors in the smart-wall business were pitching taller towers with exotic microwave transmitters and other bespoke gadgetry. For Anduril, the key to making consumer tech work was to combine it with AI. The company taught its software to identify the patterns of a person on the move, allowing it to avoid the expensive zoom lenses and thermal sensors used in competing systems, Schimpf says. “The sophistication of Nest-level technology isn’t bad,” he says, referring to the smart thermostats and motion detectors designed to automate a home. “And no one has used AI for this purpose yet. If you can identify objects with AI, you don’t need to see as far.”

Within a couple of months, Anduril had a prototype. Schimpf and his colleagues took it to a test range in Apple Valley, a two-hour drive from their Orange County office. “We lived out of the trailer there,” Schimpf says. Using open source machine-­learning training data, they taught the software how to tell humans from animals or tumbleweeds, and unearthed some glitches. In a certain light, for example, the system can mistake the rear end of a horse for a person.

What they didn’t find in affordable parts was a way to capture distant moving objects at night. Thermal cameras cost hundreds of thousands of dollars and fare poorly in the wind and dirt of the Texas border. But Luckey had an idea: Sync a laser beam to a virtual shutter, similar to flash photography. “We shoot a flash beam way, way, way out to where you are,” Luckey says. “It lights up you and the area around you, and then we’re able to pick that up with our electro-­optical sensor.” Anduril discovered it could cheaply repurpose the laser, which it bought in bulk, originally meant for a 600-watt cosmetic hair-removal device.

To test their prototype, Stephens called Will Hurd, a Republican congressperson whose district includes the nation’s longest stretch of land bordering Mexico. Hurd has long argued for a digital approach to border security, so when he heard Stephens’ pitch, he perked up. “A lot of contractors say ‘Oh yeah, I can do this,’ but the federal government’s going to have to pay for the prototypes and all that kind of stuff,” he says. “When Anduril representatives explained their approach, I was like, ‘This is pretty cool.’ ” Hurd introduced Stephens and Luckey to a rancher on the border who agreed to host three test towers.

A screenshot of an Anduril mobile notification.
Courtesy of Anduril
Anduril Mobile App screenshot of a person walking.
Courtesy of Anduril

In mid-April, Luckey, Stephens, Schimpf, and I are sailing down Highway 90 in southwest Texas in a rented SUV heading to that ranch, a road trip that started with a pit stop at an El Paso Whata­burger (Luckey’s choice). It’s a long drive through the sagebrush-­covered desert, with Schimpf at the wheel. “This is a place where machines are supposed to live,” Luckey says, “not people.” Luckey has a cold, but he chatters between sniffles about movies and technology, and he tells a story about hanging out in VR with Ready Player One author Ernest Cline. He’s wearing his trademark Tommy Bahama aloha shirt, shorts, and flip-flops; the others are in the Silicon Valley cool-weather uniform of puffy jackets and jeans.

Schimpf takes a right at an unmarked intersection. We travel over roughly 30 miles of an unpaved road populated mostly by rabbits to reach a gate with a faded sign that designates the ranch as a member of the Texas and Southwestern Cattle Raisers Association. Beyond it, inside a comfortable farmhouse, waits Ed, the fifth-­generation owner of the property. (WIRED is changing his name to spare him the attention of drug cartels operating in his area.)

With a sagging mustache and a wide drawl, Ed has the air of a canny retired sheriff in a Sam Peckinpah movie. Over coffee, he explains that for decades his land has been an unpatrolled gateway to the United States. Past trials of new technologies hadn’t worked out, but on Hurd’s urging he gave the Californians a chance. To his astonishment, their system seems to be performing well.

Take what happened on March 5. At 7:41 am local time, the system noticed activity in the valley. An alert popped up a thousand miles away, on Matt Grimm’s phone. “New person track near tower e1,” it read. Grimm, who was at his home in Orange County, opened Anduril’s app—and saw a dozen people making their way across the gullies and hills of the Texas frontier.

In an official installation, such alerts would go straight to Customs and Border Protection agents. But in this case, Grimm notified Ed. Ed called the nearest patrol station and settled into his living room couch with his laptop. Launching Anduril’s software, a wide shot of his land filled the screen. Blinking green rectangles highlighted the trespassers; zooming in, he could make out the group of figures more clearly. Between sips of his morning coffee, he watched the boxes inch across the screen as the people traversed his ranchland. “I can hardly operate a cell phone,” he later recalls. “This is beyond cool.”

Artificial intelligence identifies whether it has spotted a moving person, vehicle, animal, or tumbleweed.

Courtesy of Anduril

Later that morning, Grimm could see that a DHS helicopter was headed toward the scene. By then the visitors had traveled northeast of the towers’ range, so he couldn’t watch as border agents apprehended 12 people. In a 10-week span since the towers were installed, Lattice helped agents catch 55 people and seize 982 pounds of marijuana. (For 39 of those individuals, drugs were not involved, suggesting they were just looking for a better life.) The official test outside San Diego, ongoing at press time, led to 10 interceptions in its first 12 days.

Last July, Hurd introduced the Secure Miles with All Resources and Technology (SMART) Act, which would direct DHS to deploy technologies for “situational awareness and operational control of the border.” His nine cosponsors include two Democrats. The bill is awaiting a vote, but some of its key ideas found their way into the 2018 federal budget, which provides funds for border-security technology.

“Nobody is disagreeing with the smart wall,” says Hurd, a former CIA agent who is one of the few members of Congress with a computer science degree. The economics are an obvious factor. “A concrete structure 30 feet high that takes four hours to penetrate costs $24.5 million a mile,” he says. “A smart wall, a system like what Anduril is proposing, is about a half a million a mile.”

The prototype towers on Ed’s ranch are the epitome of cheapskate. A plain metal pole juts from the ground, propped up by a tripod and anchored by cinder blocks. Solar panels hang off its lower portion, and the top is a cobbled-together cluster of radar antennae, cameras, and more. The finished version will look more polished, but its bones will be the same. Unassembled, it can fit into a pickup truck and be installed in less than an hour. Anduril has since added small drones to the system. If the company wins a contract for the hundreds of miles of rural borderlands, where its tech is best suited, these towers will watch the movements of all who cross their line of sight—drug smugglers, job seekers, families, as well as Americans going about their business.

Staring down from our bluff toward the Rio Grande, Schimpf reflects on Anduril’s long-term goals, which include protecting private sites like oil pipelines and monitoring the battlefield of the future. “Looking at this helps you conceptualize what it would be like in Afghanistan, if you had a forward-operating base on top of this hill. It’s the same problem.”

Anduril is trying out small, cheap helicopter-­style drones.

Benjamin Rasmussen

Anduril is unusual among today’s startups for embracing the defense business. In the Valley, many believe that “if you do defense you must be an evil person,” says Joe Lonsdale, an Anduril investor and Palantir cofounder. But that wasn’t always the case. California’s tech sector was once a virtual branch of the military. “Literally 100 percent of the early microchips went to defense use,” says Leslie Berlin, project historian for Stanford’s Silicon Valley archives.

In the 1950s and ’60s, “working for the defense effort meant working for the good guys,” Berlin says. After Vietnam, that changed. “Many people in Silicon Valley today don’t feel that way.” The most recent evidence came in April, when The New York Times reported that more than 3,100 Google employees had protested the company’s work on a Pentagon-backed AI effort called Project Maven.

Lonsdale and Luckey argue that building cheaper, more efficient systems is a virtuous pursuit, saving taxpayer dollars. Anduril’s Palantir pedigree may have prepared it for criticism. As that company grew to a private valuation of $20 billion, its technology has been portrayed as Big Brother–style surveillance tools. Anduril’s leaders tread lightly on the subject of deadly force—traditionally the purview of defense companies—and have a ready answer when I ask whether the company will ever build systems that kill people.

“We’re really focused on the intelligence and surveillance piece right now,” Schimpf says. But in the next beat: Not that there’s anything wrong with building weapons. “I wouldn’t say that’s a line we’re drawing.”

Stephens jumps in to clarify. “Part of the unintended shiftiness of our answer is that no one even knows what that means. In 20 years, are we still going to be filling bullets with gunpowder? Or is this electronic warfare? Is it like sending a pulse out that takes drones out of the sky?”

Put that way, warfare sounds a bit like a videogame, an echo of the drone pilots who execute deadly missions from behind computer screens many miles away.

MythBusters cohost Jamie Hyneman is building an autonomous firefighting tank for Anduril.

Benjamin Rasmussen

In a steampunkish workshop in an industrial area of Oakland, California, Anduril houses a project called Sentry that brings this parallel to life. Sentry is a fleet of autonomous firefighting machines meant to battle blazes on California’s hills, among other applications. The idea is to hollow out armored troop carriers to hold more than a thousand gallons of water. With crinkled aluminum skin, a Sentry vehicle looks something like a battlebot tank. That’s no coincidence—Anduril’s subcontractor for the project is Jamie Hyneman, the special effects expert and former cohost of MythBusters who built one of the fiercest battle­bots in Robot Wars history.

Luckey passes me an Oculus Rift headset and a handheld controller to try driving a simulation of a Sentry vehicle. On my headset I see a stand of burning trees. I set the tank on autonomous mode and use the index-finger trigger, familiar to anyone who has used an Xbox, to shoot its water cannons at the blazes. It is exactly like playing a video­game. As the flames spread, I concentrate hard to rule over the conflagration, wanting to put in a strong performance for the Anduril team.

I leave the Oakland workshop pumped from the excitement of saving the homes of imaginary Californians. But as I steer my car through the battered chain-link gate, past graffiti-covered buildings, the lingering adrenaline from my digital immersion turns to a funny aftertaste. The California fires last summer were devastatingly real. So is warfare. Anduril is on a quest to build awesome tech, the stuff of comics and action films. But it will be deployed in situations of human desperation, a vast remove from the land of fun. Transforming consumer tech’s plowshares into swords is ultimately a dark pursuit.

It struck me after I’d wrapped up my visits with Anduril that, aside from the drug smugglers they helped intercept on the border, I had not heard the founders mention the people who might get caught in their omniscient zone. What is the right way to treat those individuals? What of the children and parents who are now being torn apart while crossing? Those are social and political questions, not technical specifications. But it is increasingly the case that the people who build new technologies trigger political consequences.

Though tech companies have been taking their knocks lately, even the ones now under the most scrutiny were launched in a glow of idealism. We once dreamed that an era of ultraconnected and infinitely empowering tech would solve the kinds of problems that lead people to flee their own countries or that propel terrorists or nations to attack. Those problems didn’t end. It now seems obvious that tech was never going to make us better human beings; we are still our flawed selves. Instead, those same technologies that once seemed full of promise are finding their way into all-too-human clashes—led by a company named after an avenging sword.

Anduril’s founders come from Oculus, Founders Fund, and Palantir.

Benjamin Rasmussen

Steven Levy (@stevenlevy) wrote about cryptographic back doors in issue 26.05.

This article appears in the July issue. Subscribe now.

Listen to this story, and other WIRED features, on the Audm app.


More Great WIRED Stories

Read more: