The technology layer is the geopolitical layer.

Karp's loudest public claim is that Silicon Valley quietly walked away from national-defense work during the 2010s — that the talent, ambition, and capital that built the internet in alliance with the federal government in the 1960s-1990s drifted into consumer apps and entertainment products instead. He argues this drift is not neutral: it cedes the technical layer that decides 21st-century geopolitics to actors with worse values. Palantir, by extension, is the company-as-counter-argument.

Karp's 2024 book (with Nicholas Zamiska) frames the West's technological lead as downstream of a now-broken alliance between technology and the state. The book is mostly a cultural argument — about the academy, the elite university's drift into ideology, the corporate boardroom's preference for consensus over conviction — and only secondarily a software argument. Karp's prescription is essentially a refounding move: rebuild the alliance, refuse the trivial work, accept the controversy.

Palantir under Karp has publicly refused certain commercial contracts (specifics named in various interviews — including walking away from advertising-targeting work that would conflict with its national-security positioning) and equally publicly defended contracts that Silicon Valley peers found controversial (ICE, Israel MoD). The pattern: Karp's defense is always the same — the alternative is ceding that work to actors who would build it worse, with worse values, against US interests.

Most defense-tech CEOs avoid extended public commentary. Karp leans into it. He speaks at length at CNAS, Hudson, the Reagan Defense Forum, and on extended-form podcasts. He writes letters to shareholders that read more like cultural essays than financial communications. The strategic effect is that Palantir, alone among defense-tech primes, has a coherent public-intellectual frame to point at — one that customers, employees, and policy-makers all read.

Palantir winning the Maven Smart System prime contractor role in 2024 (~$153M initial, expanded substantially through 2024-2025) is the concrete proof point Karp now cites. Project Maven began in 2017 as the DoD program that triggered Google's withdrawal from defense work after employee protests. Eight years later, the same workflow runs on Palantir's stack. Karp's posture frames this as exactly the predicted endpoint of his thesis: the work doesn't disappear when one company refuses it — it migrates to a company with conviction.

Palantir's hiring loop screens for conviction explicitly. Forward Deployed Engineer interviews probe what you would and wouldn't build for whom. The company tells you in the interview that you'll be embedded with government customers — ICE, IC components, DoD elements, allied MoDs — and asks if you understand what that means. The honest answer is the right answer. Pretending it doesn't matter is what gets candidates filtered out, not in.