The Patriot: Shyam Sankar of Palantir

Click here to subscribe to print for your office or home.

The Cotillion Room in The Pierre hotel has wall panels painted with pale green laurel wreaths, plush carpet patterned with grey-blue arabesques, crystal chandeliers dangling from oval recesses, and heavily draped windows overlooking Central Park—all of which makes it seem natural here, if only here, when people still refer to themselves, with a straight face, as members of “the conservative movement.” By the look of it, there are around 400 of them here tonight, ranging from ages 20 to 110. Many are sipping the evening’s signature cocktail: an Old Fashioned infused with apple cider syrup and served over a large ice cube engraved with the monogram of our hosts, the Hudson Institute, a conservative think tank in Washington. On this cold November night in Manhattan, we are gathered here to celebrate the presentation of the 2025 Herman Kahn Award to Alex Karp.

After an hour of drinks and distributing genialities around the Cotillion Room, we are steered into the Grand Ballroom, where the atmosphere closes in a bit under pressure from still more ivory, muted creams, soft golds, and twinkling crystal. People take their seats with varying degrees of resignation to their distance from the stage and to the designated VIP at their table. At one is the expressionless ambassador from Taiwan; at another is the more ruffleable looking ambassador from Israel. There is also Nikki Haley, still radiant, like a star that shines brightly on Earth long after it has ceased to exist. There is Rupert Murdoch, in a blue suit and red tie, discharging particulate mummy matter. There is Bill Barr, looking like he does.

The lights go down and our attention is directed to a hype video for the Hudson Institute, featuring footage of its resident fellows cut presumptuously with B-roll of North Korean conscripts marching, ISIS caravans barreling down the desert, etc. The video ends, the lights undim halfway, and it is time to present the annual Herman Kahn Award—named for the Cold War nuclear strategist and inspiration for Dr. Strangelove—to the co-founder and CEO of Palantir Technologies, who might have also been played by Peter Sellers. The previous year’s Herman Kahn Award was given to Mitch McConnell. The year before that to Charles Schwab. Karp is a nice change of pace around here, it seems.

Indeed, in the days before the gala, Karp had been all over the news and leaping from every screen, doing battle with Michael Burry, the investor who predicted and profited from the 2007 subprime mortgage crisis (and was subsequently played by Christian Bale in The Big Short). The previous Monday, Burry-watchers had plastered the internet with his most recent securities filing, which showed put options involving several millions of dollars’ worth of Palantir shares. Other short sellers, convinced Palantir is a “meme stock” with a valuation 200 times forward earnings, had also contributed to a slide in the share price. Burry, whose SEC filing also showed him shorting Nvidia—and who it should be said has lost a number of bold bets since the one in 2007 that made him rich and famous—explained his position by insisting that Palantir’s valuation relative to its economics is absurd; that its claims to be an “AI company” are basically bullshit; and that its stock price will collapse by about 85% in two years. Given Karp’s torrential and often opaque way of speaking, it was hard to say what his televised rebukes of Burry had accomplished.

Not that any of this spoiled the mood in the Grand Ballroom of The Pierre. On the night of the 2025 Herman Kahn Award, after all, Palantir’s market cap was still approaching half a trillion dollars. It had just reported its ninth consecutive quarter of profit and growth, and its share price was still near an all-time high of $200. More importantly to the national-security-minded Hudson crowd, their Herman Kahn Award winner was still every bit at the spearhead of “defense tech.” Most of the founders of Anduril, for example, cut their teeth working at Palantir, and several more alumni have gone on to found and staff much of the frontier defense technology ecosystem. The year, moreover, had seen billions in new contracts for Palantir with NATO, the U.S. Army, Marine Corps, the Department of Homeland Security, and the British government.

Between its work with the military and intelligence agencies; its openly ideological posture about defending “Western civilization” and “American greatness”; and the legions of fanboys that patrol the discourse around its eccentric leadership (which includes the company’s founder and chairman, Peter Thiel), Palantir in a room like this one is loved and feared, seen as both omnipotent and forbidden, mysterious but profound. Its powerful software is said to have been used in the Bin Laden raid, in the strikes on Iran’s nuclear facilities, in America’s ICE raids, in the capture of Nicolás Maduro, and to stop countless attempted terror attacks that, for all you know, might have killed you and your family.

Palantir has by now been around so long, and been the subject of so many chin-stroking think pieces on the ethical implications of its software, that it’s a wonder what a difficult time expert observers, ordinary people, and even the company’s executives still have explaining what exactly it does and doesn’t do, which civil liberties (if any) it could plausibly be accused of breaching, and why anyone should associate technology as mind-numbingly dull as “data integration and visualization” with either violations of the U.S. Constitution or of saving Western civilization. No doubt the company benefits in many ways from such opacity, which might explain why even its more literal-minded executives than the gnomic Thiel, or the circumloquacious Karp, often still struggle to discuss such matters without analogies to Hobbits, the proliferation of rhetorical questions, and tenuously generalizable examples of individual contracts.

Nor is it easy for a company to maintain an aura like that once its executives have entered the stage of their lives in which they give each other trophies in the Grand Ballroom of The Pierre. Yet the aura unmistakably trails Shyam Sankar, Palantir’s chief technology officer, as he ascends the podium to introduce Karp and present him with the Herman Kahn Award. The mononymous Shyam (pronounced Shahm, rhymes with “bomb”) is 43, 15 years younger than Karp, but has been with the company nearly as long. In March 2006, at the age of 24, he became the company’s 13th employee and never left. Grabbing the microphone in the Grand Ballroom, he appears cherubic and guileless. Yet if Karp is the duo’s oracular Greek, as more than one Palantirian put it to me, then Shyam is the Roman: the road builder, the doer.

“He’s been one of the most impactful people in defense tech, working for 20 years, and he’s done it privately, quietly, and very much behind the scenes,” said Katherine Boyle of a16z. “And I think that’s why he’s been so effective.” Ted Mabrey, Palantir’s head of commercial, said that “Without Shyam, none of it works. There are people in high-stakes environments who use our software who are alive right now because of Shyam.” “At every stage of this company he’s had a but-for role,” Karp told me. “But for Shyam it would have gone differently. But for Shyam, honestly, I’d still be sitting alone in New Hampshire as an introvert.”

Shyam begins his brief remarks by quoting from Karp’s battle with the short-sellers (“I love the idea of getting a drone and having light fentanyl-laced urine spraying on analysts that tried to screw us,” Karp had said while ostensibly promoting his book), before praising him for his conviction “that serious technology should serve serious purposes; that the West needs to defend itself with both wisdom and strength; and that you can’t protect democracy without occasionally making everyone a little uncomfortable.” “Next to my late father,” Shyam said, “no person has had a greater impact on my life than Alex.”

To a standing ovation, Shyam returns to his table as Karp ascends the stage. Taking the podium, he gazes out at the audience with that look of his in which he seems to be trying to work out whether we’re all about to have sex or going to hit him. Before getting to the meat of his speech, he begins, unexpectedly, with tears audibly stuck in his throat. “I have a lot of remarks that I wanted to make,” he began. “But I would say for all of you who are younger and on your way to build important things, that there is no greater award than getting a speech like that from Shyam.” “We all know the ups and downs we went through,” he choked up. “Shyam’s father was one of the great figures in building Palantir, because he told Shyam, ‘You may never leave Palantir.’ That’s probably why he didn’t leave.”

At the end of a speech that I think was about the need for America to be a values-based meritocracy that leverages technology and specifically AI to defend freedom and maintain hegemony, Karp returned to Shyam. “Meritocracy is the most underestimated, powerful, revolutionary tool that exists in any enterprise ever,” he said. “Shyam is a great example. Shyam was not everyone’s cup of tea.” Reaching his conclusion, Karp recalled first meeting Shyam 20 years earlier and watching him interact with his little brother. “If Shyam could love Palantir half as much as he loves his brother,” he remembered thinking, “then we are going to have the most important enterprise in the world.”

After another standing ovation, we filed into the Garden Foyer to partake of the Viennese Table Station, where there were, quite literally, several thousand miniature desserts. There were mini cupcakes, mini cream puffs, mini strawberry cheesecakes, mini bananas foster, mini apple crumbs, mini salted caramel chocolate tarts, and mini key lime pies, the delighted consumption of which by hovering adult men gave one a tinge of second-hand embarrassment.

When I looked around the room to try to catch a figure of note in the act of eating a small treat, I clocked the presence of nearly everyone else I’d seen that night, except for Shyam, who’d disappeared.

Late one night in 1984, in Lagos, Nigeria, five armed men broke into the Sankar home. They pistol-whipped Shyam’s father, threatened to execute him, and threatened to rape his mother. After one of the assailants prevented the others from carrying out the worst, they decapitated the family’s German shepherd with a machete and nearly beat Shyam’s father to death. They broke the safe in the bedroom, took the money in it, and ran.

“As the story’s been retold to me, I was just excited that there was company in the house in the middle of the night,” Shyam said as we sat in the Georgetown, Washington offices of Palantir. “But it was highly traumatic for them. When Mom talks about it, you can feel the, sort of, ‘I don’t want to go there.’ It was a formative experience for Dad. He lost everything when that happened.”

Nochur “Shan” Sankar was the youngest of nine children who grew up in a frowsy hut in the prison compound where their father worked, in a flyspeck of a village in far southern India. When he grew up, his eight older siblings pooled their wages to send him to college, where he studied to become a pharmacist while sleeping underneath his older brother’s kitchen table. His sister-in-law, tired of him living on her floor after he graduated, found a classified ad for a pharmacist position in Lagos. She drew up a resume for her brother-in-law and mailed it in.

Shan was 24 when he arrived in Nigeria in 1975 to help build the first pharmaceutical manufacturing facility in Africa. The company, Vitabiotics, provided him with a small house across the street from the factory. The venture succeeded, Shan rose quickly, and became wildly successful for his age and station. Vitabiotics entrusted the cash from the operation of the factory to a safe they asked Shan to keep in his house.

In 1980, he returned home to India to meet the girl his parents arranged for him to marry. Girija was from Bombay and had never left India. Shan confessed to her to being a bad Hindu in his personal life (he ate meat and didn’t teetotal) but promised to help Girija keep a respectable home. They wed and had a baby boy, whom Shan’s mother named Shyam, meaning “darkness,” or “evening.” (It is also the word for the type of blue used to paint idols of Krishna.)

The Sankars moved together back to the house in Lagos, where Girija, in a bit of a state of shock to suddenly be living in Nigeria, set up an open kitchen in their home, cooking Indian breakfasts and lunches for workers in the Vitabiotics factory across the street and anyone else in the community. Things went well for a little over a year, until the break-in. The man who prevented the rape and murder of Shyam’s parents—and who was also likely the one who knew that there was a safe with cash in it in the first place—had been a regular in Girija’s kitchen. None of the five assailants were ever arrested.

“He was very street smart,” Shyam said of his father. “He’d learned how to maneuver around all the institutionalized grift in Nigeria. He’d keep an attritable wallet on him for when he got held up. He’d figured out the system there. But at some point, the bankruptcy of a corrupt system just overwhelms you.”

After the break-in, the Sankars left behind all their possessions in Lagos and Shan’s lucrative job at Vitabiotics, and returned to India to figure out what to do. As fortune would have it, he got a call one day from a childhood friend who’d emigrated to Los Angeles, where he set up a company that sold knick-knacks to the theme parks in Anaheim. “There’s this place called Orlando,” he told Shan. “All the theme parks seem to be going there. It’s up and coming. I need someone there I can trust. You’ve got nothing to lose. It would be a good way to start over.”

When the Sankars arrived in Orlando in 1985, they were one of four Indian families in the city. Shan, who only a few months before had been a respected and educated pharmacist and successful businessman, took a job, along with Girija, transporting souvenirs and what they call “costume jewelry” from the knick-knack factory to the major theme parks—SeaWorld, Busch Gardens, Wet and Wild, etc.—where they’d rack them in the gift shops. They enrolled Shyam at a local preschool and dressed him in all the misprinted t-shirts. “By the time I was in second grade, my classmates were old enough to make fun of me for it,” he said. “I stopped wearing that shit real quick.”

The best schools in India had been established by the Jesuits, so Shyam’s parents convinced themselves they needed to find a Catholic school for him in Orlando. He enrolled at St. John Vianney and was the only immigrant and non-Catholic in his class. “I had a great friend group there of just normal American kids who were also in their own way part of the assimilation journey,” he said. “I went to mass every Thursday and didn’t take communion. But they were very welcoming. The nuns were great teachers. I have very fond memories of it.”

In the afternoons, Girija would pick Shyam up from school and take him to a theme park, where he would wander around while she restocked the gift shop shelves with merchandise. On certain Saturday mornings, they would get woken up by the double sonic boom of the Space Shuttle, whose re-entry flight path went over Orlando, and run outside to look at the condensation trail, before driving to visit the Kennedy Space Center at Cape Canaveral.

After eventually leaving his work at the theme parks, Shan took a position with a different company that quickly got sold; he had no equity and lost his job. He started a dry cleaners, and then a dollar store. Both went bankrupt. “I remember during one of the bankruptcies, we came home from school one night and someone was parked outside to serve him process papers,” Shyam remembered. “It was profoundly humiliating and terrifying.”

After the second bankruptcy, Shan found a job as a pharmacy technician, counting pills and filling orders, often on night shifts; Girija took a job as a supermarket bagger while studying to become a nurse. “As a kid you find that embarrassing,” Shyam said. “Your friends and their moms are there getting their groceries bagged by your mom. But I look upon that now with enormous respect. They were just grinding. I was always the last kid to be picked up from school, they’d come get me at 5-5:30 after work and then start doing whatever their next job was after dropping me off at home.”

“The opposite,” he said, when asked if the precariousness of their circumstances made him ashamed or fearful, the twin conditions of life on the edge of poverty. “As stressful as it all was for Dad and Mom, I always felt like our house was my bedrock. The stability there enabled me to go take risks and be ambitious and do things. It never fazed Dad. He always felt like, ‘We survived what happened in Nigeria, we can survive anything.’ So actually he was always so… maybe happy is not the right word. Grateful is the word. Gratitude’s much more impactful than the facile happiness of making money.”

In the meantime, Shan became the treasurer of the nascent Hindu Society of Central Florida, the first Hindu temple in Orlando, though it had no idols and hosted no religious ceremonies; it was more like a community center and function hall for the gradually increasing number of Indian immigrants to the city. By the time Shyam was in fifth grade, his father learned about the SAT, and decided to set up a Sunday school in the temple to tutor children for it, raising donations from some of the wealthier families in the community. 

“We were one of the first families there; we experienced everything first,” said Shyam. “And Dad just always wanted to institutionalize everything we learned to help out everyone else. For him it was just like: ‘There’s this thing called the SAT. If my son does well on it, he will go to a good college. If he goes to a good college, he will succeed in America. If that happens, I’ll take everything I’ve learned and help the other kids do it, too.’” The temple, and the Sunday school, survive today.

His family’s passage “wasn’t like this Hollywood story of a poor immigrant who comes here with nothing and makes it big,” Shyam said, offering me a nicotine pouch and a can of Celsius, which along with the hoodie sewn into his blazer are part of his uniform. “It was more like, poor immigrant comes here and gets kicked in the teeth over and over again. And yet we made it.”

“He was never preachy, except for one thing,” Shyam reflected. “He’d always look at us and say: ‘But for the grace of this country, we’d be dead in a ditch in Lagos.’”

Toward the end of fifth grade, Shan casted about for the best middle and high school in the Orlando area, and kept hearing about Trinity Prep. It was private, expensive, about a 75-minute drive from their home in Kissimmee, and admissions for the fall were already closed. Shan put his son in their Toyota Cressida anyway and drove to the campus. Shyam complained the whole way there that he liked his Catholic school, liked his friends, didn’t want to transfer, and they couldn’t afford private school anyway. “Then we got there,” he said, “and I vividly remember peeking into the science building and seeing this lab with equipment. It was just fucking cool. I was like, ‘I want to go here. I want to go here, Dad.’”

The tour guide informed them that admissions were closed but applications would reopen in the winter for the following year, and gave them some information about the school’s lean financial aid program. Before heading back home, Shan made Shyam stop by the enrollment office to introduce himself to Ethel Danhof, the head of admissions.

“All of us have this conspiracy of people who help us along the way,” he limned. “She’s since passed away from multiple sclerosis. But for the grace of Ethel Danhof… it was just like… for us it was like divine intervention. That was a fork in the road for me.”

Whatever he said to her in their brief meeting compelled Ethel Danhof to convince the school to give Shyam the standard admissions test, even though admissions were closed and the Sankars would be a financial liability. He took the test, and a few days later the school called to let them know Shyam could enroll that fall on a full ride. “That was the first instance I can remember,” he said, clearly thinking of his later life, “where I saw someone who wanted to do something just being willing to break the rules and fuck the process to make it work.”

For the next five years, until Shyam could drive himself, Shan and Girija drove him two-and-a-half hours each day to and from Trinity Prep. When Shan drove, he’d spend the ride downloading science history and facts into Shyam, or else they’d take turns role-playing dialogue from The Little Rascals (when Shyam was still young), or from Rocky, Rambo, The Hunt for Red October, Red Dawn, Knight Rider, and the John Ford classics (as he got older). “The Westerns were big for me and Dad,” he said. “You’re alone and not afraid. Life is hard, it is what you make of it. Even when no one knows you and your life’s at risk, it’s still important to be a hero and not a coward.” When Girija drove, she’d mostly let her son daydream out the window in silence. As he looked out at unmanicured vegetation and swampland studded by stripmalls and stoplights for over an hour each way, he fantasized about going to space. “That habit stuck with me,” he said. “To this day when I wake up in the morning, I need about an hour to be in my own head before I kind of boot up and go.”

At Trinity, Shyam was teased for being good at school; at one point he adopted the pose of an aggressive skater and made friends with the other skater kids, who eventually ostracized him. “I think they found me annoying and they didn’t know how to handle it,” he said. “So their way of excommunicating me from the group was to make a shit ton of fun of me for being Indian and an immigrant until I lost my shit, and they were hoping that that would be the way to break up with me. I don’t think they were actually racist. They just didn’t know how to say, ‘You’re annoying, go away.’” He eventually fell in with kids into health and fitness, who remained his social group through high school. When they weren’t weightlifting and going to movies together, Shyam loaded up on APs and took calculus BC at the University of Central Florida, which again required Trinity to break the rules to accommodate him.

“I was one of the few kids with no means, surrounded at a school by people with enormous means,” he said. “It wasn’t hard, really. It just puts fire in your belly. It was like, ‘Why can’t I have all this? Why not me?’”

During the summer between 10th and 11th grade, Shan asked an executive at his pharmaceutical company, “a salt of the earth, quintessential Midwesterner” named Albert Past, to give Shyam an internship. Shan had a thick accent and wasn’t very articulate, but Albert liked him and thought him smart and a worker, so agreed to help his son. He stuck Shyam on the technical team to teach him how to code. On his first day, Albert introduced Shyam to the team and said: “Teach him how to build his own Linux box, install Linux on it, install the Apache web server, and start writing Perl.”

“I didn’t know what any of this shit meant,” said Shyam. “But by the end of the summer I built a production application. It was like a digital fax machine—you would send your prescription faxes in, it would process it, it would store, route into the system. It’s toy-like now when I think about it, but that was when I first felt the power of how you could use technology to just build things. My code was never the cleanest; some people can just bang out code like poetry. I was more like brute-forcing my way through it. But it gave me an opportunity to understand the systems from the lowest level up and the integration of those into business processes. And I just got the builder bug.”

That was the first instance I can remember where I saw someone who wanted to do something just being willing to break the rules and fuck the process to make it work.

Shyam wanted to go to MIT but didn’t get in, likely because his essays sucked. “Who knows and who cares,” he said, “but you could speculate that if you have the scores to get in and didn’t, you probably failed the likability test. It’s an old story. The guy in my class who got in there who didn’t have the scores wasn’t Asian.” He enrolled instead at Cornell and studied electrical engineering. He graduated in three years, which included an attempt to dropout to join the CIA after 9/11 (his application went unacknowledged), and a stopout to build a company that would provide finance, accounting, and IT services for small companies (it failed).

After Cornell, he got a master’s in management science and engineering from Stanford; in one of his classes that assigned the creation of a mock business plan, he drew one up for an electronic remittance provider. As part of the project, he researched potential competitors, which included Wells Fargo, MoneyGram, and a new startup in San Francisco called Xoom. He decided to email the founder, Kevin Hartz, who agreed to get coffee with him. Ten minutes into their meeting, Hartz took Shyam up to Xoom’s offices to meet the other founders, Roelof Botha and Keith Rabois. The company was backed by Peter Thiel. It was the spring of 2004, and Shyam hadn’t heard of any of these people. By graduation day in Palo Alto, he had two offers: one from Boston Consulting Group, and the other from Xoom.

As badly as he wanted to join Xoom and hated the idea of being a consultant, it was a wrenching decision. It was a choice between making real money immediately—which among other things meant being able to help his parents—or taking a flyer at the little startup, with the smart and exciting people who nevertheless couldn’t make him any financial or professional promises.

“Dad had been kind of trying not to be directive of what I should do,” Shyam said. “But when the time came, he put his foot down. He was like: ‘I didn’t make all these sacrifices so that you could take the easy way. Go take a risk at the startup with the people you like. What do you have to lose?’”

“He was just this perfect boy scout, very kind of prim and proper,” Hartz told me. “He was very well-spoken, very enthusiastic and positive; very good energy and intensity. I hired him to be in business development. His main duty was to travel to all these different countries, forge these relationships, get the economics right, get us going in different markets. He was kind of the perfect beast for that in terms of being well-spoken, intense, highly intelligent. He ended up being very influential on the business model of Xoom.”

Shyam’s first task was to rescue the company’s launch in the Philippines, which was turning into a disaster. In essence, the company’s spec back in San Francisco was not meeting the ground reality of what the engineers in the Philippine banks were dealing with. The workflow was wrong, the file format was wrong, the two systems couldn’t transmit state, and as a result, the banks weren’t able to take the Xoom-facilitated remittances and upload them into their general ledger system. Shyam, not yet 22 years old, embedded with the banks’ engineers and fixed the on-site technical issues, relayed back to San Francisco what the problems were with the spec product, and cooled down the bank managers who’d been open to working with Xoom in the first place. “I wouldn’t have had the words for it at the time,” he recalled. “But I was a forward deployed engineer.”

Nearly two years had gone by at Xoom when one day Shyam went swimming with a friend at the South End Rowing Club in San Francisco Bay, near Fisherman’s Wharf. He confided to his friend that the work at Xoom hadn’t been feeding his “lizard brain” of late—he felt he wasn’t learning enough and that the company wasn’t growing fast enough. (Xoom went public eight years later and was subsequently acquired by PayPal.) Shyam’s friend, thinking of ways to help get him out, mentioned that his college roommate was starting a company to catch terrorists. “You should meet him,” he said. “His name’s Joe.”

Shyam went to meet Joe Lonsdale in January 2006 at the old Straits Cafe near Page Mill Road. Lonsdale had him come to the office soon after to meet the 11 other employees of their startup, including Bob McGrew, Aki Jain, Stephen Cohen, and Alex Karp. It was called Palantir, named for the “seeing stones” used by the kingdoms of the West to view things at a great distance in The Lord of the Rings. Shyam remembers being bewildered by the janky startup’s claims to be building one of the world’s most important companies, the insistence that their software would bend the arc of the American Republic, their seemingly collective prior experience as interns at PayPal and Clarium Capital (Thiel’s now-defunct hedge fund), and the incessant repetition of the fact that Karp had gotten a PhD in German philosophy by studying under Jürgen Habermas. “I was like, I’m a STEM student who works in remittances,” Shyam remembered. “Who the fuck is Habermas? And why are all these guys un-Googlable?”

“But they were so full of piss and vinegar,” he said. “They were building software to fight terrorism. And their thesis was: After 9/11, the political discourse was about privacy versus security. Americans were arguing that one or the other was more important, but of course if you had more of one you’d have to accept less of the other. But as engineers, they were like, why would we accept that tradeoff? Why can’t you push out the efficient frontier? Why can’t we build technologies that allow you to have more privacy and more security for democracy to choose from? And how do you bring the West together as an alliance so you can stop these terrorists?”

“If I’m honest, at the time I thought it probably wouldn’t work,” he remembered. “But after those interviews I was kind of ruined. Once I found out these guys were working on this, there was no going back. It was the last time I ever considered working at a big company or wasting my life building a calendar app or working at YouTube or whatever. It was just like, OK, I’m going to pour my soul into this. It’s just all risk every day from here on out. Once again it was a fork in the road where Dad was right.”

When Shyam joined Palantir in March 2006, it had been around for three years and had 12 employees. Thiel’s kernel insight—that the post-9/11 “security vs. privacy” debate was a false tradeoff that could be broken by software innovation in data engineering, and that perhaps the fraud-detection systems pioneered by Max Levchin at PayPal could provide a roadmap for creating enterprise intelligence software for spies—had convinced In-Q-Tel, the CIA’s venture arm, to make a $2 million investment. But it had not yet resulted in a product. That wouldn’t happen until the launch of Palantir Government (later renamed Gotham) in 2008, at which point the company was assisting the U.S. Army with counter-IED operations in Iraq and receiving visits from MI6. The troika that largely got it there consisted of Aki Jain, who led backend engineering; Bob McGrew, who led engineering overall; and Shyam, who was originally hired to lead “business development,” but had to create a new role that made more sense for a startup that had nothing to sell.

The problem that led to the intelligence failure of 9/11, and which still hadn’t been fixed or even really addressed five years later, was how America’s various intelligence agencies (there were 13 in 2001, and 16 by 2006) stored, shared, and analyzed the intelligence gathered by their systems and spies. The analysts’ primary software tool was called i2 Analyst Notebook, which allowed the user to create PowerPoint-like link charts nearly as primitive as the corkboard and red yarn kind used in Hollywood depictions of pre-internet municipal police stations: a picture of a guy’s face with a line pointing to another guy’s face and the words “reports to” or “travels with,” etc. Worse, the data that facilitated the link charts was stored in individual Excel files and other desktop apps that in turn were stored on over a dozen different systems and had to be manually uploaded from one system to another, which often meant analysts literally emailing each other file attachments. Worse still, the analysts generally served in their positions for only two years at a time; when their replacement arrived, they would have to make sense of Excel tables that essentially represented the idiosyncratic thought processes of a different person.

An example of this problem, and how long it continued to persist even after the catastrophic intelligence failures of 9/11, was the 2009 bombing attempt of Northwest Airlines Flight 253 by Umar Farouk Abdulmutallab, aka the underwear bomber. Roughly five weeks before the attack, Abdulmutallab’s father personally visited the U.S. Embassy in Abuja, Nigeria to warn the officers there that his son had become a religious extremist, was potentially planning something against the United States, and was likely in Yemen. The father’s tip was logged into the U.S. National Counterterrorism Center’s central database but was never cross-referenced with any other available intelligence, such as NSA intercepts confirming Abdulmutallab’s presence in Yemen, the fact that the UK had recently rejected his visa request, or that the CIA was aware of deranged anti-American posts he’d been making on social media. Because the selfless father’s heroic tip didn’t trigger a revocation of his son’s U.S. visa by the State Department, or his placement on the No Fly list by the FBI, Abdulmutallab capered onto a flight from Amsterdam to Detroit with high explosives in his underwear, which only failed to detonate due to degradation from sweat and dirt.

Nearly a decade after 9/11, in other words, U.S. intelligence analysts still didn’t actually have access to all the data the U.S. government collectively harvested, which meant, in a practical sense, that for all the concerns about how much data Big Brother possessed, a lot of it didn’t really exist. The first problem Palantir sought to solve, therefore, was the creation of an enterprise knowledge store. The second was how to actually integrate the information that already existed in multiple different databases into that centralized store.

The third problem was how to replace rigid database tables like Excel spreadsheets with a simple, transparent, and attractive user interface that would give analysts easy access to all the data that existed—an “ontology,” or generic data model, that categorizes the real world into distinct objects (e.g., people like Abdulmutallab, places like Yemen, objects like entry visas, media like Facebook posts, and events like the father’s tip). According to this ontology, these distinct objects could be defined by their specific properties and the relationships between them, and could be represented by data of any kind (e.g., smartphone videos, satellite imagery, wiretap audio, bank statements, flight logs, emails, social media posts, etc.)

“I was like, I’m a STEM student who works in remittances,” Shyam remembered. “Who the fuck is Habermas? And why are all these guys un-Googlable?”

The prototype Palantir took to its 2004 meeting with In-Q-Tel was famously built in eight weeks by Stephen Cohen, who believed that the aesthetics of the software should be the same as that of a video game. It wasn’t the Minority Report interface, but Cohen’s prototype did try to draw more on the mental models people had of such systems than of the soul-sucking desktop applications being churned out by Microsoft, let alone the government contractors making i2 Analyst Notebook. In Cohen’s demo, you’d begin by right-clicking on an entity—a person, say—and a visually attractive circle menu would appear around it with different events, flights, communications, or other people linked to that person. As the user pursued different searches, surfaced different data, and made connections between them, the software would quickly generate compelling graphs, structure the connections into pretty-looking networks, and lead to a subsequent set of logical actions or search options. At the end, the system would generate a PowerPoint export summarizing the investigation that could be automatically uploaded to a centralized knowledge base. If nothing else, it was smooth, fast, and nice to look at, and promised to automate away a lot of the bullshit work analysts spent their days doing, like emailing people to ask for Excel files and making PowerPoint summaries of their investigations.

The problem was it was just a demo, and while it worked within the limits of a prearranged scenario, it didn’t work well outside that scenario. Another problem was that it was clear to the CIA analysts they showed it to—including the ones who appreciated their coding chops and enthusiasm, and were equally horrified by the quality of the Agency’s existing technology—that the Palantir kids didn’t actually understand what intelligence analysts do all day. Nor could they be enlightened about it, as none of them had security clearances. “They would only talk to us using these strained metaphors about baseball cards,” McGrew told me. “Or they’d tell us, ‘Hey, that’s not really what I do. I do something else.’ They would never tell us what they did, just what we got wrong, which we had to use to figure out what they did.”

“So we had the ontology that you could map the data on and build semantics into, but we didn’t know how to do that mapping ourselves,” said McGrew, who eventually left Palantir after 11 years to lead research at OpenAI. “We needed actual spies to figure out how to do that mapping. It was Shyam who quickly realized that there was an opportunity here.”

The concept of a “Forward Deployed Engineer” (FDE) has since been the object of much imitation and mockery. While several B2B and enterprise SaaS startups have tried (and mostly failed) to graft the FDE label onto conventional sales roles, skeptics and antagonists have argued that it’s a glorified name for “consultant,” which Palantir uses to justify its business model as “a set of applications that were very expensive to install because you had to hire their consultants after you bought the software just to install and learn it,” as Michael Burry believes.

The truth is both more striking and more prosaic than either Burry or Palantir would have it. Shyam says it began with a claim Karp had made one day that French restaurants are good because the wait staff are part of the kitchen staff—rather than merely fulfilling orders and transporting plates, their knowledge of the kitchen’s techniques, methodology, and combinations, he explained, is as intimate as the cooks’. Karp’s point was that he didn’t believe in boffins sequestered in a lab polishing their Platonic ideal of a product; he wanted them to go figure out what customers actually wanted, build it, and show evidence that they loved it.

Shyam extrapolated this idea into the proposition that Palantir shouldn’t have engineers building a product in Palo Alto and sales people selling it in Virginia, but instead “forward deploy” engineers to potential clients to adjust the spec on-site and then send reports on the experience back to Palo Alto, where the home engineers could use the feedback to build not an out-of-the-box product, but a bricolage that could be recustomized for each client.

For all this creative froth, which looks inevitable only in hindsight, Palantir was at the time still seen as a small and somewhat demented chickenhawk in the aviary of Silicon Valley. It was constantly running out of money, garnering very little venture capital interest, operating a buggy demo without a real customer in sight, and working out of a Stanford-owned office built into the side of a hill that nearly every early employee who spoke to me for this story described as “moist.” The In-Q-Tel investment, however, landed the startup a few key opportunities that Shyam took charge of.

One was a field trip that In-Q-Tel arranged for a handful of CIA analysts to visit Silicon Valley, where they stopped by Palantir and received the demo. One analyst in particular, Sarah Adams, was smitten with its potential. When she returned to Langley, she set up three meetings for Palantir, inviting only other intelligence analysts and targeters to the meetings, not the CIA’s IT employees, whom she shrewdly suspected of gatekeeping the Agency’s existing software platforms. Shyam gave the demo to over 100 CIA analysts, and then did the same at the Defense Intelligence Agency. His talent was for presenting the software in a cogent and confident way, but also being able to handle stump-the-chump technical questions from the inevitable IT staffer who snuck in. His new work rhythm became flying to Virginia every Sunday night, taking meetings and giving demos Monday through Thursday, flying back to Palo Alto Friday, downloading what he learned to McGrew and Jain for them to code over the weekend, then flying back out to Langley with an improved demo that Sunday.

“Maybe it was the naivete of not understanding how bureaucratic the place was,” he said of Adams, “but she just kind of forced this thing. Talk about the conspiracy of people who change your life along the way. She willed this thing into existence when there wasn’t a contract. Those meetings became the avenue where we could show people this, and then once they saw it, they couldn’t unsee it. They were like, ‘I could be using this, and I’m not?’ It just created overwhelming momentum.”

The next stop was with the newly established Counter-IED Operational Integration Center (COIC), a multi-agency intelligence unit tasked with disrupting the networks manufacturing the improvised explosive devices that were killing and maiming Americans in the War on Terror, which by 2007 were responsible for roughly 70% of U.S. troop deaths in Iraq and 40% of those in Afghanistan. The COIC allowed Shyam and his small team of FDEs to embed with them for two weeks with their buggy product in tow. “The software was so rough,” said Shyam. “We didn’t know how rough it was, to be honest. We went in there like all right, let’s do this! Let’s go catch some terrorists! But we had no idea what we were getting into.”

The FDE team spent the next 14 days, 19 hours a day, in a classified information facility (SCIF), showing the COIC analysts the product, integrating their data into it, showing them how it worked, asking for feedback, coding that feedback back into the product, building custom applications, monkey patching, experimenting, and sending reports back to McGrew and Jain in Palo Alto. Karp remembers that because speakerphones aren’t permitted in SCIFs, Shyam “taped a phone to his head and secured it with an elastic band” so he could have both hands free to code while a COIC analyst spoke feedback into one ear and Jain talked to him from Palo Alto in the other.

At the end of the two weeks, the COIC analysts, improbably, declared the product useful. “But I was so sleep deprived, I got super fucking grumpy and humorless,” Shyam remembered. “Before we left I called Karp and I was like, ‘This is not gonna work. It’s not sustainable. We’re fucked.’”

“And Karp just laughed, like really laughed,” he said. “It was probably 30 seconds, but I remember it like it was 10 minutes. And finally he was like, ‘It’s not gonna work? Shyam, how else is it gonna work?’”

“Because, you know, when I think back on it, Karp had been under so much pressure to hire ex-generals, ex-executives from Booz Allen, people like that to Palantir, not people like me,” he said. “And in that moment he just felt so vindicated. That was the crucible where we proved the thesis—that you can’t get the product to be ready for these environments without being in the environment. And it ended up being proof that this relentless commitment to making products that actually matter could create this wildly positive feedback loop.”

The success at COIC led to the FDE team’s first actual deployment, with the 10th Special Forces Group to Balad Air Base in Iraq in 2010. Each morning, the FDEs would await the return of the soldiers who went outside the wire with Palantir Forward (a field laptop with Gotham installed), get their feedback on what worked and what didn’t, and spend the day coding while the soldiers slept so that it worked better on their next mission later that night. In 2011, they did the same at Camp Leatherneck in Helmand Province, Afghanistan. “When we got to these, the warrant officers were like, ‘Who the fuck are these kids and why do I have to find beds for them?’” Shyam remembered. “But then two weeks into it, the warrants were our protectors. They were like, ‘these guys are gold.’”

“I used [Gotham] when it was brand new in Afghanistan,” I was told by Gen. HR McMaster, who commanded international forces in Afghanistan from 2010–12 and later served as U.S. national security advisor. “When we stood up the task force to combat corruption and organized crime, and we were shifting dramatically our collection and analysis capabilities to understand these criminalized patronage networks and their connections to the insurgency, but also their connection to individuals and government and the degree to which they were fatally weakening the Afghan state, the Palantir guys they seconded to us did a fantastic job. Those [FDEs] were sharp, man. They allowed us to come up with a way to portray these networks, understand nodes in the networks and flows through the networks of people, money, weapons, narcotics, precursor chemicals, not just in Afghanistan but internationally. We just blew out the visibility on these networks with their assistance.”

As a result, Gotham’s reputation started to gain on the U.S. government’s legacy software systems more generally. In somewhat technical terms, what Palantir’s ontology had solved was the “impedance mismatch” between how computers store data (i.e., in rows and columns) and how human beings actually view the world (through objects and relationships). Rather than using spreadsheets, Gotham modeled data as objects to create what Shyam describes as a “pointillist painting” that integrated disparate information into a cohesive picture. In more ordinary terms, what it did was allow an organization to see itself and the outside world more clearly, ask better questions given the available data, and, ideally, come to better, faster decisions.

Karp remembers that because speakerphones aren’t permitted in SCIFs, Shyam “taped a phone to his head and secured it with an elastic band” so he could have both hands free to code while a COIC analyst spoke feedback into one ear and Jain talked to him from Palo Alto in the other.

Over the next several years, Shyam continued to lead the forward deployments, McGrew and Jain kept refining the monolithic platform, and Karp earned his reputation not just as the company’s beloved manager of engineers and product turns, but as “one of the ultimate sales people in the world,” as Ross Fubini, one of Palantir’s earliest investors and advisors put it. The widespread claim that Gotham was used in the 2011 raid that killed Osama bin Laden has never been proven (or disproven), but there is no doubt the rumor helped supercharge Palantir’s business. During the rest of the decade, the company won contracts with the CIA, DIA, NSA, FBI, CDC, NIH, IRS, DHS, ICE, every branch of the military, and municipal police departments.

In 2015, Palantir launched Foundry, the company’s commercial counterpart to Gotham that also operated on the FDE model, and which solved the problem caused by the continual emergence of new data—what Shyam calls “managing the entropy of the universe that drives toward corrupting your data.” Foundry won contracts with JPMorgan Chase, Thomson Reuters, Hershey’s, Airbus, Merck, BP, Morgan Stanley, Ferrari, Fiat Chrysler, PG&E, Credit Suisse, and dozens of others. An early example of its value came from the Airbus factory in Toulouse, France, where the company’s engineers had spent two years unable to diagnose a recurring A380 fuel pump fault. Palantir FDEs organized their sensor data into Foundry and within two weeks, they found the culprit: fuel sloshing away from the pump on ascent, a trivial fix that saved Airbus $40 billion in orders. 

In 2016, Palantir marked Shyam’s first decade with the company by suing the U.S. Army over its plan to buy the next version of its main intelligence-data software system. Palantir argued that the army violated the Federal Acquisition Streamlining Act by pursuing costly custom development without considering commercial off-the-shelf systems like Gotham, which field commanders had repeatedly requested. The Court of Federal Claims eventually ruled in Palantir’s favor, a decision the Federal Circuit later affirmed. The case forced the Army to reconsider its acquisitions process, and set the legal precedent that arguably gave birth to the modern “defense tech” sector, which in turn is populated by many of Shyam’s former FDEs.

In the meantime, Palantir as a whole, which began as a mix between the brainchild of Thiel, the creation of Cohen, and the cult of Karp, had started more and more to become the culture of Shyam, who’d slipped off the chain of a boy scout personality. Teams of engineers working on a new product got used to him stopping by to use it, build a demo, break it, and write a report on what was wrong with it. Ten-thousand-word critiques of products or ideas, which came to be known as “Shyam Bombs,” would get emailed to every person in the company in the middle of the night, on weekend afternoons, or when everyone thought he was at a dinner he’d apparently decided to skip. “Fatwas” were new Shyam-issued rules or ideas followed by an invitation to argue with him about it with the whole company cc’d. To prove the “flatness” of the organization, he began a practice of prompting junior hires to tell him, with his face inches away, to go fuck himself. “Save the Shire,” “Metabolize pain, excrete product,” “There are no silver bullets, just millions of lead ones,” and other enduring Palantir memes began as Shyam-isms. He had each of the company’s lawyers officially retitled “Legal Ninja.” He started wearing the blazer with the hoodie sewn in.

“Shyam’s personality is very much built into the culture of the brutality with which we look at problems, which is what made the FDE structure actually scalable,” Karp told me. “His ability to understand what a product looks like, the way in which he intuits, and just as importantly anti-intuits what’s working and what’s not, and being very honest about it—I’ve seen people really struggle with that, but it’s just not something Shyam struggles with.”

“Shyam invented the term Forward Deployed Engineer and all the nomenclature that goes with it, but also the principles that outlined what it meant to be one,” Jain said. “He just poured his heart and soul into what users were trying to do in the field to identify what we actually needed to build to win, and then worked closely with Bob and me to figure out how to turn that into bigger product movements. And then all the other byproducts of that, like the real radical transparency in our culture, the spirit of building from the front line back to the product, the freedom for junior engineers and senior leadership to get in substantive fights with each other in front of everyone. Those were all things Shyam developed.”

“If you don’t know him, it seems like he’s being a little brash or aggressive,” said Ted Mabrey. “But the reason he can do it is because it’s not personal at all. He completely depersonalizes it.”

“I remember one time, we were sitting in the Toronto airport together working on our laptops at a coffee shop, having a totally pleasant conversation, he was in a great mood,” Mabrey recalled. “And then all of a sudden I got a Shyam Bomb in my inbox. He’d been writing it while we were sitting there. It was honestly an interesting moment because it was like, Oh, I can see he’s not angry at all. He just thinks, ‘This is what I owe you all in order for us to win together.’”

Although it still hadn’t turned a profit and had lost almost $4 billion since its inception, by the end of the decade, Palantir was approaching $750 million in annual revenue. It had around 125 customers, nearly half of which were commercial. Sixty percent of its customers were international (though the company has refused to ever do business with China, Russia, or other U.S. adversaries). It had around 15 offices globally. After Google withdrew under pressure from employee protest, the Pentagon awarded its $800 million contract for Project Maven—an initiative to develop AI image analysis for military drone operations—instead to Palantir.

In February 2020, Shyam’s father was diagnosed with stage IV esophageal cancer. He and Girija had been living with Shyam in Palo Alto, but as head of the FDE teams, their son had barely been home for most of the previous 14 years. With his father’s diagnosis and the world shut down during COVID, Shyam moved with his parents to Texas for treatment at MD Anderson; during Palantir’s IPO in September 2020, he was living in lockdown with his dying father in a fourth-floor townhome in downtown Houston.

“My family was very proud, but most of all Dad was proud,” he said. “I’m very glad that he got to live to see the company go public. It was immensely satisfying that I got to share that with him.”

When Shan Sankar died in February 2021, Shyam had been named chief operating officer of Palantir. Its market cap was $60 billion.

Not everything in the business was a bed of elanors and niphredils. Palantir’s own 2020 IPO filing noted, if somewhat proudly, that “Activist criticism of our relationships with customers could potentially engender dissatisfaction among potential and existing customers, investors, and employees with how we address political and social concerns in our business activities.”

By then, those political and social concerns had included a lawsuit against Palantir by i2 that, among other things, accused Shyam personally of conducting “industrial espionage” (i2 later settled out of court for $10 million). It also included a 2011 scandal revealing Palantir’s complicity in a plan by Bank of America’s law firm to discredit WikiLeaks and target the journalist Glenn Greenwald (Karp subsequently issued a public apology and ended the contract); a 2013 scandal involving JPMorgan Chase, whose security chief had used Palantir software to surveil bank employees; Palantir’s contracts with ICE (which began during the Obama administration’s large-scale deportations of illegal immigrants, but only garnered greater public protest in Donald Trump’s first term); accusations that the company’s software was facilitating “predictive policing” for the New Orleans and Los Angeles police departments; and more general dismay that the company, whose founder Thiel was an early and prominent supporter of Trump, had helped make real the U.S. intelligence community’s supposedly long-held dream of panoptical and unconstitutional surveillance capabilities via a private-sector cutout. Starting in 2018 and continuing to this day, activist groups began protesting outside of Palantir offices on a near-daily basis.

The company’s IPO had also made its apparently brittle business model a matter of public interest. Investors had long worried that Palantir was overly reliant on government contracts, which are subject not only to “political and social concerns” but also to unpredictable appropriations cycles, fickle Congressional resolutions and omnibus awards, bid protests, and lawsuits. In turn, that made Palantir’s story of compounding commercial growth particularly salient, such as when it reported year-over-year U.S. commercial revenue growth of 136 percent in the first quarter of 2022, and also particularly prone to panic, such as when it reported a collapse to just 8–12 percent by the fourth quarter of the same year. Palantir was also accused of “buying its own demand” by taking equity stakes in potential customers, losing hundreds of millions of dollars on those investments, and relying heavily on stock-based compensation for employees, diluting the shares of investors. And by the end of 2022, after nearly two decades, Palantir still had never turned a profit. Complaints about its capital structure were repeatedly rebuffed by the reality that Palantir’s founders—which included Thiel, after all—retained total voting control. 

Salvation of a sort arrived, as it did for many, with the abrupt appearance at that very moment of Generative AI. It was perhaps propitious that Bob McGrew, Palantir’s head of engineering in its first decade, had spent the previous four years as chief research officer of OpenAI.

Its more contemporary claims to have always been an artificial intelligence company of sorts notwithstanding, Palantir had long resisted repeated government and commercial inquiries about “AI,” and for understandable reasons. Clients had always been hoping for something more like Minority Report, in which magical software could autonomously identify all the terrorists and business opportunities, eliminating the need for boot-leather detective work. Palantir was a set of software tools, not a silver bullet, they constantly reminded clients; human beings would always be required to do the analysis and make the decisions.

When McGrew happened to visit Palantir earlier in 2022, things changed a bit. As McGrew explained to Karp and Cohen what he was working on, they were interested but dubious that this was really happening the way McGrew said it might, which was that something real was in fact happening, and that the models would only get better, not worse. When ChatGPT debuted in November 2022, they came to the same realization everyone else did. After GPT-4 debuted in March 2023, Shyam made a call.

Every company, of course, was having a similar idea. Shyam’s insight was that while OpenAI built powerful models, it would have little interest in enterprise data integration, which he believed was the thing that would make its models operationally useful. Palantir’s existing platforms had spent almost 20 years integrating its clients’ data and making it accessible. At the beginning of 2023, Shyam and Karp put the majority of Palantir’s engineers on the construction of an AI layer for Gotham and Foundry, and wheeled the company onto what would become its Artificial Intelligence Platform (AIP), developing it in a partnership between the FDEs and their customers.

“The thing I most respect about Shyam is his willingness to pivot everything when he realizes it’s needed,” McGrew said. “There are multiple times in the history of Palantir when we pivoted the whole company because there was some unique insight, and we thought we could make it happen, and we would put half or two-thirds of the company on some new problem, which is an extremely risky thing. That was what enabled us to get deployed to Afghanistan, for example.  And they did this again with AIP. That was the difference between Palantir at $10 and Palantir at $160 or whatever. Just a hugely important thing for the company to do. And that was Shyam’s leadership.”

“Our view is that all the economic value accrues to chips and ontology,” Shyam explained. “The ontology is a semantic representation of the enterprise. It’s the common thread through Gotham and Foundry and it’s why those platforms can interoperate. Foundation models understand semantics, and Palantir’s ontology becomes a set of tools for those models to use: You can ask them to read cables from certain sources, do some analytical work, write the results back. The ontology becomes the tool through which you communicate and coordinate these things. If you decompose a complex process like insurance underwriting into discrete steps, where each one is mediated by writing state and progress to the ontology, the models become much more reliable. It’s a way of getting the stochastic genie into a straitjacket.”

We were sitting in his office in New York, and had only a few minutes left until I was to receive demos of what he was talking about. Apparently alarmed by the look on my face, he asked if I felt like I understood what he was saying. I said I felt a bit like a member of the HBO Max social media team, which during the most recent of several rebrands had to explain that while HBO is a premium cable network Max is a streaming service that houses content from many brands including HBO and that all HBO series stream on Max but Max series do not air live on HBO and both are owned by Warner Bros. Discovery which plans to rebrand Max back to HBO Max. Not really, in other words.

He blinked a few times and smiled. “Think about the stack of the modern AI economy: power, data centers, chips, frontier models, AI infrastructure, AI applications,” he clarified. “Where does value accrue? The model companies have trained these multibillion-dollar models that are essentially commoditized the second they’re released, partly because China is fucking with us, but also because it’s a race to the bottom. They recognized long ago that to monetize this miracle they’d have to build software, to move up the stack toward AI infrastructure and applications. Meanwhile, all these AI application companies—AI for accountants, AI for lawyers, AI for this or that—run headfirst into the problems we’ve already solved with AIP. They end up realizing they need AI infrastructure, and now they’re on a 10-year journey to build something we already built. Both ends of the market are converging on the same realization. The valuable part is the AI infrastructure.”

Palantir, in other words, exists to solve the mismatch between how computer systems store data and how human beings think about things. Every large enterprise—whether the U.S. Army, an oil company, or an airline—has hundreds of different databases and software systems that don’t talk to each other. By modeling everything as an ontology, “hydrating” that ontology by integrating data from all the previously siloed sources, and allowing users to ask questions of it, Gotham mirrored how people actually see the world. Foundry solved the problem caused by the continual emergence of new data—“the entropy of the universe driving toward corruption”—by catching problems before they cascade. The Apollo system is what automates the deployment of the software’s management; for example, by allowing an FDE sitting in a rental car with a laptop to write code and ship it to a classified network within 15 minutes. AIP integrates LLMs with the ontology to automate workflows and make it easier for human beings in positions of commercial and government leadership to make better decisions faster.

That, at any rate, is what the market by November 2025 valued at $420 billion. It is also what people like Michael Burry would call bullshit.

On the morning of the Herman Kahn Award gala, I was welcomed into a large meeting room in Palantir’s New York office by four young people who introduced themselves as team members of Hospital Operations, Warp Speed, Maven, and Hivemind, respectively. Furnished by Shyam with another nicotine pouch and can of Celsius, I sat at a long conference table and received demos of the company’s products, feeling determined not to be impressed.

During the first demo, I watched Foundry take different data pipelines like electronic medical records, staff schedules, and lab orders from Tampa General Hospital, and map them onto Palantir’s ontology, which linked individual patients to their doctor’s visits, notes, medicines, orders, and assigned staff—data that is traditionally recorded, I was told, by nurses jotting it down on whiteboards with erasable markers. One feature showed LLM agents continuously comparing each patient’s chart against temperature thresholds, respiratory rates, signs of organ malfunction, and other criteria for sepsis; when a patient triggers the criteria, the system alerts the hospital’s rapid response team. (According to Tampa General, Foundry saved several hundred patients from septic shock during the COVID Delta wave of 2021, when Palantir also reportedly played a role in helping the U.S. and UK governments distribute vaccines.)

The second demo, of Warp Speed, was described as Shyam’s “big pet project” to reinvigorate America’s defense manufacturing base. In a user interface more befitting industrial rather than hospital aesthetics, a notional shipbuilding scenario showed the system tracing a problem backward: When it noticed that weld inspections had started failing, or when the vision model detected a defect, the ontology connected the finding to a new welding tool recently in use, and a part that was recently redesigned. When the user confirms that a fix is needed, the system alerts the factory floor to stop work on that particular area, automatically notifies the supply system to reorder parts, and generates a cost and timeline impact. Traditionally, I was told, all that work is done by “bills of materials 30 layers deep, two-year lead times for raw materials, a design engineer, mechanical engineer, manufacturing engineer, and supply planner who all hate each other, and who all maintain their own documents that fall out of sync.”

With the nicotine and Celsius wearing off and the borborygmus coming on, I was relieved when the next presenter, a former submariner, began the demo of what I was told I could refer to only as “a defense system.” A continuously updating map of a notional battlespace in Ukraine showed Russian and Ukrainian unit positions, orders of battle, trench lines, and 48-hour fire patterns. Each of these, I was told, is continuously pulled from human and electronic intelligence, overhead sensor detection of flashes on the ground, satellite imagery, and computer vision detection of grid coordinates gleaned from social media posts. When asked a ridiculous question like “How many U.S. Army Black Hawk pilots speak fluent Hungarian and are available to deploy to Korea tomorrow?”, the system generates an immediate answer. When asked to generate a strike package, the system’s AI layer does so in about 30 seconds, including by listing the available assets in each service branch, calculating time-on-station, and coordinating air defense suppression with the ensuing bomber waves. I was ready when the former submariner told me that the human design of such strike packages, even in the most urgent scenarios, traditionally takes hours, not seconds.

Hivemind, the final demo, had not existed six weeks prior. Like with any chatbot, the user types in a simple prompt; we asked it for a proposal to modernize the Federal Aviation Administration. The system then “spawned” multiple LLM agents to simultaneously research the problem, think of solutions, critique those solutions, and produce a proposal. Then more “critique agents,” each running code (one wrote a Python simulation modeling passenger behavior), refined the proposal’s operations, budget, security, safety, and directions for human implementation. I was unmoved by what looked like a simple aggregation of existing chatbots until Hivemind was asked to generate a plan to take out [a well-known longstanding overseas threat to U.S. interests] circa May 2025. While I can’t vouch for the claim (nor for Hivemind’s FAA reforms), I was told that the system, without being fed any classified information, produced in less than two hours a plan that precisely matched [a high-profile U.S. military strike on that overseas target which took place in reality shortly thereafter]. Reader, need I tell you how long it took the poor human planners to come up with their strike plan?

Following the demos, I attended a lunchtime AMA Shyam held for dozens of interns, which he concluded by leading them all in a chant telling him to go fuck himself. After the interns scurried back to their hovels, I followed Shyam back to his office, where he asked what I thought of the demos. I said that Palantir sees too much but not enough to justify its valuation.

“The whole, ‘You must be a services company masquerading as a product company’ was actually very protective for us for years,” he said unspooling in his chair, appreciating the joke I’d tried on other Palantirians, mostly to blank stares. “I think that criticism probably doesn’t need me to eviscerate it now because it’s been eviscerated. Now the valuation, look. Which companies have had 10 quarters of accelerating growth off of a meaningful base? There’s a gravity-defying element to that. By conventional metrics maybe it seems richly valued, but who else has a positive second derivative at this sort of scale right now? And what does that mean about the market? Again, it’s our thesis that the value in AI accretes at chips and ontology.”

He then snapped back with a straight spine and an earnest look on his face. I asked if he was thinking of something like the recent news that ICE had reported a detention population of over 60,000 people, a 75% increase since Trump’s second inauguration and the highest number in its history, and plans to hire thousands of “skip tracers” to verify the addresses of up to 1.5 million illegal immigrants.

“The most misinformed take on Palantir is that we should not be building the panopticon or like a Minority Report surveillance state,” he said. “I agree with that! It’s just rooted in a deep misunderstanding about Palantir. We don’t collect any data, we don’t harvest data. We’re a software company. We’ve built classification-based access controls, roles-based access controls, attribute-level access controls into the foundation of our infrastructure. That’s how we establish this zero-trust environment where we can essentially guarantee that the right people, and only the right people, have the right access at the right time to make specific decisions.”

I asked whether these tools should nevertheless not exist, knowing that human nature is forked, politics are fallen, and people in positions of power over others will find ways to abuse it, regardless of technical controls. “That’s just crazy to me,” he replied. “We should expect our institutions to operate really well. They serve the citizen at the end of the day, and you know, it’s not okay that doors fall off planes. It’s not okay that Intel was almost run into the ground because we cared more about financial engineering than actual engineering. These things speak to our national security and prosperity. Why would we indulge in this nihilistic managerialism that’s just ruining everything? And then extend that to government: Are the American people sure that the military we have is the military we need and expect? Of course these things need to work. Now, you can and should change the rules by which they work. But that’s not the job of unelected engineers in Silicon Valley. That’s the job of democracy.”

“What people don’t understand is that the argument that, you know, ‘Maybe our institutions are too powerful and we should be throwing sand in the gears, not making them more effective’—that just has deeply corrosive effects,” he said, breaking the facade of childlike optimism for the only time during our interviews, growing slightly indignant as he reached for the same words he’d used to describe what nearly swallowed his family when he was a boy. “It’s a form of institutionalizing corruption. Like, why do you think nothing works in Nigeria or India? It’s because the cultures are corrupt. The systems are bankrupt, and people just live with it or get overwhelmed by it. The entropy of the universe is towards that sort of corruption.”

“America’s always been able to push back against that and basically have things that function,” he concluded, switching registers and settling back into his chair. “It waxes and wanes throughout history. But we’re in a period of really fighting for that now.”

When asked why I was interested in writing about technologists in general and him in particular, I’d told Shyam that “technologist” is the category of public figure I feel a sense of intimacy with, and whose psyche I can understand to some extent through the public record, whereas what goes on in the mind of Timothée Chalamet or Stephen Miller is a total mystery to me. Yet with him, I’d noticed, this hadn’t quite been the case—which is why I’d wanted to talk.

Over the previous year, I’d watched from afar as Shyam transitioned from the most pivotal but invisible figure behind the country’s most mysterious company into something of a public voice. He’d published a treatise, “The Defense Reformation,” in which he argued for leveraging venture-backed defense startups to break the Pentagon’s defense monopsony and rebuild the American industrial base before it’s too late to deter a great power war. He’d become a regular on Fox and CNBC, where he discusses AI, warns that World War Three has already begun, and defends the vagaries of Palantir’s share price in a suit and tie. He wrote for Pirate Wires on why he and his comrades aligned behind Donald Trump, and for The Free Press on his commission as a lieutenant colonel in the U.S. Army’s new Executive Innovation Corps, which he compared to the industrial giants deputized by Franklin D. Roosevelt to build the Arsenal of Democracy. He’d become a regular at The White House, and increasingly on podcasts. He announced a book, which elaborates on the treatise. There is a Substack, which elaborates on the book. He’d started a movie studio. If for two decades he’d been a private man with a controversial but quiet mission, Shyam had suddenly become a torchbearer with a message.

As the most coherent explainer of Palantir and the reformation it begat, Shyam’s decision to become an evangelist for his cause is as understandable as his service is honorable. His message, moreover, is a serious one: that decaying federal institutions have presided over a 30-year schism between American commercial enterprise and national defense, and allowed the country’s defense capacity to corrode to the point where we can no longer deter or fight a civilization-threatening war. Yet there was also, I felt, a distinct sense of unease in Shyam’s public turn.

Part of the reason is structural. The contemporary media landscape Shyam has embraced with unexpected eagerness—of podcasts, of TV news, of social and “new” media—has so thoroughly absorbed the pose of sincerity, or what technologists and venture capitalists like to call “conviction,” that the thing itself and the knowing performance of it have become practically indistinguishable. Each is rewarded with ever more attention, and ever more capital, as if they were the same, and as if the very real problem of deteriorating U.S. military deterrence were as urgent and true as the “question” of whether aliens built the pyramids. The tension in Shyam’s newfound participation in these media is not any insincerity, but a certain self-awareness, almost palpable through the screen, of the trap it lays—of a concern that we won’t be able to distinguish between genuine warnings about the entropy of the universe and a rhetorically sophisticated strategy for business development; an awareness of the reason that prophets have not historically been chief executives of publicly traded companies.

Such is the conundrum for any Cassandra with that kind of day job. The reason it matters in Shyam’s case, however, is that his message is important, and he does actually believe it. That much is clear from Mobilize: How to Reboot the American Industrial Base and Stop World War III, his forthcoming book, which argues again for empowering “heretics” and “founders” working in government and venture-backed technology to restore American deterrence, and the industrial overmatch required to fight and win if cold war turns hot.

Critics may read the book in part as a veiled retort to the bears convinced that Palantir is a galactic bubble, or to those who believe that the $130 billion worth of venture capital poured into “defense tech” more generally over the last few years will end in a giant sucking sound. They might also seize on what Mobilize leaves out: that peacetime mobilizations are as much about industrial capacity as they are about the kind of social bargaining, political coalition building, diplomatic finesse, and institutional endurance for which “heretics” are ill-suited. Such mobilizations have also, historically, led to deterrence or victory as often as they’ve succumbed to an internal logic of timetables, momentum, and self-justification that made the wars they were meant to prevent more likely, and, once begun, harder to stop.

Yet the book is strongest, and most revealing, in its profiles of the heretics Shyam offers as proof of his thesis. To make its point about “the primacy of people,” Mobilize tells the stories of Colonel Drew Cukor, the Marine intelligence officer who bypassed a 17-year acquisition cycle to launch Project Maven from a basement office; Admiral Hyman Rickover, who created the mission-obsessed culture that launched the first nuclear submarine in under seven years; General Bernard Schriever, who sidestepped “nitpicking” Pentagon officials to develop the Atlas, Minuteman, and Titan intercontinental ballistic missiles in five-year sprints; Andrew Higgins, the Louisiana boatbuilder who spent years fighting the U.S. Navy for the right to compete, and whose landing craft eventually comprised over 90 percent of the fleet at Normandy; and Bill Knudsen, the General Motors president who became Roosevelt’s “Production Czar” and made the Arsenal of Democracy a reality. Kelly Johnson, Wernher von Braun, and John Boyd all come in for similar treatment, as do more contemporary figures like Palmer Luckey. Against the entropy of the universe, Shyam seems to be saying throughout the book, the heretics are those who have held its corruption at bay.

It is the point he circled repeatedly in our interviews but never quite made, and which he seems to want to make as much to himself as to the world. What we’re left to consider, of course, is whether a place in this pantheon will be reserved for him.

“Please don’t let my tombstone just read ‘defense tech,’” he’d said as we sat in his office, Manhattan bombinating and sheeting rain outside. “‘American Greatness’ is the thing I care about. It’s the primacy of winning, defeating nihilism and cynicism. You can’t achieve American Greatness through zero-sum divisiveness. There has to be a unifying vision. There has to be something positive-sum. We’ve kind of ceded that narrative formation entirely. It’s not possible for any civilization to do great things if it’s not proud of itself. That’s kind of a root-cause issue right now; we are confused about whether we are a noble nation. That’s what led me to something like starting a movie studio around the American cinematic universe.”

“Saving Private Ryan, Schindler’s List,” he said, when asked for some of his favorite films. “And then the ones I grew up watching with my Dad. Rocky IV and Rambo III, The Hunt for Red October, you know, Red Dawn. The great Westerns.” He paused to ruminate before repeating a thought from one of our earlier conversations. “The cowboy as the kind of quintessential archetype of America,” he said. “Life is hard. It is what you make of it.”

When he was at Trinity Prep, which was 40 percent Jewish, Shyam had been through the whole bar and bat mitzvah circuit by the time they screened Schindler’s List for his ninth-grade class. “I remember after we watched it,” he said, “I told myself, like, if the world came to this, I was going to be the Righteous Among the Nations. I was going to be Schindler. And I grew up with this sort of protective sensibility around Jews. So now I’m kind of double-fucked, right? I’m Indian, and I also study Torah every week.”

“You what?”

“It hasn’t been that long,” he said, “I just started last year. I read the parsha every Sunday with Rabbi [Adam] Mintz, he’s this well-known Orthodox rabbi on the Upper West Side.”

When asked if he ever discusses such things with Karp, he replied: “Even back in the early days, I realized that a lot of his most genius business insights came from these like ancient jokes, which of course are more wisdom than joke. There was this one time early on where I was freaking out at him, like, ‘How did you sign us up for this work? I don’t know how to fucking do it. This shit is not possible.’ And he was just like, ‘Let me tell you a story.’”

“Go on,” I said.

“So there’s a bunch of Jews living in a goy kingdom,” Shyam regaled, channeling Karp. “And one day the goy king summons the rabbi. ‘Look,’ he says. ‘I have good news and bad news for you, rabbi. The bad news is I’m gonna kill you and all your people. The good news is if you can get this monkey perched next to me to talk, I’ll spare you all.’ The king is mocking him, obviously. But the rabbi says, ‘I can do it.’ And the king is like, ‘What?’ And the rabbi says, ‘But such things are not possible immediately. It will take me five years.’ And the king says ‘Alright rabbi, if you can get my monkey to talk in five years, I’ll spare you. And if you don’t, I’ll kill you all.’”

“So the rabbi goes back to his helpers, and they’re like, ‘What the fuck? How are you gonna get the monkey to talk? Monkeys don’t talk.’ And the rabbi says: ‘Guys, a lot can happen in five years. The king could die. The monkey might die. And who knows? Maybe he’ll learn to talk.’”

I was thinking of that joke when I got back to my hotel after the Herman Kahn Award gala, and saw in my inbox some attachments regarding Shyam’s new movie studio. I expected to find a series of flag-humping movies at various stages of production, and wasn’t disappointed. There were movies about an ex-commando liberating a forced labor mine in Kenya; the operation to assassinate Qassem Soleimani; a defector from Chinese intelligence; a mission to rescue the first American-trained Afghan commando; a biopic of Hyman Rickover; a special forces soldier exfiltrating civilians from Abbey Gate; and the improvised civilian rescue operation at Ground Zero, the largest maritime evacuation in history, that began within minutes of the towers being hit.

I wondered if, beyond the obvious, there was a deeper connection between such ideas and the movies Shyam grew up watching with Shan, acting out the dialogue in the car every day from Kissimmee to Trinity Prep, Florida swampland and stripmalls rifling by in the window. There is, of course. They are tales of men who are incorruptible. Some are make-believe. Others are based on true stories.

Jeremy Stern is the editor-in-chief of Colossus.