Who he is
Affiliation: Co-founder (2003, with Thiel and three others) & CEO, Palantir — role check 2026-09-21: still CEO per his own Form 4 (filed 2026-08-24), the Q2 2026 letter (2026-08-03) and CNBC (2026-09-17). Co-author with Nicholas Zamiska of The Technological Republic: Hard Power, Soft Belief, and the Future of the West (Feb 2025). Born 1967; Haverford BA, Stanford JD (1992), PhD in neoclassical social theory, Goethe University Frankfurt (2002) — asked Habermas to supervise, was passed to a colleague. File correction: v1 said "Stanford PhD" (Wikipedia; NYRB, 2025-09-25). Net worth above $18B (2025), nearly all PLTR. Subject of Steinberger's The Philosopher in the Valley (2025-11; "a chilling portrait of genius in the yoke of authoritarianism" — Kirkus). Time 100 (2025).
One-line position: AI is the decisive technology for Western democratic security — Palantir builds the software that governments and militaries need to wield AI effectively, and pretending this isn't a power question is naive and dangerous. Since 2026 the thesis has a product name, "AI sovereignty" (running AI on your own hardware, data and model weights): "Demand for AI sovereignty has now been unleashed" (Q2 2026 release, 2026-08-03).
Discipline & technical bet
Not an engineer: a social theorist who runs a software company and writes quarterly essays quoting Augustine, Huntington and Houellebecq (Fortune, 2025-10-16). The wager: the value in AI sits in the integration layer, not the model. Palantir makes no frontier model; it sells Foundry/Gotham/AIP/Ontology — the plumbing that makes messy data usable by any model — and since 2026-06-29 an engine for running NVIDIA's open Nemotron models inside air-gapped networks so the customer owns the weights (Quartz, 2026-06-29). His bet: the pay-per-token model business — "the token industrial complex" — is "completely, irresponsibly, oversold"; enterprises are in "revolt" against handing model-makers "the keys to their institutions" (CNBC via Forbes, 2026-07-01; Q2 letter via Constellation, 2026-08-03). Military variant: Maven Smart System (MSS), Palantir as prime integrator over 179+ data sources (CSIS, 2026-06-02).
Key claims (Says)
- AI is a weapon of statecraft: whoever has the best AI-powered intelligence, logistics and decision-support will dominate geopolitically. Held (mechanism) — Operation Epic Fury (Iran, from 2026-02-28) used MSS "to conduct strike missions across the entire battle space": 13,000 targets in 38 days, 20 billion tokens a day at peak, an "insatiable appetite" per the Pentagon's AI chief (Breaking Defense, 2026-05-12); NATO bought MSS in one of its fastest procurements ever (NCIA, 2025-03-25). "Dominate" is graded under prediction 1.
- "The Technological Republic" (2025): democracies need a new social contract between technologists and the state; the technologist has obligations to democratic governance, not just to shareholders. Held (still endorsed; normative) — restated as a 22-point manifesto on X, 2026-04-19 (32M views): engineers have "an affirmative obligation to participate in the defense of the nation," national service should be universal, "hard power in this century will be built on software" (Al Jazeera, 2026-04-21; full 22 points).
- Palantir's model is the template: software that integrates messy government data into actionable intelligence, deployed across defense, intelligence, healthcare and immigration — the operating system of the state. Held (US) — US government revenue $809M in Q2 2026, +90% (8-K above); a $10B, 10-year Army enterprise agreement folding 75 contracts into one (US Army, 2025-07-31); MSS made a program of record (a permanent budget line, not a pilot) 2026-03-09, $2.3B in the FY2027 request (CSIS). Drifted (valence, Europe) — the same integration now reads as a dependency to escape: the Bundeswehr won't contract Palantir, the Netherlands wants out in two years, Denmark seeks a replacement, Switzerland has rejected nine-plus bids, a UK committee called the NHS dependency "an unacceptable point of weakness" (Euronews, 2026-06-05).
- The builder-operator distinction matters: Palantir doesn't advise — it builds and deploys working systems inside agencies; different from consultants or academics offering policy papers. Held — "we support war fighters... we get to bask in the part of the glory that's downstream from you guys winning" (UNITED24, 2026-05-12). Wrinkle: Palantir staff now advise the State Department on "countering digital surveillance" (The Intercept, 2026-08-07).
- Silicon Valley's reluctance to work with government is a national security risk: the best AI talent won't work on defense because of cultural stigma. Drifted — true when the book shipped, aged out within a year: by mid-2025 every frontier lab held a Pentagon contract (Anthropic's $200M, July 2025); the 2026 fight was over terms, not participation (dispute timeline). His posture inverted: "Are we really going to outsource the battlefield of this country to the consensus view in Silicon Valley? That is effing insane" (CNBC, 2026-07-01). What survives is the demand that the Valley serve on the state's terms.
- Privacy and security are complements, not tradeoffs: access-control architecture enables powerful analytics with audit trails and data governance — "the state and its agents can see only what ought to be seen" (Karp, via ACLU, 2026-03-24). Drifted — the observable is an independent audit and the company refused one: 56% of non-insider shareholders voted for a Human Rights Impact Assessment on 2026-06-03; it failed at 13% overall because founders' Class F shares (super-voting founder stock) carried it (8-K, 2026-06-03; IASJ, 2026-06-11). ICE's Palantir-built ELITE tool reportedly pulled Medicaid data to map raid locations (EFF, 2026-04-20); Palantir's 2026-03-06 letter denies any "master database."
Notable predictions — with falsifiable checks
- (Ongoing) AI-powered defense will be decisive in future conflicts — v1: supported by Ukraine, attribution difficult. Held (mechanism) — DoW officials and Palantir describe Epic Fury as the first major conflict with an AI targeting system as primary planning tool (Breaking Defense); Ukraine's defence minister says Palantir tech is integrated into planning deep strikes inside Russia (Meduza, 2026-05-12). "Decisive" (wins wars, not just speeds targeting) is ungraded. Check: by 2027-12-31, does an authoritative after-action assessment (DoW, RAND, CSIS, IISS) attribute the Iran campaign's strategic outcome to AI targeting tempo, or find tempo without decision?
- (2025) Democracies that don't integrate AI into governance will be outcompeted by adversaries — v1: TBD. Unfalsifiable as written — rewritten, not graded. Check: by 2028-06-30, either (a) Germany/France field a sovereign, NATO-interoperable AI planning layer at scale (ChapsVision, Helsing or equivalent), refuting his claim that sovereignty rules are "backfiring" (France 24, 2026-09-03), or (b) they buy MSS or a US equivalent anyway, confirming it.
- (Implicit) Palantir's model will be the template for state AI deployment — v1: "partially supported; $2.87B government revenue (2024), growing." Held, number corrected: $2.87B was total 2024 revenue; 2024 government revenue was ~$1.37B ($1.195B US) (FY2025 8-K, 2026-02-02). Corrected trend is stronger: US government $1.855B in FY2025 (+55%), $687M Q1 2026 (+84%, CNBC, 2026-05-04), $809M Q2 2026 (+90%). Check: FY2026 US government revenue above $3.0B and a second program-of-record by 2027-06-30 keeps Held; a termination wave reaching the UK MoD (£240M contract; £1.5B pledge, Computing, 2025-09) moves it to Drifted.
- NEW (2026-07-01): the frontier-model business is "oversold"; enterprises are in "revolt" against token dependency. Check: by 2027-06-30, labs' enterprise revenues decelerate while Palantir US commercial (guided above $3.424B for FY2026) holds above 100% growth = Held; labs still compounding = Drifted.
- NEW (2026-09-17): frontier labs will seek nationalization, or hand Washington a large stake, to cap liability — "these businesses have to be nationalized because if you don't nationalize them, every single one of my clients is going to sue" (his characterization of the labs' view; Motley Fool, 2026-09-17; Yellow, 2026-09-18). Check: by 2027-12-31, a federal liability cap or government equity stake in a lab = Held; a liability regime that lands on integrators too = Wrong.
Revealed behavior (Does)
- Money out: sells on a schedule. 2026-08-20: 975,000 Class B RSUs vested; 402,348 shares converted and sold for tax withholding (~$70.3M) plus 90,000 under a Rule 10b5-1 plan (pre-scheduled sales) adopted 2026-03-12 (~$15.7M); 6,432,258 Class A held directly afterward (Form 4, filed 2026-08-24). Earlier: $65.9M on 2026-02-20 (Investing.com); $50M+ in May 2025 (CNBC). Plan-driven, not conviction trades.
- Money in, personal: first/lead investor (personally, not Palantir) in ousted Ukrainian defence minister Mykhailo Fedorov's new company — battlefield robotics, jet-powered interceptor drones, cheap AI-guided missiles; amount undisclosed. Fedorov ran the ministry from January to 2026-07-15. Karp: "We've worked together for a half decade. I have been absolutely inspired by what you and the Ukrainians have pulled off" (Defense News, 2026-09-01; Kyiv Post). First move into hardware, and a private stake in the channel Palantir sells into.
- Time: Kyiv (2026-05-12, Zelensky and Fedorov); the G20 in Chapel Hill lecturing France (2026-09-03); Horse Guards for the UK £1.5B partnership (2025-09-18); a Tel Aviv board meeting "to embarrass other people who are pro-Israel in private but somehow not in public" (Times of Israel, 2024-01). Partners: the NVIDIA sovereign engine (2026-06-29; +7.8%, ~$22B of market cap on 2026-07-01) routes around the frontier labs; a "miniscule and shrinking sales headcount" (Q2 letter).
- Governance: the Class F bloc defeated both human-rights proposals at the 2026-06-03 AGM (Proposal 5: 290.9M for / 2,042.2M against; Proposal 4, independent conflict-zone due-diligence report: 193.0M / 2,137.4M) against a coalition including Norges Bank, CalPERS and NYC's pension funds (NYC Comptroller, 2026-02-04). Does: keeps control. Says: technologists owe obligations to democratic governance.
- Sides taken: backed the DoW's "wide license" in the Anthropic fight while calling himself "very sympathetic with arguments against using these products inside the U.S." and proposing self-imposed limits via a Valley consortium (Fortune, 2026-03-13). A federal judge found the blacklisting "illegal and baseless," "unlawful retaliation in violation of the First Amendment" (CNN, 2026-08-27). He picked the side that lost in court.
Feels
Wants to be read as a philosopher, is reviewed as a contractor — "You run Palantir!" in a critic's margin (NYRB). Underneath: grievance at the left he came from (Oct 7, 2023 and the campus protests "cemented" the break — Steinberger via Axios, 2025-10-23) and a creed from Huntington: the West rose "by its superiority in applying organized violence... Westerners often forget this fact; non-Westerners never do" (Q4 2025 letter, 2026-02-02). Fear: a West too squeamish to use its power. Hope: vindication as the one who saw it first.
Hears
Dead thinkers by choice — Huntington, Augustine, Nixon, Houellebecq, the New Testament (Fortune, 2025-10-16); living ones by proximity — Thiel, Zamiska, the Pentagon's CDAO, Jensen Huang, Fedorov and Ukrainian war-fighters. Structurally deaf to AI-safety researchers, European regulators and his own outside shareholders.
Sees
What the Pentagon actually does with AI at combat tempo — the real usage curve (20B tokens a day, 89% month-on-month classified growth) is visible from inside MSS and nowhere else. Frontier-lab dependency from the deployment side: "paying for tokens that create no value" is customer-channel intelligence. Ukraine's kill chain from inside. Which European governments keep Gotham running while announcing its replacement — France's DGSI keeps Palantir through ChapsVision's build-out (France 24).
Incentive map
Re-derived this cycle: maximal, pointed in every direction he talks. A ~$427B company (StockAnalysis, 2026-09-18; $37B end-2023, $172B end-2024, $424B end-2025) at roughly 150 times trailing earnings, priced on two stories he narrates: US government AI spend, and "AI sovereignty" — fear of frontier-lab dependency turned into a product. Who pays: the Pentagon; ICE ($81M+ in new contracts since Jan 2025, incl. the $30M ImmigrationOS prototype through Sept 2027 — PX14A6G, 2026-04-28); NATO; the UK and Israeli MoDs; US commercial ($764M in Q2, +149%). Can't say: that frontier models make the integration layer a commodity; that a domestic-surveillance use has occurred; that Europe is right to fear dependency; that Iran was strategically inconclusive. Convenient alignments: "liability, not regulation" (2026-09-17) puts unlimited liability on model-makers — his suppliers and rivals — and none on integrators; the "token industrial complex" attack landed two days after the NVIDIA open-weights deal; the manifesto is marketing to the customer it flatters (Bellingcat's Higgins: "completely normal" for a contractor whose revenue depends on the politics it advocates). Personal: an $18B+ PLTR position sold on schedule; a private Ukrainian arms stake; a Class F structure (Karp, Cohen, Thiel: 49.999999% of votes) that makes him unremovable. Rule-25 discount: maximal on "governments must buy AI" and every frontier-lab critique; high on Ukraine; moderate on the philosophy.
Theories aligned with
- AI as state power / democratic security — the bench's purest exponent, hard-power variant
- Technological statecraft — distinct from Schmidt (advisor); Karp is a builder/operator who ships production software into government
- Techno-optimism — the state-directed strain (cf. Andreessen's American Dynamism), not the abundance strain
- cf. Post-political progressive — self-described "progressive (but not woke)," since 2025 a "classic liberal" (Wikipedia); a test case for Rich's frame and a warning about where it can land
- NEW: "Sovereign AI" as an industrial thesis — open weights on your own hardware beat rented frontier models; shared commercially with Huang since 2026-06-29
What he's reacting against
- Silicon Valley's apolitical posture toward AI — argues AI is inherently a power tool and the West must use it or lose to adversaries
- The assumption that AI governance means restricting AI — insists the real risk is adversaries deploying AI that democracies won't
- "Move fast and break things" culture applied to government — argues government AI requires precision, secrecy, and institutional trust
- The EU regulatory approach — positions Palantir's operational deployment as the alternative to regulation-first governance; 2026: sovereignty rules are "backfiring" (France 24, 2026-09-03)
- The tech industry's moral ambiguity about defense contracts — Karp is explicit that Palantir serves Western militaries and intelligence agencies, and that this is good
- NEW (2026): the frontier labs — the "token industrial complex," labs "stealing the weights and alpha" of enterprise customers (quote log, 2026-07-01), a Valley "consensus" setting the terms of its own military service
- NEW (2026-09): the slow-down camp — Amodei's 2026-09-13 call to "slow the pace at which we improve the capabilities of AI models," endorsed by Altman and Musk (ABC, 2026-09-13). Karp's third front: "reasonable guidelines" but "you can't do it" by regulation; liability is "the first line of defense" — "if you build something that can destroy 10% of the world, there are civil and criminal liabilities attached" (CNBC, 2026-09-17)
- Still: the left he came from — "left-wing people fostering right-wing populist movements because they won't be adults about these issues" (Fox Business, 2024)
Where he overlaps / splits (with Rich)
- Overlaps with Schmidt on AI as a geopolitical competition; splits on role — Schmidt advises, Karp builds and deploys
- Overlaps with Luckey (Anduril) on defense-tech building; splits on surface — Luckey does hardware/autonomous systems, Karp does software/data integration; narrowing since the Fedorov venture
- Overlaps with Scharre (CNAS) on military AI policy; splits on posture — Scharre researches and regulates, Karp deploys and profits
- Splits with the AI safety community on priority — they worry about alignment, Karp worries about adversaries using AI first; sharpened 2026-09 against Amodei and Tegmark-style pause framings
- Splits with Mazzucato on the state's role — Mazzucato wants the state to direct innovation, Karp wants the state to buy from companies like Palantir
- Overlaps with Thiel structurally (co-founder, Class F bloc, the manifesto's politics); not independent sources on Palantir
- Product-partnered with Huang since 2026-06-29 (corrected 2026-09-22: the NVIDIA release describes a product partnership, no equity found) — yet Huang sits in the "market forces suffice" camp with Zuckerberg while Karp wants liability with teeth
- Splits with Varoufakis, who read the manifesto as adding "the AI-driven threat to humanity's existence" to militarised state power (Al Jazeera)
- Splits with Zuboff — for her Palantir is the surveillance state's contractor; for him the audit log is the privacy protection
- Splits with Bradford and Tooze on Europe — the 2026 de-Palantiring is Brussels-effect logic applied to defence software, the live test of his thesis (prediction 2)
- Overlaps with Toner only on governments, not labs, setting military terms — splits on whether "any lawful use" is a term or an abdication
- With Rich: overlap on "AI is a power question, not a product question"; split on where the power should sit and whether audit logs substitute for oversight. His "progressive but not woke" is a mirror for Rich's post-political frame — same start, very different destination
Track record
- Palantir deployed in US intelligence community since ~2004 — two decades of government AI deployment; ICE case management since 2014 (ACLU)
- Ukraine deployment (2022–present) demonstrated military-AI value in real conflict — the strongest evidence until 2026; Karp claimed in 2023 to be "responsible for most of the targeting in Ukraine" (via PX14A6G); 100+ companies now train 80+ models on a Palantir-built platform there (Meduza)
- Iran (Epic Fury, 2026): the first AI-primary air campaign, 13,000 targets in 38 days on MSS; Venezuela (Maduro capture, reported Feb 2026) also ran on Palantir systems (CSIS). Strategic verdict open
- Palantir stock performance (post-2022) validated by market — though stock price ≠ strategic impact: $37B (end-2023) → $172B → $424B (end-2025) → ~$427B (2026-09-18); Rule of 40 (growth plus margin; 40 is good) at 155%; more Q2 2026 profit ($1.06B) than Q2 2025 revenue
- Deal ladder in the window: NATO MSS (2025-03-25) · Army $10B (2025-07-31) · UK MoD £240M plus £1.5B pledge (2025-09-18) · Maven program of record (2026-03-09) · NVIDIA sovereign engine (2026-06-29). Reverse ladder: Germany (Apr 2026), Netherlands, Denmark, Switzerland, France (DGSI to ChapsVision), UK NHS exit recommendation for 2027; divestments by Storebrand (2024) and ABP (over €800M, Apr 2026)
- Controversial: ICE contracts, predictive policing associations, civil liberties concerns — the deployment surface includes applications many consider harmful. 2026: ImmigrationOS ($30M, through Sept 2027), ELITE with HHS/Medicaid data, the "compile data on Americans" story (Palantir's rebuttal to the 2025-05-30 NYT piece), the Sisters of St Joseph of Peace / NYC Comptroller / Norges Bank coalition
- The book: bestseller, critical failure — "fails to describe the 'Technological Republic' he invokes," "a LinkedIn post mercilessly stretched to nearly three hundred pages" (NYRB); the 22-point restatement drew "technofascism" (philosopher Mark Coeckelbergh) and 32M views
- Pattern: exceptional at converting a thesis into contracts, poor at converting it into legitimacy; every attempt to make him answer to outsiders (audit, AGM, the courts) has been settled by control, not persuasion
Empirical vs normative
- Empirical: government AI deployment data, contract revenue, military applications, Ukraine and Iran operational evidence; the enterprise "revolt" against token dependency (visible in Palantir's numbers, not yet in the labs')
- Normative: democracies must use AI for security; technologists have obligations to the state; the alternative to Western AI deployment is adversary advantage; hard power precedes moral appeal; universal national service; liability, not regulation, should govern AI risk
- The normative framework is explicitly political — this is not a neutral technology company; it's a company with a thesis about state power, and since April 2026 a published manifesto about it
- Conflict note (CONVENTIONS rule 25): every empirical claim above is also a sales claim for a ~$427B company he controls; discount highest on his frontier-lab critiques, where a competitor is describing competitors
Weak spots / open questions
- Palantir's government work includes applications (ICE, predictive policing) that civil liberties organizations strongly oppose — "AI for democracies" framing obscures controversial uses. 2026: the company reversed its own 2020 decision not to serve ICE removal operations (NYC Comptroller) and now advises State on "countering digital surveillance" (The Intercept)
- The "technological republic" thesis assumes Western governments will use AI wisely — but Palantir has sold to governments with mixed democratic records, and in 2026 the US government it serves was found to have blacklisted a vendor as "unlawful retaliation" for safety speech; Karp sided with the department
- Revenue from government contracts creates structural incentive to advocate for more government AI spending per CONVENTIONS rule 25 — now $2.3B in one budget line for the system he integrates
- Privacy-and-security-as-complements claim is contested — Palantir's access to data is vast, and audit trails are only as good as oversight; the oversight was requested by a majority of outside shareholders and refused by founder votes
- Less engaged with alignment, consciousness, or economic questions — the frame is purely about state power and security. Partly retired: he entered the alignment debate on 2026-09-17, on his own terms (liability) and against the labs; still nothing on consciousness or labor
- NEW: the liability position is untested against himself — if a Palantir-integrated system misfires in a strike or a deportation, does "you're liable for your own actions" reach the integrator? Unsaid. The Fedorov stake: personal money in a company selling into the channel Palantir serves, terms undisclosed. "Decisive" has no strategic scorecard — tempo proven, outcome not
- NEW: political identity is unstable in the record — "socialist," "progressive but not woke," "classic liberal," "full MAGA" (Axios), "voting against Trump" (2024) — grade by contracts and manifesto, not labels
- Source gaps: the Q2 2026 letter is JS-rendered, quoted via Constellation; Bloomberg, CNBC excerpts, Forbes and Axios were paywalled, quoted via secondary outlets; the Venezuela operation's name/date differ between the PX14A6G ("Absolute Resolve," Jan 2026) and CSIS (Feb 2026) — CSIS used
Rich's take
- (your synthesis here)
Delta log
2026-09-21 — v1→v2 migration + validation (bulk batch)
- Grades: 6 Held / 2 Drifted / 0 Wrong on Karp's own claims and predictions, plus 1 unfalsifiable prediction rewritten with a per-branch check and 2 new 2026 predictions with checks. Two Wrong facts were the file's, not his: "Stanford PhD" (Stanford JD; Frankfurt PhD) and "$2.87B government revenue (2024)" (total revenue; government ~$1.37B, FY2025 8-K).
- Held: AI as weapon of statecraft (Breaking Defense, 2026-05-12); Technological Republic contract, still endorsed (Al Jazeera, 2026-04-21); OS of the state, US branch (8-K 2026-08-03; CSIS 2026-06-02), Europe-valence Drift logged separately (Euronews, 2026-06-05); builder-operator; AI decisive in conflict (mechanism); template prediction (number corrected). Drifted: "Valley won't work with government" (premise dissolved, posture inverted — Forbes, 2026-07-01); privacy/security complements (audit refused by Class F vote — 8-K 2026-06-03; EFF 2026-04-20). Unfalsifiable, rewritten: outcompeted-democracies thesis — Europe-branch check due 2028-06-30.
- Most surprising delta: the persona filed as "less engaged with alignment" became, on 2026-09-17, the named third pole of the AI-safety fight — liability, not regulation, against the Amodei/Altman/Musk slowdown and the Huang/Zuckerberg market-forces camp (Motley Fool; CNBC, 2026-09-17) — and the pole he chose puts unlimited liability on model-makers and none on integrators like Palantir.
- Employer/role check (mandatory): no change — CEO, per Form 4 (2026-08-24), Q2 letter (2026-08-03), CNBC (2026-09-17). Incentive map re-derived from scratch: NVIDIA (2026-06-29), the personal Fedorov stake (2026-09-01), the Class F vote (2026-06-03), ICE totals (PX14A6G, 2026-04-28).
- Post-review sources: the v1 "last-reviewed 2026-06-16" file held nothing after the Feb 2025 book — no NATO (Mar 2025), ImmigrationOS (Apr 2025), Army $10B (Jul 2025), UK £1.5B (Sep 2025), Iran war (Feb 2026), program of record (Mar 2026), manifesto (Apr 2026), or the AGM vote 13 days before its review date. Sixth v1 file from the May/June-2026 pass confirmed as a no-current-events review.
- Migration: Weak spots, Track record, Overlaps, Theories, Reacting-against, Empirical/normative and Sources carried and extended; Rich's take carried as-is (placeholder). Tier: quarterly — a letter he writes every quarter, an AI-run war, European contract churn, a new personal venture and a live liability debate. Confidence stays medium: thick record, but the central geopolitical claims stay unfalsifiable at the outcome level and every empirical claim is also a sales claim.
Sources
- Primary, not linked above: Q1 2026 letter to shareholders (2026-05-04) · Q2 2026 letter (2026-08-03) · Hansard, Commons debate on MoD Palantir contracts (2026-02-10) · CNBC, Karp backs Fedorov venture (2026-09-02) · palantir.com (SEC filings, government contracts)
- Cited in-text, dated, primary preferred: SEC 8-Ks and Form 4, PX14A6G, Karp's letters and interviews (Fortune, CNBC, UNITED24, France 24), US Army, NCIA, CSIS, Breaking Defense, Meduza, Defense News, Kyiv Post, Euronews, ACLU, EFF, NYC Comptroller, IASJ, The Intercept, ABC, CNN, Al Jazeera, NYRB, Axios, Kirkus, StockAnalysis, Wikipedia
- Carried from v1: The Technological Republic (2025); interviews (Bloomberg, CNBC, various defense-tech conferences); Ukraine deployment: publicly reported military AI support
migrated v1→v2 2026-09-21 · backup: _archives/alex-karp.html.bak-20260921 · converts-from: people/alex-karp.md · AI & Society domain