Philosopher in the Valley
Alex Karp, Palantir, and the Rise of the Surveillance State
Journalist Michael Steinberger's biography of Palantir CEO Alex Karp — from philosophy to Silicon Valley.
Palantir spent two decades being dismissed as overpriced surveillance technology with more government contracts than customers who'd admit to using it. Then AI happened, its stock went vertical, and suddenly everyone wanted to understand the company. Steinberger's *The Philosopher in the Valley* arrives at exactly the right moment, and almost gets the book right.
The central subject is Alex Karp, who remains one of the stranger figures in tech. A philosophy PhD who wrote his dissertation on aggression in social life, severely dyslexic, former progressive who's shifted steadily toward the MAGA orbit while maintaining what sounds like a genuine philosophical unease about some of what Palantir does. Karp co-founded Palantir in 2003 with Peter Thiel, and in a pattern Steinberger traces well, the two men have diverged in almost every direction since. Where Thiel's politics were always readable, Karp's shift from neo-socialist to Trump supporter came later and seems more tortured — driven by a conviction that woke ideology represents a real civilizational threat and that Palantir's work defending the West is not something to apologize for. Whether you find that convincing depends on priors Steinberger can't resolve for you.
One of the axioms of this data-soaked age is that data can lead to better decision making in almost every sphere of human activity, from the operating room to the assembly line, to the battlefield to the baseball diamond.
— Steinberger, *Philosopher in the Valley*
What Palantir actually does is harder to explain than the company's critics acknowledge and easier than its boosters admit. The core product integrates data — messy, siloed, incompatible data from dozens of sources — into something an analyst can actually use. The intelligence failure that led to 9/11 was, at its root, a data integration failure: agencies had pieces of the picture and no way to fit them together. Palantir built the software to fit them together. From that starting point, the platform expanded into defense logistics, healthcare, financial crime, immigration enforcement, and now AI. The AI Platform they launched post-2022 is essentially Palantir doing what it always did — integrating data, surfacing patterns — but with large language models in the stack. For an organization drowning in unstructured data (emails, video, documents, chat logs), that's a genuine capability leap. Palantir's trajectory from 9/11 counterterrorism to a $400 billion company is not luck; it's one bet placed correctly in 2003 and extended through every subsequent wave.
Palantirians took pride in being different. Instead of making gadgets and games, they believed they were on the front line of a battle to preserve America's way of life.
— Steinberger, *Philosopher in the Valley*
Steinberger had real access to Karp and uses it well in stretches. The biographical material is genuinely interesting: the biracial Jewish background, the vulnerability it produced, how that shaped Karp's conviction that Palantir's defense work is morally serious rather than just profitable. Where the book runs thinner is on the technology itself and on the "surveillance state" claim in the subtitle — the harder questions about what happens when a private company holds this much data infrastructure for governments get raised but not answered. Several reviewers noted that Steinberger's own political leanings shade the framing in ways that feel like editorializing; that's a fair criticism, and readers across the political spectrum landed on it from both directions.
My biggest fear is fascism.
— Karp, quoted in Steinberger, *Philosopher in the Valley*
For anyone who thinks Palantir is either pure defense contractor or pure surveillance menace, this book will complicate the picture. It won't satisfy people looking for a takedown or a celebration. What it delivers is a solid portrait of a company whose business is genuinely consequential and whose CEO is genuinely strange. That's worth 300 pages.
Read the longer summary
What the book is actually arguing
Steinberger has set himself a tricky double task: write a biography of an unusual CEO, and through him explain how a single software company ended up plugged into the most sensitive parts of the American national-security apparatus and large chunks of the commercial economy along the way. The thesis, when you flatten it, is that Palantir is the company that built the data plumbing for the post-9/11 world, that Karp is the unlikely person who steered it there, and that what comes next — AI everywhere, in government and on the battlefield — runs on infrastructure that looks a lot like what Palantir has been quietly shipping for twenty years.
Karp matters in this story because he is genuinely odd. The book’s title isn’t a stretch. He has a Stanford law degree but never practiced. He has a PhD in social theory from Goethe University in Frankfurt, where he wrote a dissertation on aggression. He’s biracial — Jewish father, Black mother — severely dyslexic, has no formal training in either business or computer science, and was openly progressive for most of his adult life before flipping into vocal Trump support. He owns somewhere around ten houses, most of them clustered near cross-country ski trails, and skis 12 to 15 miles a day when he can. He calls himself, in public, a “batshit crazy CEO.” This is the person Peter Thiel handed Palantir to in 2003.
How a CIA experiment became a $400 billion infrastructure company
The origin story Steinberger tells is cleaner than most Silicon Valley founding myths because it’s narrow. After 9/11, Thiel had an idea: the antifraud system PayPal had built to spot stolen credit cards might be retooled to spot terrorist activity buried in intelligence data. The bet was that the September 11 intelligence failure had less to do with missing information and more to do with information that was sitting in the wrong databases, never joined together, never queried in the right way. Palantir was founded in 2003 around that bet, with the CIA as an early backer through its venture arm. Karp came on as CEO — a Stanford classmate of Thiel’s, a friend, and a person with no obvious qualifications for the job other than that Thiel trusted him.
What the company built — Gotham for intelligence and defense, Foundry for commercial customers, more recently AIP for the generative-AI era — is best described not as analytics in the dashboard sense but as data integration with operational hooks. Take a customer with files in twenty places, no shared schema, classified material mixed with unclassified mixed with raw chat, and turn it into a queryable model where an analyst can pull a thread and act on it. Dashboards matter; the workflow that wraps the dashboard matters more. For an analyst trying to track a supply convoy out of Kabul, this is closer to the operating system of the mission than to a BI tool.
The customer list Steinberger walks through is genuinely striking. Every branch of the US military. The CIA, FBI, IRS, SEC, and Department of Homeland Security. The Mossad. The UK’s NHS, where Palantir software helped run the COVID vaccine rollout. Twelve national governments using the technology for pandemic response. Corporate buyers including Airbus, BP, JP Morgan, Bridgewater, Home Depot, and Thomson Reuters. Use in the operation that killed Osama bin Laden. Use in Ukraine’s defense against Russia. Use in Israeli military operations including hostage extractions and the strikes on Hezbollah’s senior leadership in Lebanon. Use by US Immigration and Customs Enforcement to build dossiers on individuals — pulling in healthcare records, driver’s license data, mobile-tower pings — and to assign each person a “confidence score” predicting whether they’ll be home when ICE comes for them.
That last one is where the book gets uncomfortable, and where it should. A federal judge in California, Steinberger reports, last year issued a temporary injunction stopping HHS from feeding ICE the personal data of Medicaid enrollees. Palantir denies direct involvement in Israel’s “Lavender” targeting system, but its software was in the loop on raids, missile defense, and the Lebanon strikes. The company’s roster is not a clean story about software for good. It’s software at the seams of state power, full stop, and Karp seems to like it that way.
The Karp pivot
Half of what makes Karp interesting in this book is that he is, by his own description, a child of the European left who has ended up running the American security state’s preferred software vendor and openly cheering for Donald Trump. Steinberger spends a lot of pages trying to explain how that happened, with mixed results.
The cleanest version of the story Karp tells about himself is identity-driven: as a Jewish man, the question of where Jews are safe is structural, not academic, and after October 7, 2023, he no longer trusts a Democratic Party that he sees as accommodating anti-Israel sentiment. The book’s broader read is more deflationary. Karp got rich, then very rich, then absurdly rich — $1.1 billion in compensation in 2020 alone, the highest of any public-company CEO that year — and at that altitude you stop spending time with anyone who isn’t also a billionaire. The peer group does the work the politics used to do.
The Habermas detail is small and telling. Karp claims Jürgen Habermas — yes, that Habermas — advised his dissertation in Frankfurt. Multiple sources Steinberger talked to say Habermas declined. It isn’t a defining flaw, but it sits awkwardly next to Karp’s persistent self-presentation as a serious philosopher whose ethical engagement separates him from the rest of Silicon Valley. If you’re going to lead with the doctorate, the doctorate has to hold up.
What Steinberger does well here is sketch a person made of contradictions without trying too hard to resolve them. Karp critiques wokeness for breaking meritocracy, then leverages his own identity in the standard ways when it suits him. He rails against remote work, then has employees travel to wherever he happens to be skiing. He inveighs against tech-billionaire excess while owning the ten houses. He talks like a moralist and operates like a CEO whose first job is the share price. None of this is news to anyone who has watched a public company up close, but the texture of it — the specific contradictions of this specific person — is the book’s strongest material.
Where the book is sharpest
When Steinberger sticks to the company and the man, he’s good. The chapter material on Palantir’s early scramble to figure out what intelligence analysts actually needed — Thiel and his cofounders, by Steinberger’s account, had no real grasp of the workflow they were claiming to fix — is funny and honest about how much luck and good engineering papered over a missing strategy. The Ukraine and Afghanistan-evacuation passages give you a real sense of what happens when the data exists but lives in twenty places that don’t talk to each other, and why a company that solves that specific problem is going to print money.
The portrait of Karp is the book’s other strength. Steinberger had unusual access — many hours with Karp, plus interviews with current and former Palantir employees — and he uses it. We get the cross-country skiing, the houses, the chaotic management style that employees describe as half intelligence and half no plan, the cultivated outsider posture that the company has wrapped around itself as a brand. Steinberger describes Palantir as a company of “perennial outsiders, always the barbarians at the gate” — and the line works because the company has worked hard to make that frame stick, even as it has become arguably the most insider tech vendor in Washington.
Where the book is weakest
The subtitle is a check the book doesn’t fully cash. The Philosopher in the Valley: Alex Karp, Palantir, and the Rise of the Surveillance State promises three things, and the third — the surveillance state itself, what it is, who it watches, what changes when its software vendor is publicly traded — gets noticeably less attention than the other two. Steinberger names the controversies. He doesn’t dig into them. ICE, Lavender, Cambridge Analytica, the Medicaid-to-immigration-enforcement pipeline: each could carry a chapter. Each gets paragraphs.
The technology coverage is similarly thin. For a book about a software company, you finish it without a clear mental model of what Gotham and Foundry actually do, how AIP differs from a generic LLM wrapper, or what’s distinctive about Palantir’s approach to entity resolution and ontology. Readers who already work with data will fill the gap themselves. Readers who don’t will leave with a vague sense that the software is powerful and not much more.
The third weakness is the one that has dominated the book’s reception: Steinberger’s own voice keeps wandering into the frame. The author is a Haverford classmate of Karp’s, which gives him access; he’s also a writer whose default register is liberal-magazine commentary, and that register doesn’t always serve the material. A biographer’s job is not to disappear, but it is to keep the spotlight on the subject. In a few chapters — particularly toward the end — Steinberger’s own takes on Trump and the contemporary right start to crowd Karp off the stage. Whatever you think of those takes, they aren’t what readers came for.
This isn’t a fringe complaint. Reader reviews fall into a specific pattern: right-leaning readers say the politics ruined a good biography, left-leaning readers say Steinberger pulls his punches on Israel and on the company’s role in immigration enforcement, and the people in the middle say a tighter edit could have cut a third of the book. They’re all describing the same problem from different angles: a book that promised a portrait of a complicated CEO and a hard look at his company sometimes settles for being a column.
What’s actually missing
A few questions a more disciplined version of this book would have answered.
What is Palantir’s moat? Steinberger gestures at the forward-deployed-engineer model — sending real engineers to live inside customer organizations until the software fits the workflow — but doesn’t pressure-test it against AWS, Snowflake, Databricks, or any of the in-house data platforms that big customers could plausibly build instead. The lock-in is real; the explanation of why is hand-wavy.
What do Palantir’s actual customers say when the salespeople aren’t in the room? The book has Karp on tape at length. It does not have, say, an anonymous Pentagon program manager describing what it’s like to depend on a single vendor for mission-critical software, or what the switching cost actually looks like in practice.
What does Karp believe about AI, beyond the press-release version? He’s running a company whose pitch increasingly leans on AIP being the way enterprises plug LLMs into their proprietary data. We get his opinions on the Democratic Party. We don’t get his model of where the technology is going.
How does Thiel actually relate to the company now? Thiel is everywhere in the early chapters and recedes to caricature later, mostly as a Trump-supporting foil for Karp. The two of them are clearly the most interesting double act in this story, and the book underuses it.
Who should read it
If you want a fast, readable orientation to who Karp is, how Palantir got where it is, and why a company most people have never heard of is now worth more than most banks — this book delivers, and it’s the only one currently doing so. For a developer or technical reader trying to make sense of the headlines, the first two-thirds are genuinely useful background.
If you want a real treatment of the surveillance state, of how AI changes the math of state power, or of the specific trade-offs Palantir’s customers are making, you’ll need to read this alongside something else. Pair it with Karp and Zamiska’s own The Technological Republic to get the company’s preferred self-image, then with reporting from places like The Intercept or 404 Media on ICE and Lavender to see the texture of what the software actually does in the field. Steinberger gives you Karp; you’ll have to assemble the surveillance state yourself.
It’s a flawed book about an important subject, and on balance we’d rather read it than not. Just don’t mistake it for the definitive account. That book hasn’t been written yet.