Agostinho Serrano@EducatingwithAI
Read this document once as political philosophy. Then read it a second time knowing what it actually is: a corporate prospectus in the vestments of civilizational concern, written by the CEO of a company whose revenue depends, thesis by thesis, on precisely what the manifesto prescribes: more military contracts, more predictive AI, more surveillance-as-governance. Every proposition reads differently once you notice that its author profits from its adoption. That does not refute the arguments. It obliges the honest reader to place the conflict of interest on the table before assessing anything else.
Some of the diagnoses are real and worth keeping. The critique of the tyranny of apps, of the psychologization of politics, of the gleeful vanquishing of opponents, of the ruthless hollowing of public life; this is genuine cultural criticism in the lineage of Bloom (1987), Lasch (1979), MacIntyre (1981). Thesis 20, on elite intolerance of religious belief as a symptom of the closing of a supposedly open project, is perhaps the sharpest line in the book. But the scaffolding fractures at three critical points.
First, the manufactured inevitability. Thesis 5 ("AI weapons will be built; the question is who and for what purpose") is classical arms-race rhetoric, structurally identical to every military-technological escalation of the twentieth century. Its function is to dissolve the substantive ethical question ("should we?") into a tactical one ("if not us, them"). Oppenheimer used precisely this argument before spending the rest of his life repenting it. The trick works only if one accepts that tools are ethically neutral.
They are not - and here Vygotsky (1978) cuts far deeper than Maslow's "law of the instrument" ever did. Maslow described a perceptual distortion: whoever holds a hammer sees nails. Vygotsky's theory of mediation, extended by Leontiev (1978), Stiegler (1998), and activity theorists such as Engeström (1987), shows that tools do not merely distort the perception of a pre-formed subject. They constitute the subject. The instrument first appears on the interpsychological plane, as a social artifact, then migrates inward, where it reorganizes the very architecture of consciousness. A hammer used long enough does not limit what you see; it reconfigures who the "you" that sees has become.
Applied to Palantir, this is not ideological critique but diagnosis. Karp is almost certainly sincere, and the sincerity is the measure of how thoroughly the instrument has colonized the mediation. Two decades of building target-identification systems, signal correlation, and predictive-deviance models produce a subject whose cognition has been literally reformatted in the terms of the product itself. The book reads as it does, actually, absolute conviction braided with structural blindness, because the hammer has finished its work on the hand that holds it.
There is an aggravation Vygotsky did not live to develop but which contemporary political economy makes unavoidable: the capitalized hammer produces its own nails. It is not enough to distort perception of existing problems: the instrument, to reproduce itself as commodity, must manufacture the problems that justify its existence. Every inflated geopolitical threat, every pre-crime forecast, every border requiring vigilance is the hammer generating nails to sell more hammer. Stiegler (1998) named this the pharmacological ambivalence of technics: the same instrument can be remedy or poison, and under capital it tends structurally toward poison.
Second, the Pax Americana as selective mythology. Theses 13 and 14 celebrate a century of great-power peace. Peace for whom? Not for the millions killed in Korea, Vietnam, Iraq, Afghanistan, nor for the Latin American democracies dismantled by US-backed intervention — Chile 1973, Operation Condor, the Brazilian military dictatorship installed in 1964 (Kornbluh, 2003; Weiner, 2007). That is not peace. It is peace among nuclear powers while violence was outsourced to the Global South. The claim that "no country has advanced progressive values more than this one" becomes almost comic when read from Porto Alegre, Santiago, or Nairobi, as though Scandinavian social democracy, European universal healthcare, or Brazilian labor law counted for nothing. They do not count in Karp's ledger because the implicit metric is opportunity for individual ascent of non-hereditary elites, which conveniently happens to be what Silicon Valley sells.
Third, and deepest, the anthropology of permanent threat. The Enlightenment wager, from Kant's Perpetual Peace (1795) to Elias's Civilizing Process (1939) to Pinker (2011), was that violence is exosomatic: an evolutionary appendix, a residue of the state of nature that institutions, education, law and commerce would progressively atrophy and expel. The republican project, including its Catholic social variant (Maritain, 1936), assumes the well-ordered polis can push violence toward the margins of the improbable. The implicit anthropology in Karp is the exact opposite: violence never recedes; it only migrates vectors; the role of technology is not to extinguish it but to anticipate its next spatial location. This is a theology of permanent containment, not transformation.
The eschatology is visible in its consequences. If violence is inherent and irremovable, then peace is only temporarily undetected violence, maintained by exhaustive vigilance. This is not peace in Augustine's tranquillitas ordinis (ca. 426) — the tranquility of right order. It is carceral peace, the quiet of a maximum-security facility where inmates are still because every movement is monitored. The classical tradition (shalom, eirene, pax) understood peace as a positive plenitude, a rightness of relation, an ordo amoris realized. Palantir-peace is subtraction: violence minus effective surveillance. The more surveillance, the more peace. The equation requires violence to remain infinite in the denominator, because on the day violence reaches zero the product loses its market.
Foucault (1975) anticipated this without the tools. The panopticon does not require that someone always watch; it requires only that the watched believe they could be watched at any moment. Predictive AI modernizes Bentham's architecture by making observation effectively continuous. The subject that emerges: returning to Vygotsky, has had her interiority reformatted for self-surveillance, internalizing suspicion as a permanent structure of self-experience. Philip K. Dick's The Minority Report (1956) is no longer science fiction; it is a product specification. In the story the precogs disagreed among themselves, and the dissenting report was suppressed to preserve the system's appearance of infallibility. This is how every real predictive system must operate: suppress its own uncertainty to remain marketable as omniscient.
A theological irony deserves to be named. The manifesto presents itself as a defense of religious seriousness against empty secularism (thesis 20), but its operative cosmology is profoundly anti-Christian. A civilization organized around hard power, mocking grace toward the vanquished (explicitly rejecting thesis 11 when convenient), elevating builder-billionaires to Promethean figures (note the quasi-religious vocabulary around Musk in thesis 16), dissolving conscription into universal civic purification (thesis 6), this is pagan in the precise sense. It is Rome, not Jerusalem. It takes only the dark half of the doctrine of original sin (homo homini lupus made permanent) and discards both the imago Dei that grounds it and the redemptive possibility that completes it. What remains is secularized Calvinism without Christ: total depravity, statistical predestination, surveillance as substitute for grace.
Where we already failed, and where the answer must begin. The civilizational pact of modernity, Westphalia, the Enlightenment, the UN of 1945, the Universal Declaration, wagered that institutions well-designed could substitute for deep transformation of the subject. The bet: get the rules, incentives, and checks right, and minimally rational agents in well-designed structures will suffice. That wager has failed, or has proven catastrophically insufficient, because it underestimated the modern subject's capacity to desert the very institutions built to restrain him. Institutions can be captured, hollowed, instrumentalized, and when they are, nothing holds, because the wager ruled out the need for an internal bulwark. The paper endures anything; the soul of the citizen does not.
This is why education is not one response among many. It is the response. Aristotle's Politics, Freire (1970), Dewey (1916), Maritain (1936): the common intuition is that no institutional design compensates for the formative failure of the subject. What Vygotsky adds, making the argument contemporary rather than nostalgic, is that ethical formation does not happen through declarative moral content. Values taught as curricular content slide off consciousness without reshaping it. Formation happens in the structure of mediated activity. You do not form an ethical subject by teaching ethics; you form one by organizing activities whose internal structure requires the operation of higher ethical functions, genuine negotiation with the other, assumption of responsibility for consequences, reformulation in the face of counter-argument, suspension of judgment before the unfamiliar. Ethics enters the subject through the zone of proximal development, mediated by instruments and by more-competent humans, and is internalized as psychological structure.
Generative AI is the question on which the next century pivots. The same instrument, depending on the activity structure that surrounds it, produces either the citizen of Minority Report or the human of Fringe's parallel universe. In that series the "other side" has the same surveillance technology, compulsory biometric ID, airships in the skies, amber sealing reality — and yet the shared moral sense is more cohesive, more civic, more solidary. The difference was never the hardware. It was the anthropology that wielded it. Two universes, nearly identical physics, opposite humanities. The fork is not in the tool; it is in the activity structure and the cultural matrix that frames it.
Which brings us, at last, to Pope Leo XIV, who is acting, deliberately and, I believe, consciously, as the precise counterforce to the cosmology this manifesto encodes. His repeated interventions on artificial intelligence, on the digital environment, on truth in the age of synthetic media, should not be misread as conservative moralism. They are, in substance, a Vygotskian intuition dressed in theological language: the instrument is reformatting the soul, and no civilization survives that reformatting without having first decided, deliberately, toward what it wishes to be reformatted. Where Karp naturalizes violence as permanent anthropology to justify permanent vigilance, Leo XIV retrieves the older and harder wager, that violence is exosomatic, an appendix to be expelled through the formative labor of the human spirit across generations. Exophagy against endogenization. Education against surveillance. Imago Dei against statistical predestination.
The choice between these two cosmologies will not be made in manifestos or white papers. It will be made in thousands of small formative decisions, in classrooms, in the supervision of student projects, in how the next generation of AI tutors is designed, in whether a university allows a physicist with twenty years in education to teach pedagogy. The civilizational pact, if it is rebuilt, will be rebuilt one activity structure at a time.
Palantir is betting the other way. The rest of us had better notice what is on the table.
