Maurice Blondel on Artificial Intelligence and Divine Action: The Insufficiency of Technical Work

Abstract

Confronted with generative artificial intelligence, a lot of secular humanism defends the worth of the human person by naming capacities that machines are said to lack, a strategy Leif Weatherby calls remainder humanism. The ground so defended shrinks with every advance, but the deeper fault is that the strategy misdescribes what it defends. This article brings Maurice Blondel’s philosophy of action, largely absent from philosophical discussion of emerging technology, to bear on that anthropological problem, proceeding by Blondel’s own method of immanence: it begins from what is given in the experience of acting and asks only afterward what must be true of reality for that experience to be what it is. It argues that artificial intelligence is an instance of what Blondel calls the volonté voulue, the willed will, a determinate object of a willing whose reach it can never equal, and that the disproportion between the two leaves technical work permanently open to a completion it cannot itself supply. It is that opening, and not the artifact, in which divine action is located, and so located it is no rival to the work. The classical distinction between immanent and transitive action, which Blondel takes up in his own idiom, supplies the positive account the subtractive strategy lacks: what is proper to the person is not a residue of competences but an acting that perfects the agent and cannot be transferred to any product. Blondel’s analysis of superstitious action then supplies a better account of the remainder than the contemporary debate possesses. The disproportion is not a defect to be engineered away but the permanent condition of technical work.

1. Introduction

Anyone who has argued about artificial intelligence knows the moment when the ground of the argument shifts. Machines cannot really hold a conversation, and then they can. They cannot write anything a person would want to read, and then the essays arrive. They cannot be creative, and then creativity is redefined. The redefinition is not always quiet, and it is worth distinguishing two ways it happens, since they fail differently. Sometimes the concession is made tacitly, presented as a clarification the defender had meant all along. Sometimes it is made openly and prescriptively: what the system does is simply declared to be what the word meant, and the declaration is announced rather than smuggled, as when a model is proclaimed to have crossed into consciousness or understanding. The first move retreats; the second legislates. Both leave the concept in the same condition, since a definition revised every time an engineer succeeds is not a definition, whether it is revised in silence or by proclamation. It is a receding line, and the people who move it are not the ones defending it.

Leif Weatherby has a name for the humanism that argues this way: remainder humanism. A book that does not sell is remaindered, returned as unsold stock, and the word fits what is left of the human in these defenses: “a humanism without a theory or doctrine of what is human, in which humanity is remaindered, like a book past salability” (Weatherby 2025, p. 34). The strategy defends human worth by pointing to capacities machines are said to lack, so that “[t]he human here is defined by technology’s creep, but only negatively” (Weatherby 2025, p. 34). That is a losing position, since every concession narrows the next defense’s ground. But my concern in this article is not that the strategy is losing. It misdescribes what it is defending, because it presupposes that the human is what is left over once the machine’s competences are subtracted.

The retreat is understandable, and I want to credit what motivates it before objecting to it. Those who reach for the remainder are responding to something real: the phenomena surrounding this technology, companionship sold to the lonely, intelligence sold back to us as a utility, are moral before they are technical, and the available secular vocabulary strains under them. My disagreement is not with the seriousness of the diagnosis, but with the anthropology it presupposes. A defense that begins from a perception this accurate deserves a better description than the subtractive strategy can give it.

The resources for a better description lie in the philosophy of action developed by Maurice Blondel and carried forward by Catholic philosophers and theologians after him, Henri Bouillard among them, and largely neglected in philosophical discussions of emerging technology. L’Action (1893) is not a treatise on human capacities but a study of what Blondel calls the phenomenon of action: an inquiry that begins from what is given in the experience of acting and asks what must be true of reality for that experience to be what it is. The book opens with a question that makes the subtractive method unavailable from the outset: “Yes or no, does human life make sense, and does man have a destiny?” (Blondel [1893] 1984, p. 3). That question is not answered by an inventory of capacities, because it is not a question about capacities at all. It asks whether what a person does can ever be brought into equality with what that person wills. Both remainder humanism and the enthusiasm it opposes have already answered it by inventory, taking the human to be whatever an enumeration of competences leaves standing. The irreducible element, I will argue, is not an item on any list, which is why no list can settle the question.

The argument makes four claims, stated here so that each step can be held to its warrant. First, an artifact is never a neutral instrument standing outside its maker: every made thing carries the personal coefficient of the persons whose choices constituted it, so the question about artificial intelligence is not what it can do but what kind of willing it expresses (Section 2). Second, placed within Blondel’s distinction between the willing will and the willed will, such a system is the most ambitious determinate object human willing has yet produced, and the disproportion between the two wills is structural, disclosed rather than diminished by every success (Section 3). Third, the classical distinction between immanent and transitive action, which Blondel takes up and reworks, locates what is proper to the person in an acting that perfects the agent rather than in any product that acting leaves behind; this supplies the positive account of human worth that the subtractive strategy conspicuously lacks (Section 4). Fourth, the true remainder is that disproportion itself, not a residue of unmatched competences; treating a finite work as able to absorb it is what Blondel analyzes as superstition, and receiving the insufficiency as an opening is what allows technical work to be a locus of divine action rather than a rival to it (Section 5 and Section 6). Section 7 answers four objections, and Section 8 states the conditions under which the thesis holds. My method is Blondel’s own, and the order of the argument reflects it. The argument is philosophical throughout. What the method of immanence yields is an exigency internal to action, a demand action makes and cannot meet; whether it is answered is a question philosophy can pose and cannot settle. The magisterial texts cited along the way are corroborating testimony from an independent line of reasoning, never premises.1

2. Technological Action and the Personal Coefficient

The subtractive defense and the enthusiasm it opposes share a premise, and the premise is what needs examining first: that the artifact stands outside the person who makes it, so that the two may be compared as rivals. By artifact I mean, here and throughout, the determinate thing that making leaves behind: a hammer, a bridge, a trained model. The word names the product of an act, never the acting that produced it, and much of what follows depends on keeping the two distinct. Michael Hanby states the difficulty precisely: “But of course technology is not merely an instrument. It is not merely something we use to shape the world; rather it profoundly shapes us” (Hanby 2011, p. 200). The word “technology” is doing two jobs at once here, and the argument cannot proceed until they are separated. In one sense it denotes an instrument: a determinate made thing, available for use, describable without reference to any particular user. In another, it stands for technē (τέχνη) fused with logos: a way of understanding what things are, which is not available for use because one is already inside it. The Greek term is itself equivocal, as Pfau notes, since it “blurs the boundaries between methodical ‘making’ and inspired ‘creating’“ (Pfau 2022, p. 46), and that equivocation is precisely what the neutrality thesis exploits. Taken in the first sense, the thesis that technology is neutral is nearly trivial; taken in the second, it is false, since a civilization that builds in a certain way has thereby committed itself to an account of nature. The neutrality thesis survives only by trading on the first sense while its conclusions are drawn in the second. Hanby puts the historical form of the claim more sharply elsewhere: the “technologization of being” is “the very essence of modernity,” with the stipulation that this does not preclude other true descriptions of that essence (Hanby 2016, p. 352). I take the qualification seriously, and press it further than Hanby does, since what follows depends on a description he does not give: an account of the willing that does the technologizing.

Polanyi arrives at the same conclusion from the side of knowing. Every application of language to things, he argues, is unformalized, and therefore rests on a skill the speaker endorses in exercising it: “This personal coefficient of all affirmations inherent in the use of language will be presently reconsidered in the wider context of ineffable knowledge and ineffable thought” (Polanyi 1998, p. 84). The knower is a condition of there being knowledge at all rather than an interference to be filtered out. Copernicus, on Polanyi’s reading, exchanged a terrestrial for a solar standpoint on the strength of an intellectual satisfaction the available evidence could not then justify (Polanyi 1998, pp. 2–4). The personal coefficient is a submission to reality and not a retreat from it, and this is what licenses the transfer from knower to maker: the transfer is not an analogy but an instance, since engineering is applied knowing, and whatever holds of the unformalized ground of knowledge holds of the unformalized ground of design.2 What holds of the astronomer holds of the engineer. The choice of an architecture, the sense that a training run is converging on something real, the judgment that a benchmark measures what matters: none of these is derived from a rule, and each is a commitment of the person who makes it. The coefficient does not make engineering arbitrary any more than it made astronomy arbitrary; it makes both personal, in Polanyi’s sense of a person staking himself on contact with a reality he did not invent. “To know that a statement is true is to know more than we can tell,” Polanyi writes, “and that hence, when a discovery solves a problem, it is itself fraught with further intimations of an indeterminate range” (Polanyi 2009, p. 23).

Two consequences follow. The first is epistemic. A machine-learning system is not a knower that has dispensed with a personal coefficient; the coefficient was exercised, in advance and irrevocably, in the choices that constituted the system: of objective, of corpus, of what would count as success. The personal element has not been eliminated from such a system. It has been displaced within it, from the moment of use back to the moment of construction, where it is harder to see and no less operative. The second consequence is anthropological, and here the language of expression that has served so far must give way to something more exact: Karol Wojtyła’s analysis of the act. Person and Act is written, as he says at the outset, to study “the act that reveals the person”: the act is approached not as something a person is presupposed to produce but as the place where the person is disclosed, as “the dynamic reality itself, which at the same time reveals the person as its efficacious subject” (Wojtyła 2021, pp. 103, 114). What does the revealing is efficacy. Wojtyła judges that the manuals’ distinction between actus humanus and actus hominis marks a difference rather than explaining it, and locates the real basis in something given in experience: the lived-experience of efficacy, which separates the structure “man acts” from the structure “something happens in man” (Wojtyła 2021, pp. 168–69). Where efficacy is lived, there is an efficacious “I”; where it is absent, there is an occurrence and nothing more. The scholastic axiom Wojtyła invokes here, in the synopsis of Person and Act that he drafted before 1969 and that this edition prints for the first time, is operari sequitur esse, acting follows upon being: the act “has its own essence and its own existence while remaining most closely linked to the essence and existence of the person,” so that the act is “in a sense the person (persona) in actu” (Wojtyła 2021, p. 23).

The claim needs qualifying before it can be used, because in its unguarded form it says more than the evidence will carry. Three claims must be kept apart. The weakest is that such a system bears personal coefficients at all, that somewhere in its constitution are commitments no rule dictated; this is a presupposition the argument requires, and it is hard to deny. The strongest is that the system expresses a particular human will, so that one could read off from its behavior the intention of an identifiable maker; this the argument does not require, and I do not assert. Between them lies the claim about responsibility, that the persons and institutions who constituted the system are answerable for it; this I do assert, and it is a moral claim rather than an expressive one. The middle claim is the one that matters, and it survives everything that makes the strongest claim false. A trained system is not the work of one hand. Its coefficients are plural and often in tension: the researchers who chose an architecture, the firms that set the objective, the annotators whose judgments fixed what counted as a good answer, the vast and largely unexamined corpus whose authors never consented to the use, the users whose interactions retrain it, the regulators who constrain its deployment. Much of what such a system does was chosen by no one, and some of it surprises the people who built it, since a fitted model generalizes in ways its makers neither specified nor foresaw. None of this restores the neutrality thesis. It distributes the coefficient rather than removing it, and a coefficient distributed across many persons and institutions is still a coefficient, still unformalized, still the commitment of persons who staked themselves on judgments no rule compelled. What follows is not that the artifact speaks for a particular will, but that there is no point in its constitution at which persons drop out, and a neutral instrument takes over.

Efficacy is in turn the entrance to what Wojtyła calls the transcendence of the person in the act, and the term carries a precise sense that is easily mistaken. It does not mean intentionality. The crossing from subject toward object that occurs whenever we know or want something, he terms horizontal transcendence, and he says plainly that this is not what he means. What he means is “the transcendence that we owe to self-determination, the transcendence by the very fact of freedom, the fact of being free in action,” which he calls vertical; every authentic “I will” reveals it, and it is on this account that he can write that “the person is transcendent in his own action because he is free—and inasmuch as he is free” (Wojtyła 2021, pp. 221, 241; cf. 213).

This is the distinction the case at hand requires, and it is sharper than the one usually reached for. In a trained system, something happens. Parameters are adjusted so as to fit a probability distribution, in the case of a language model, a distribution over the possible next elements of a text, from which each successive output is drawn; and at no point does anything in the system undergo the lived experience of efficacy that Wojtyła describes. What the word distribution names here is a mathematical object, an assignment of probabilities across a space of possible continuations, and not a distribution of goods or of anything else that could be called a decision. The modality of this claim should be stated exactly, because it is criterial and not empirical: I do not claim to have inspected the system for a hidden inner life and found none. The claim is that nothing in the concept of a fitted distribution requires one, and that the efficacy in the case is fully accounted for by the persons who fixed the objective, assembled the corpus, and decided what would count as success. It is those persons whom the artifact reveals, in the way any made thing reveals its maker. Nor is the point that such a system lacks a capacity we happen to have. It exhibits much that resembles the horizontal movement, being directed at objects and built toward ends; of the vertical it exhibits nothing, because there is no self there to determine itself. The point is easily misheard as a verdict about the technology’s future, so it is worth saying where the vertical transcendence Wojtyła describes is actually exercised: at the desk where the objective was written, in the meetings where the corpus was approved, at the moment someone decided the outputs were good enough to ship. The freedom in the system’s operation is real, and it is all upstream. Jove Jim Aguas states the consequence exactly, reading Wojtyła alongside Aquinas: because the person is sui juris and alteri incommunicabilis, his own master and not transferable to another, “nobody can decide for us, nobody can want for us, no one can substitute his will for ours” (Aguas 2009, p. 56). That last clause is the whole argument of this article in miniature, and Section 7 will return to it.

The recent magisterium has said as much in its own idiom, and it says it from both directions at once. Leo XIV, writing on artificial intelligence, denies that technology is antagonistic to humanity and describes it as having “formed part of our history since the beginning as ‘a profoundly human reality, linked to the autonomy and freedom of man’“; the inner words are Benedict XVI’s, and the continuity is deliberate, since the encyclical frames its teaching on artificial intelligence as an extension of the claim that technology belongs to human freedom rather than standing against it (Leo XIV 2026, no. 4). John Paul II, addressing communicators, was more direct still: “Do not be afraid of new technologies!” (John Paul II 2005, no. 14). And Benedict XVI called the digital technologies “truly a gift to humanity,” insisting that their benefits be measured by whether they serve those most disadvantaged (Benedict XVI 2009). None of these texts treats technology as neutral. Each treats it as human, which is to say as belonging to the same order of act, and therefore falling under the same order of judgment, as anything else a person does. The claim is stronger than neutrality and differently directed: neutrality would exempt the artifact from moral analysis altogether, whereas humanity subjects it to the whole of that analysis, including its capacity for idolatry. Leo XIV describes the disposition in question under the heading of “a technocratic paradigm,” a phrase he takes from Francis, and identifies it as the tendency to let efficiency, control and profit alone determine what matters and what may be discarded (Leo XIV 2026, no. 92); the vocabulary of idolatry is my own gloss on that tendency, and he reserves the word itself for the idolatry of profit (Leo XIV 2026, no. 10). Wojtyła had drawn the ordering that follows from this with unusual directness: to acknowledge the dignity of the human being “means to place people higher than anything derived from them in the visible world,” so that “human beings do not live for the sake of technology, civilization or even culture; they live by means of these things, always preserving their own purpose” (quoted in Aguas 2009, p. 57).

If the artifact is not outside the maker, then the question of what artificial intelligence is cannot be settled by comparing its competences with ours. It must be settled by asking what kind of willing it expresses. That is the question Blondel’s distinction was built to ask.

3. The Willing Will and the Willed Will

Blondel’s answer turns on a distinction internal to willing itself. Cathal Doherty states it as compactly as the literature allows: the science of practice investigates the relation between “la volonté voulante,” the willing will, “which is knowable only by reflexion on specific exercises of the will,” and “la volonté voulue,” the willed will, which “corresponds to decision, the choice of one finite motive,” while the willing will “corresponds to the infinite power to decide, i.e., the power to choose one finite motive out of an unlimited number of other finite motives” (Doherty 2017, p. 53). Blondel’s own formulation, in the analysis of freedom, gives the distinction its Latin precision and its affective weight:

That is why, in proposing freedom as our end, we feel a disproportion between the willing will, quod procedit ex voluntate (what proceeds from the will), and the willed will, quod voluntatis objectum fit (what becomes the object of the will). We experience the difficulty of a choice and a sacrifice. That infinite power of contraries that enveloped and dominated all our ways of being and desiring has to impose a determination on itself.

(Blondel [1893] 1984, p. 134)

Blondel’s verbs are worth dwelling on: we feel the disproportion, and a sentence later we experience it. It is not first a conclusion drawn from a doctrine of the will but something given in the experience of choosing, and given most sharply at the moment a choice is made, when what is chosen appears against the background of everything the chooser might have willed instead. The disproportion is not, then, a defect of particular choices to be remedied by choosing better. It belongs to willing qua willing. It is structural, and it is what makes the whole of L’Action a single question: “for one who limits himself to the natural order, is there any concordance between the willing will and the willed will; and does action, which is the synthesis of this double willing, finally find in itself the wherewithal to be self-sufficient and to define itself?” (Blondel [1893] 1984, p. 53).

Placing artificial intelligence within this distinction does real work, and it does it in two directions. Considered as volonté voulue, such a system is the most ambitious determinate object our willing has yet produced: an attempt to give durable external form to the operations of intelligence itself, and thereby to secure in a finite artifact something of the infinite power to decide. But considered from the side of the volonté voulante, the same system is evidence that the willing exceeded it, and the evidence is of a specific kind: no capability, once achieved, functions as a term; each redescribes the goal instead of reaching it, and the redescription is not a failure of the engineering but a disclosure of what the engineering was answering to. If the willed will could equal the willing will, some sufficiently ambitious artifact would close the gap and be received as final. That none is so received, and that the most impressive artifacts sharpen the demand rather than settle it, is what the disproportion thesis predicts and the sufficiency thesis has to explain away. Blondel formulates the resulting task exactly: “The whole art of living, the whole effort of sincerity, will be to bring about as perfect a conformity as possible between the initial will and the positive will” (Blondel [1893] 1984, p. 135). The sentence turns the disproportion from a datum into a task: sincerity is not a matter of finding the object that would finally suffice but the ongoing labor of conforming what one posits to what one is. Applied to the technological case, this reframes what an engineering vocation is. The task is not to build the artifact after which nothing remains to be willed, since there is no such artifact; it is to keep one’s building answerable to the willing it expresses, a discipline of the builder and not a property of the built. Why this artifact rather than any other should be the one that sharpens the disproportion is a question the third objection in Section 7 answers at length; the short answer is that only here is the willed will built to give finite form to the power of determination itself, and so only here does it present itself as a candidate substitute for the willing will.

The distinction can be held in something as domestic as a letter. Dhananjay Jagannathan, teaching Aquinas (1920) on the angels to undergraduates who, as he notes, increasingly hand their intellectual work to a chatbot, observes that “ChatGPT cannot really write a letter. To be sure, it can compose a text that resembles a letter, much more quickly than any of us can. But to write a letter is to take a certain kind of journey, to choose—and continue to choose—one’s words reflectively” (Jagannathan 2025). The text that results is the volonté voulue: a determinate object, complete, available for delivery, and producible by the system in seconds. The continued choosing is the volonté voulante at work, and it is not in the text, nor could it be added to it; a reader who receives the machine’s letter has received everything except someone’s having meant it. What the student who submits the generated essay has skipped is therefore not labor but the act itself, in Wojtyła’s sense, that the assignment existed to occasion. The product was never the point, or never the whole of it, and the product is the only thing the machine supplies.

This is also why the option of standing aside is not available. Blondel’s dilettante, who declines to will anything in particular, is not thereby exempt from willing; in willing not to will he “necessarily and actually wills something,” and so, as Doherty puts it, “practices what we might call a performative contradiction” (Doherty 2017, p. 57). A refusal to participate in technological development is itself an act with its own efficacy. What follows from the disproportion depends entirely on how it is received, and the two ways of receiving it are the subject of the next section.

4. Immanent and Transitive Action

The distinction drawn provisionally in Section 2, between the determinate thing that making leaves behind and the acting that produced it, has a classical name, and naming it properly is what allows the positive account to be stated. The tradition divides action in two. Aquinas, following the ninth book of the Metaphysics, puts it compactly: action is twofold, for actions of one kind “pass out to external matter, as to heat or to cut; whilst actions of the other kind remain in the agent, as to understand, to sense and to will” (Aquinas, Summa theologiae I, q. 18, a. 3, ad 1). The names that later usage settled on are transitive for the first and immanent for the second, and the difference that matters is not where the action goes but what it perfects. Aquinas states it in the next clause: “the former action is the perfection not of the agent that moves, but of the thing moved; whereas the latter action is the perfection of the agent” (Aquinas, Summa theologiae I, q. 18, a. 3, ad 1). Heating perfects the water. Understanding perfects the one who understands. Nothing in the first case accrues to the agent as such, and everything in the second does.

Blondel takes the distinction up, and he does something with it that the manuals do not. He registers it first as a point of vocabulary, in a note that repays more attention than it usually gets. The word act, he says, “refers rather to the first initiative of the internal effort,” to “the entirely subjective part of the agent,” while the word action “refers rather to the passage from the intention to the execution that embodies it, and often, consequently, to the result or the work itself of this transitive operation” (Blondel [1893] 1984, p. 121n). The two words the argument has been using all along are therefore Blondel’s own, and they mark his version of the classical line. But he immediately denies that the line can be drawn as a partition. What he says of action in general is that it “always appears as transitive, that is, it seems to be in a perpetual becoming,” and that “each end, as soon as it appears reached, is already only a stepping-stone for going beyond” (Blondel [1893] 1984, p. 203). The reason he gives is the one this article has been pressing: “precisely because there is in him insufficiency, penury, imbalance between the initiative and the results of his internal operation,” the agent goes out of himself, “seeking his completion and his equation” (Blondel [1893] 1984, p. 204). Transitive action is not a second department of life alongside the interior one. It is what the interior insufficiency drives us into.

This is why Blondel’s treatment of the body belongs here and not in a separate anthropology. He devotes an entire stage of L’Action to what he calls the body of action, and his claim is that the body is not the instrument the will picks up but the first place where willing meets what it did not posit. “In the eyes of consciousness,” he writes, “the body is a beginning of subjective life. It is what resists the immediate expansion and the ideal reality of willing, but it is also what submits and lends itself to it” (Blondel [1893] 1984, p. 157). Seen from within, it is “a consequence and like a prolongation of our subjective nature: we will it to be” (Blondel [1893] 1984, p. 153); seen as obstacle, “the body of action is everything, in ourselves and outside, that still separates us from ourselves” (Blondel [1893] 1984, p. 153). The body is therefore the original case of the transitive, and every artifact stands downstream of it: “Every act that comes out of the human organism is, even outside of ourselves, an organism of signs and a symbol expressive of subjective life” (Blondel [1893] 1984, p. 198). A trained model is a recent and elaborate member of that series, continuous with the hand and the voice, not a new kind of thing.

What follows for interiority is the opposite of what the subtractive strategy wants from it. Interiority, on this account, is not a private compartment the machine has failed to reach. It is the immanent side of one act whose other side is out in the world, and Blondel’s formulation of the relation is as compact as it is unexpected: “it is through action that the soul gets a body, and the body gets a soul; it is their substantial bond; it makes of them a natural whole. In us transcendence implies immanence” (Blondel [1893] 1984, p. 181). Awareness is not a spectator of the transitive work but the mode in which the agent is present to it, which is why Blondel can say that within the individual, action “constitutes a closed circuit” that “fashions body and soul for us,” and then immediately say that “this moving circle closes in on itself only to open up again once more and to become the point of departure for a new expansion” (Blondel [1893] 1984, p. 176). The circle that will not stay closed is the whole argument in a figure, and Section 5 will show what happens when someone insists on closing it.

Now the positive account can be stated, and it is the account the subtractive strategy never reaches because it is looking on the wrong side of the line. What is proper to the person is not a residue of competences left over when the machine’s are subtracted. It is immanent action: an acting that perfects the agent in the doing of it, that is not itself a product, and that cannot be handed to anyone else. Wojtyła’s vertical transcendence names the same thing from the side of freedom, since what self-determination determines is first of all the self that determines. Aguas’s formula names it from the side of incommunicability, and it is exact: no one can want for us. The reason is now visible, and it is not a limitation of any available technology. A transitive act can always in principle be performed by another agent or by a machine, because its perfection lies in the thing produced, and the thing produced does not care who produced it. An immanent act cannot, because its perfection lies in the agent, and there is no mechanism by which my understanding could perfect you. Human worth, on this account, rests on the person’s standing as the subject of immanent acts, which is a standing and not a score, and which therefore no advance in transitive capability can erode. The ground does not shrink, because it was never the sort of thing that could be occupied.

Two consequences follow immediately, and both bear on the debate this article entered. The first is that the question of whether a machine will eventually match some human competence, however it is answered, does not touch the account, since competences are transitive achievements and worth is not located among them. The subtractive defense and the enthusiasm it opposes are arguing about the same column of the ledger, and the entry that matters is in the other column. The second is that transitive work is not thereby demoted. It is the medium in which immanent action is exercised and exposed: the engineer who writes a training objective is performing an immanent act whose transitive residue is a model, and the model, precisely because it is transitive, will never contain the acting that produced it, which is not to say that the model is without value. The disproportion of Section 3 is now specified. It is not merely that our works fall short of our willing in degree. It is that they fall on the other side of a line, and no accumulation of what lies on one side ever amounts to what lies on the other.

5. Superstitious Action and the True Remainder

Blondel’s analysis of superstition is the least read part of L’Action and, for the question before us, the most useful. He is not describing a primitive stage of religion. He deliberately relocated the critique of superstition out of the treatment of religion and into the phenomenon of action itself, and the placement is principled: superstition is one of the two things action can do with its own insufficiency, and it belongs to the phenomenology of acting on that ground alone.3 What it describes is action trying to complete itself from its own resources. “There is no act, no matter how vile, where man has not placed the divine, no act that has not aroused some idolatry,” he observes, and then asks what the superstitious cult is actually directed at:

What then, in short, is the object of the superstitious cult? Under a borrowed form (for here the need creates the organ and the food that nourishes it), it is the expression of that inexhaustible depth of the interior life which no particular act has equaled, it is a desire that embodies itself, the desire for an infinite response to an infinite tendency.

(Blondel [1893] 1984, p. 288)

The mechanism, on this account, is the attempt to close the circle: “For if superstition means to bring action to a full circle and to make of it a closed system, if it tries to ‘loop’ human life in and to organize the perfect city where, hypothetically, everything would be end and means, then the most diverse forms of activity must concur in preparing and nourishing the faith and the cult” (Blondel [1893] 1984, p. 289). Superstition, so understood, is not a failure of intelligence but an excess of ambition in a finite work, and this is precisely the shape of the technological promise when it offers, in the same breath, a solution to mortality, to labor, to loneliness, and to meaning.4

Here Blondel supplies the word that the contemporary debate has appropriated and misused. Man “always has a superstition, because always he discovers in his action as it were a remainder for which he does not find the use” (Blondel [1893] 1984, p. 289). The remainder, in Blondel’s sense, is not a residue of capacities the machine has failed to replicate. It is the excess of willing over everything willed, which no achievement whatsoever removes, and which is disclosed most sharply by success rather than by failure. To look for it among the competences is already to have reduced the person to an inventory, which is the very move the defense was meant to resist; and this is why remainder humanism, which looks for the remainder in the shrinking list of what we can still do better than a machine, finds its ground eroded by every advance, while the remainder Blondel identifies is untouched by any advance whatever, having never been a competence in the first place. The status of this claim should be marked: that Blondel’s is the better description is not stipulated here but argued in Section 6, where the test is whether it can hold what is true in each rival account while explaining that account’s over-extension.

The word superstition is strong, and it would be worthless if it applied to every hope anyone has ever placed in a machine. It does not, and the criterion that limits it is already in Blondel’s account. What he describes is the attempt “to bring action to a full circle and to make of it a closed system” (Blondel [1893] 1984, p. 289), and what makes an investment superstitious is therefore not its size or its confidence but its structure. Ordinary technical confidence is proportionate, defeasible, and bounded: it expects a determinate result in a specified domain, it names in advance what would count as failure, and it leaves the question of what the work is finally for untouched, because it never purported to answer it. A radiologist who expects a classifier to catch tumors that the eye misses, an engineer who expects a compiler to be correct, a physician who expects a model to shorten the search for a drug candidate: none of these is superstitious, and all of them may be perfectly well founded. Nothing in this article counts against them, and it is worth saying plainly that the ordinary competence of technical work is presupposed by the argument rather than impugned by it. What marks the other disposition is the transfer of a question the work cannot take. The investment becomes superstitious at the point where the finite work is expected to close the circle: where it is offered, or received, as the answer to mortality, to labor, to loneliness, and to meaning at once; where no result would be allowed to count as failure, because the hope has detached itself from anything a result could settle; where the language of ultimacy attaches to an artifact. The difference is not one of degree along a single scale of enthusiasm. A modest expectation may be superstitious in structure if it treats a finite work as self-completing, and an enormous expectation may be perfectly sober if it remains defeasible and knows what it is for.

The most careful versions of the subtractive strategy do not look for the remainder in performance at all. They look for it in something the machine is not, locating it either in interiority or in the body, and I take their negative claims to be correct; together they mark the best ground the strategy has.5 But the positive claims belong on the other side of the line drawn in Section 4. Interiority is something a person has, and so is a body, and what one has can be inventoried; the subtractive method is perfectly willing to extend its list past cognition to consciousness and embodiment and defend that instead. Blondel’s remainder is not a further item for any list. It is the disproportion between the willing and everything willed, which is disclosed in the exercise of any capacity whatever, bodily or interior, and is therefore not the sort of thing another capacity could replace.

This distinction can be tested against experience rather than left standing as exegesis. Consider what is actually given at the moment a hard technical problem is solved. What appears is not repose. The engineer who has spent months bringing a system to the point where it finally does what it was built to do knows the experience precisely: the achievement, in the instant of arriving, discloses a range of intimations it does not itself satisfy. It is what Polanyi described from the side of knowing: a discovery is “fraught with further intimations of an indeterminate range” (Polanyi 2009, p. 23). The remainder is given in that experience rather than inferred behind it; it manifests itself in the moment of accomplishment. And it is given most sharply to those who succeed, which is why an account of the human built on what machines cannot yet do is looking for it in the one place from which every success will eventually remove it. The observation also yields a prediction that could in principle be checked, and I offer it as a prediction rather than as a finding, since I have no survey to appeal to and the claim is not one philosophical analysis is competent to establish: if the remainder is disclosed most sharply in accomplishment, then those closest to the making should be the least tempted to mistake the artifact for the power, having watched it fail for months. Whether the people who build these systems are in fact their soberest assessors is an empirical question, and the answer would bear on the account without deciding it.

Cathal Doherty’s study is the most sustained recent treatment of this material, and its governing thesis bears directly on the argument here. Doherty’s concern is “the specific difference between sacrament and superstition,” and his claim, against the Enlightenment reduction in the first to the second, is that “an abyss separates sacraments” from superstitious practice (Doherty 2017, p. xi). What distinguishes them is not the outward form of the act but whether the finite act is taken to produce what it signifies or to receive it. I extend the distinction to a case Doherty does not treat, and the extension needs its terms fixed before it can be used. In its proper theological sense, a sacrament is an efficacious sign instituted by Christ, which does not merely represent the grace it signifies but conveys it, and conveys it as something received rather than manufactured by the one who performs the rite; the sacramental character impressed by baptism, confirmation and orders is precisely not an achievement of the recipient but a mark conferred on him. Nothing of the kind is claimed here for any artifact, and the usage that follows is analogical and much weaker: it borrows from the sacramental structure only the single feature that a finite act may be held as receiving what it cannot itself produce. In that restricted sense, technological work is superstitious insofar as it is expected to produce the completion it gestures at, and sacramental insofar as it is held as receiving what it cannot produce. The same artifact, made by the same hands, falls under either description according to the disposition in which it is held. Nor does the scale of the achievement bear on which description applies. Even a technical accomplishment of the first order would leave the willing will unsatisfied, so that a superstitious reading of it is not merely premature but false.

The civilizational form of the first disposition has been named exactly, and it is the form that the argument of this article has been circling.6 John Paul II treats scientism not as one error among others but as a threat: the doctrine that “refuses to admit the validity of forms of knowledge other than those of the positive sciences” thereby “consigns all that has to do with the question of the meaning of life to the realm of the irrational or imaginary” (John Paul II 1998, no. 88). What makes the doctrine dangerous is less that it is believed than what it licenses. Where reason is narrowed to what the positive sciences can certify, persons come to be judged “by pragmatic criteria based essentially upon experimental data, in the mistaken belief that technology must dominate all,” and a technical capability ordered to no end beyond utility “could soon prove inhuman and even become potential destroyer of the human race” (John Paul II 1998, nos. 5, 81). The temptation then takes the form Blondel’s analysis would predict: certain scientists, lacking any ethical point of reference and sensing the opportunities of technological progress, seem to succumb “not only to a market-based logic, but also to the temptation of a quasi-divine power over nature and even over the human being” (John Paul II 1998, no. 46). That is superstitious action named from the other direction. A model of the willing is one more thing the willing has produced, and the producing is not among the things modeled.

None of this amounts to a case against building these systems. It amounts to a case about what building them cannot accomplish, and it is that negative result which turns out to be theologically productive.

6. Insufficiency as the Locus of Divine Action

Blondel’s argument does not end in the frustration it describes. The disproportion between willing and willed, pressed to its limit, yields a positive finding, and the finding is the hinge of the whole work:

In my action there is something I have not yet been able to understand and equal, something which keeps it from falling back into nothingness and which is something only in being nothing of what I have willed up to now. What I have voluntarily posited, therefore, can neither surpass nor maintain itself. It is this conflict that explains the forced presence of a new affirmation in consciousness; and it is the reality of this necessary presence that makes possible in us the consciousness of this very conflict. There is a “one thing necessary.”

(Blondel [1893] 1984, p. 314)

Blondel is careful about the modality of this result, and I mean to be careful about it in the same way. He insists that “there is nothing in this argumentation, absolutely nothing, that is a deduction”; the demonstration “does not result from a logical construction of the understanding” but is “a matter of grasping precisely what is already there” (Blondel [1893] 1984, p. 314). What philosophy establishes is a hypothetical necessity: an exigency that action cannot satisfy and cannot abandon, whose satisfaction, if it comes, must come as a gift. The scholastic name for the appetite in question is desiderium naturale, a natural desire whose object exceeds what nature can procure for itself. It matters that this is not the merely passive potentia oboedientialis of the extrinsicist account, on which the supernatural is superposed on a capacity that contributes nothing; Blanchette records that it was exactly such a purely passive obediential potency, addressed by an extrinsic dictate, that Blondel’s method was framed against (Blanchette 2010, chap. 7, p. 242). This is the sense in which Blondel spoke of showing “the natural necessity of the supernatural and the supernatural reality of the natural” (Blondel 1994, p. 44).

The modality deserves a closer look than it usually receives, because it is where the argument’s philosophical credentials are decided, and the sharpest pressure on it has come from within Blondel’s own tradition. Henri Bouillard, expounding Blondel against his critics, insisted that the passage from the indeterminate supernatural reached by the method of immanence to the Christian supernatural is not analytic but synthetic: nothing in the philosophical result entails the Christian determination of it, and the philosophy would be falsified as philosophy if it did (Bouillard 1969, pp. 66–67, 72). The point is easy to miss and decisive for the present argument. If the passage were analytic, the analysis of technological insufficiency offered here would be crypto-theology, and the charge of smuggling would stand. Because it is synthetic, the analysis stops exactly where philosophy stops: it establishes that technical action posits an end it cannot supply, and it leaves entirely open whether that end is met, by whom, and under what description. The theological reading of the opening is a further step, taken on other grounds, which the philosophy makes intelligible without making necessary. That is not a weakness of the argument. It is the argument’s precision about what argument can do. It also explains why the argument cannot be run in reverse: nothing here moves from God to a verdict on the technology. A reader who declines the theological identification keeps everything else: the critique of the subtractive defense, the relocation of the remainder, the diagnosis of the closed circle. The philosophy is hospitable to the theology without being hostage to it, and the hospitality is itself a philosophical result, established by Bouillard’s distinction and not by appeal to any doctrine.

Because this is the step the argument is most likely to be challenged on, it is worth setting out in order exactly what carries the weight and what does not. Four claims are in play, and only the first three are philosophical. The first is that action posits ends it does not reach: the willed will does not satisfy the willing will. Section 3 and Section 4 argued this from what is given in choosing and in succeeding. The second is that this shortfall is structural rather than technical, since it follows from the relation between immanent and transitive action and not from any contingent limitation of means; no improvement in the means touches it, because the improvement falls on the transitive side. The third is that the shortfall is not a privation to be repaired but a determinate openness: action does not merely fail to complete itself, it posits a completion under a description it cannot supply, and it goes on positing it. That much is a result about action, reached by analysis of action, and a reader who grants it has granted nothing theological. The fourth claim is the theological one, that the openness so described is where God acts, and it is not entailed by the first three. What the first three establish is that there is a place of a certain shape; that something occupies it is a further affirmation, made on other grounds, which the philosophy renders intelligible without rendering necessary. This is exactly Bouillard’s synthetic passage, and the reason the argument does not collapse into crypto-theology is that the fourth step is marked as a step rather than smuggled in as an implication. Blondel himself refuses the alternative that would make the passage easier and the philosophy worse, the retreat into an interiority purified of transitive work: he asks whether faith might find “a region of security and of rest where it is enough to adore in spirit and in truth, without mixing the miseries of transitive action in with this higher life,” and answers that this is “a subtle and widespread error,” since “it is, on the contrary, through practice that faith develops and is purified” (Blondel [1893] 1984, p. 372). The insufficiency is met in the transitive work, not away from it, which is why an account of technology and an account of divine action belong in the same argument at all.

Read against this result, each of the positions canvassed above holds insofar as it captures one aspect of an action that cannot complete itself. What has to be denied in each case is not the insight but its over-extension: the step from having perceived something real to having given the whole account of it. Remainder humanism is true insofar as it insists that something in the human is not captured by any machine, and false only in its localization of that something: the uncapturable element is the willing itself, the infinite power to determine oneself toward one finite motive out of indefinitely many, and not any of the products or competences in which that power happens to have left a trace. The enthusiasts are right that these systems genuinely extend human power; they are wrong only in supposing that extended power bears on the disproportion at all, since every new capacity is one more determinate object set against the same undetermined power to determine, and the more impressive the object, the harder it is to mistake for the power. Hanby is right that technology carries an account of being, and Polanyi right that it carries the person of its maker, and neither description can be dispensed with in favor of the other, since the first concerns what is made and the second who makes it. Del Noce is right about the social pathology that follows when a civilization denies the gap and reorganizes itself around the finite good it has mistaken for the infinite one, and those who locate the remainder in interiority or in the body right that no inventory of a machine’s competences touches what is at issue. What gathers all of these into one account, without asking any of them to give up what it saw, is Blondel’s insufficiency. This is the test Section 5 promised: the description that can hold each rival’s truth while explaining its over-extension is the better description, and the disproportion passes it while no list can. The portrait that results leaves every contributor recognizable in it, and reduces none of them to any of the others.

This is what allows the human–technology relation to be read as a place held open for God’s action, without either sacralizing the technology or making God a rival to it. If the insufficiency were a defect, it would call for a better artifact. Because it is structural, it functions instead as an opening: the point at which action, having exhausted what it can produce, is disposed toward what it can only receive. Thomas Pfau supplies the epistemic form of the same claim. A strictly conceptual and propositional outlook, he argues, could never by itself move the subject to action, since “authentic, intelligible action arises from internal motivation rather than external (efficient) causation” (Pfau 2022, p. 613). The distinction is worth pressing here, because it isolates precisely what a trained system does and does not do: whatever else such a system achieves, it produces outputs by external causation, and the internal motivation belongs to the persons whose act it is. The engineer’s work, on this reading, is neither an alienated output nor a substitute for grace; it is the ordinary medium in which a person’s willing is enacted, exposed as insufficient, and thereby held open. Doherty’s abyss between sacrament and superstition marks the difference in disposition that decides which of the two a given piece of work becomes, and the difference is not legible from the artifact alone (Doherty 2017, pp. 267–68). Blondel’s own remark on the idols is the appropriate deflation of the alternative: “what importance is there in the idols, more or less purified, which human intelligence manages to propose for itself?” (Blondel [1893] 1984, p. 315).

7. Objections and Replies

Four objections press on the argument, and they share a shape worth naming in advance: each, in order to be stated at all, must grant the very limit the argument claims.

First, that the analysis baptizes technological expansion. If every act of making is a locus of divine action, the objection runs, then no particular making can be criticized. The argument does the opposite. Precisely because every artifact is a determinate finite object of an infinite willing, none can bear the weight of an ultimate end, and any project presenting itself as bearing that weight is identified, on Blondel’s own terms, as superstitious action. Enrico Beltramini, arguing on Guardini’s ground, has shown that ethical protocols cannot forestall machines whose wrongdoing would constitute an existential risk, since in a technological culture the ethos of action has displaced the logos that such protocols would have to invoke, leaving ethics with nothing by which to govern what technology does (Beltramini 2021, p. 264). The constraint must then come from elsewhere, and the present analysis says where from the structure of the act that produces the artifact, which falls under judgment as any human act does, and not from rules bolted onto the artifact once it exists. The gap between the willing will and the willed will is what gives that judgment its content, and three constraints follow from it directly. The first bears on what a system is offered as. Any deployment presented as supplying what the user would otherwise have to will for himself, whether companionship, consolation, judgment or meaning, sets a willed object in the place of the willing; that is the superstitious structure exactly, and the offer is false whatever the quality of the output. The second bears on what a deployment displaces. A system that relieves a person of transitive labor takes nothing from him, whereas one that relieves him of an immanent act takes the act itself, and the student’s essay, the judge’s judgment and the physician’s diagnosis are losses of the second kind however great the gain in speed. The third bears on who answers for the result. Because the personal coefficient is displaced within these systems rather than removed from them, responsibility cannot be laundered through the artifact, and the sentence “the system decided” is not available to anyone who built or deployed it. None of these constraints waits on a forecast of capability, because none of them is a claim about what machines can do. This is why the analysis yields criticism rather than blessing: it supplies the question to put to any project, whether the making is held as a finite work or as a candidate completion, and Section 5 showed the second answer is always available and always false. Leo XIV applies the same test to the specific promise that the human condition is to be surpassed (Leo XIV 2026, nos. 115–17). The objection concedes that no finite work can complete action, which is the thesis it was raised against.

Second, that hypothetical necessity smuggles theology into philosophy. Blondel’s careful modality is the reply: philosophy does not establish that the supernatural is given, but only that action posits an end it cannot itself supply, and it is silent about whether that end is met. Blondel’s Thomist critics concede as much explicitly, since the concession is the premise of their own objection. J.-H. Nicolas grants that the method of immanence reaches an undetermined supernatural whose necessity is shown to purely rational reflection, so that “man without revelation appears to himself as unachieved, incomplete,” and objects that this compromises the gratuity of grace (quoted in Trethowan 1994, p. 100). Whether that further objection succeeds is not my question here, and Bouillard’s synthetic-passage point, set out above, is the standing reply to it: a gratuity compromised would be a gratuity entailed, and the method entails nothing. Nicolas and I agree about where the boundary between philosophy and theology falls and disagree only about what follows from it, and his objection is intelligible only if the boundary is where I have said it is. That a philosophical result should prove hospitable to a theological one is no defect in the philosophy, unless one has already assumed the scientism that Del Noce and John Paul II identify as the position needing defense rather than the position from which defenses are demanded.

Third, and most precisely: if every artifact is a volonté voulue, the description fits a hammer as well as a trained model, and the Blondelian apparatus does no work that a general philosophy of making could not do more cheaply. The objection is right about the genus: a hammer is indeed a determinate object of a willing that exceeds it. But it misses the specific difference. A hammer gives finite form to a determinate operation; a trained system is built to give finite form to the power of determination itself, not to one choice among many but to the capacity for choosing among indefinitely many. The disproportion belongs to artifacts per se, but is made manifest by this one alone, because only here does the willed will present itself as a candidate substitute for the willing will. A substitution must be available before it can be believed, which is why superstition attaches to this technology as it never attached to the hammer. What is on offer, in Aguas’s terms, is precisely the substitution he says cannot be made: no one can want for us. A system can produce what we would have produced had we wanted it, and that is a real and sometimes valuable thing; it cannot want it, and the difference is not a shortfall in the system but a category to which the system does not belong. D. C. Schindler reaches a convergent conclusion from the side of the artifact, arguing that the extreme case discloses the ordinary one: in the “grief tech” that reanimates a dead child as a chatbot we see not an aberration but “the very nature of AI, its inner logic, at work to some analogous degree in every use, no matter how apparently innocuous” (Schindler 2025). What such a system offers is a simulation of the one thing that cannot be simulated, and the offer is credible in proportion to how badly it is wanted. The credibility is not an accident of marketing: because the system is built to give finite form to the power of determination itself, its outputs arrive wearing the grammar of decision, and a thing that speaks in the first person invites the transfer of the one feature no output can carry. The hammer never tempted anyone this way, because the hammer never said “I.”

Fourth, an objection from the analytic side of the aisle: that the distinction between willing will and willed will is only a French dress for the familiar distinction between an act and its object, so that the apparatus imported here adds nothing a standard action theory does not already possess. The two distinctions cut differently, and the difference is where the work is done. The willed will is not the object of an act; it is everything determinate that willing has posited, our decisions and habits and works included, so that my own past acts stand on the willed side of the line just as their products do. An act/object distinction has nothing to say about the relation between the power of self-determination and the accumulated mass of its determinations; Blondel’s distinction is about exactly that relation, and it generates predictions the standard distinction does not. It predicts the phenomenology of choice recorded above, the felt disproportion at the moment of decision, since every determination is made against the background of the undetermined power that makes it (Blondel [1893] 1984, p. 134). It predicts the dilettante’s failure, since a refusal of all objects would still be an exercise of the power and so cannot escape the analysis. And it predicts what Section 3 called the settling of nothing: if the relation were merely act-to-object, there would be no reason why the most successful objects should sharpen rather than satisfy the demand, whereas on Blondel’s account that is just what they must do. A distinction earns its keep by what it explains, and this one explains what the standard apparatus leaves as brute facts.

8. The Conditions of the Thesis

It is worth stating exactly the conditions under which the thesis holds, since a claim without its conditions asserts more than it argues. The thesis requires, first, that these systems be artifacts in the sense defined at the outset, products of human acts of construction. No party to the debate denies this. It requires, second, that the disproportion between willing and willed be structural rather than technical, which Section 3, Section 4 and Section 5 argued from the experience of choosing and of succeeding, not from any doctrine about machines.

It does not require that machine consciousness be impossible, now or ever. That is a concession about what the argument depends on, not a prediction, and it should not be read as one. Nothing in the operation of these systems, in the hidden layers or anywhere else, gives any reason to hold that conscious thinking goes on there, and Section 2 gave the reason why: what a fitted distribution requires is stated in full by its construction, and an inner life is no part of that statement. Whether a made thing could be a subject at all is a further question, and a serious one, which this article neither settles nor needs to.

What the conditional concedes is therefore narrow, and it is worth saying exactly what it grants and what it does not. Suppose some future system were conceded an inner life. It would not thereby become human, and the distinction drawn here would not collapse, because that distinction was never between our species and its artifacts. It falls between immanent and transitive action, and so between an agent and a product. A system conceded an inner life would have ceased to be an artifact in the sense the first condition requires, a product of human acts of construction and nothing more, and would have become an agent of some kind; the question to put to it would then be the question to put to any finite agent, whether its willing exceeds what it wills. The answer would be yes, since the disproportion belongs to finite willing as such rather than to humanity as a species. Its productions would stand to its own willing exactly as ours stand to ours, and the disproportion would reappear within the machine rather than being abolished by it. The substitution of any will for a person’s would remain what Aguas names impossible, since it is substitution as such, not the substitute’s poverty, that the argument rules out: a second agent, however capable, still cannot want for me. What I hold, then, is that the thesis is indifferent to how the consciousness question is finally answered. What I do not hold is that the question is live for anything now built, or that a machine granted an inner life would be a human being rather than one more finite willer subject to the same measure.

Nor does the thesis rank technologies, forecast capabilities, or legislate about particular deployments; it concerns the structure of technical action, and the ethics of any given system is further work that the structure constrains but does not settle. These limits mark what the argument is entitled to, and an argument that knows its entitlements can be tested.

Finally, the thesis is answerable to failure conditions, and naming them is part of stating it honestly. It would fail if some willed object could be shown to have equaled and closed the willing that produced it, received as final and sharpening no further demand; the burden of exhibiting such an object lies with the sufficiency thesis, and every candidate so far offered has instead disclosed the gap again. And it would fail if the felt disproportion of Section 3 were shown to be an artifact of description rather than a datum of acting, a possibility the method takes seriously and answers in the only way a phenomenological claim can be answered: by returning to what is given in choosing and in succeeding, and inviting the objector to find it otherwise. Stated so, the thesis invites the right kind of disagreement: about the structure of action, which is its subject, rather than the pace of engineering progress, which is not.

9. Conclusions

The defense of the human by subtraction fails because it accepts the terms of a competition that no inventory of capacities can settle. Blondel’s analysis relocates the question. What distinguishes a person is not a residue of unmatched competences but a willing whose reach no work of that person’s hands has ever equaled, and which every genuine achievement discloses more sharply than any failure. Artificial intelligence is a limit case rather than an exception: the most ambitious volonté voulue yet attempted, and for that reason the clearest available demonstration that the willing will is not satisfied by what it wills.

The positive account this yields can be stated in a sentence, and it is what the subtractive strategy was reaching for and could not formulate. Human worth rests on the person’s standing as the subject of immanent acts, an acting that perfects the agent rather than a product that leaves him, and that no one can perform in his place. So stated, it is not a claim about what machines will never do, and it is therefore not exposed to the next demonstration. Whatever is built will fall on the transitive side of the line, and the worth in question was never located there. This is why the account does not need the ground it refuses to defend: it is not holding a position that advances in capability could overrun. Leo XIV states the practical form of the same judgment in Magnifica Humanitas, warning against a vision in which the human being is “treated as something to be perfected or surpassed,” since it “becomes easier to accept that some lives are less useful, less desirable or less worthy” (Leo XIV 2026, no. 117). The philosophical result and the pastoral warning meet at the point this article has been pressing: what cannot be surpassed is not a level of performance but a standing.

What this leaves is not a diminished account of technical work but a more exacting one. The engineer who writes a training objective is performing an act in the full sense, one that reveals her as its efficacious subject, expressing a personal coefficient she cannot fully articulate, and producing something that will shape how a civilization understands what a mind is. Held as an ultimate end, that work becomes the closed circle Blondel describes, and it will disappoint in proportion to the hope invested in it. Held as what it is, it remains open: a form of the ordinary labor by which persons enact their willing in the world and discover its measure. What that space asks of us first is not mastery but attention. When we look steadily at what we have built, what we perceive in the very impressiveness of it is that so much remains to be willed, and this is not the disappointment it is usually taken for. It is the beginning of wonder, and the point at which a philosophy of action reaches its limit and hands the question on to theology. A technology practiced as a vocation rather than as a cult would be, in Blondel’s sense, exactly where one should expect to find the action of God: not in the artifact, but in the space its insufficiency keeps open.7