Salvation by Compute: The Promise of Defeating Aging

We should begin by quoting it accurately, because the caricatured version of this promise is easy to demolish and is not the one in circulation.

In October 2024, Dario Amodei published Machines of Loving Grace, a text of some forty pages describing what he calls a compressed twenty-first century. The central claim: a sufficiently advanced artificial intelligence could fit fifty to a hundred years of biological progress into a span of five to ten. Infectious disease, cancer, genetic disorders, mental illness; the list is explicit, and so is the timetable. Ray Kurzweil, working in an older and more openly declared register, has long spoken of longevity escape velocity, the point at which life expectancy would gain more than a year for every year that passes, making death by aging technically optional.

Behind the texts there is money, and that changes the nature of the object. Calico, an Alphabet subsidiary founded in 2013 with an explicit mandate to take on aging. Altos Labs, announced in late 2021 and launched with roughly three billion dollars, hiring away a substantial share of the laboratories that matter in cellular reprogramming. Retro Biosciences, whose seed round Sam Altman funded single-handedly. These are not supplement shops. They are richly endowed institutions, run by serious scientists, with peer-reviewed research programs.

And then there is the note of exhortation, which is the most interesting detail in the corpus and the most revealing. Kurzweil made it the title of a book: Fantastic Voyage: Live Long Enough to Live Forever, written with Terry Grossman in 2004. The program there is explicit, organized around three successive bridges: hold on long enough on the first, the health regimen available today, to reach the second, biotechnology, which is in turn meant to carry you to the third. Aubrey de Grey supplied the genre’s most quoted line when he suggested that the first person to live a thousand years might already be alive. Ask these people what one should say to someone suffering today from an incurable illness, and the answer that keeps coming back is a variation on: hold on. Hold on a few more years. Whatever is killing you will be solvable in your lifetime if you last long enough to see what is coming. This is not a prediction. It is a consolation addressed to a dying person, and a consolation addressed to a dying person has a history, a form, a function. It belongs to a literary genre several thousand years old.

What sets this promise apart from the millenarianisms that preceded it deserves to be stated at the outset, since without it the rest of this essay would be pointless: it is made by people who actually ship systems that work. Amodei is not selling vapor. He is selling models that millions of people use daily and whose capabilities are advancing at a rate nobody anticipated, skeptics included. I have argued here before that the question was never whether to celebrate or lament this acceleration, but to understand what it displaces. It is precisely that operational credibility which rules out treating the subject with mockery, and which makes examining it necessary.

From cure to transcendence

The escalation happens at a precise point in the argument, and it is worth locating.

Curing disease is a conventional medical aim, continuous with two centuries of practice. Treating aging itself as a pathology is something else: it reclassifies as malfunction what the whole of human history had filed under condition. Aging becomes a decodable process with identifiable mechanisms, therefore with points of intervention, therefore, by the logic native to engineering, a solvable problem.

That reclassification is not free, and this is where the actual state of the science has to be laid out, not for the pleasure of detail but because it supplies the argument.

The levers exist. Senolytics, the compounds that selectively clear senescent cells, have moved from mouse models to early-phase clinical trials in humans, and then to a randomized placebo-controlled trial in idiopathic pulmonary fibrosis. Partial reprogramming with Yamanaka factors, which rejuvenates a cell’s epigenetic state without stripping it of its identity, has produced results in mice that nobody would have believed fifteen years ago, up to and including the restoration of vision after optic nerve injury. Epigenetic clocks, since Steve Horvath’s 2013 work, make it possible to measure a biological age distinct from chronological age, which gives the field something it badly lacked: a metric. And the mapping of the mechanisms of aging has grown considerably sharper over a decade.

All of this is true, serious, and published. In fact one should go further, because critics of techno-optimism consistently underrate this point: these intermediate results are more solid than most skeptics will admit. Partial reprogramming is not a laboratory curiosity; it is a mechanism that independent teams have reproduced. Epigenetic clocks are not a public relations gimmick; they transformed a field that until then had no way of measuring the thing it was talking about. Anyone who waves the whole body of work away has not read the literature, and discredits himself as thoroughly as the man promising immortality by 2035.

And yet the gap remains entire. Having mechanistic handles on a process is not the same as knowing how to reverse it. No demonstration of reversed aging in humans exists to date. No compound has been shown to extend the maximum human lifespan. The passage from mouse to man is, in the history of biology, the graveyard where most of the promises of the last thirty years lie buried.

The point is not that the promise is false. I have no idea, and neither does anyone else. The point is that the certainty plainly exceeds the available evidence, and that this excess, the space between what has been demonstrated and what is being asserted, is exactly the volume that faith occupies.

The counter-position, under the same scanner

Against this construction one can set a sentence of five words: God exists. This will not happen.

It is a position, and a coherent one. Aging and death are not design flaws in this view but constitutive elements of an order. Finitude is not a bug inherited from evolution; it is the very structure of created existence, the thing that makes transmission possible, and the succession of generations, and the seriousness of a life that is not played twice. The limit is a given, not a badly posed engineering problem. It is the same intuition that leads me to write, on an entirely different subject, that a child is not something you order: there are domains in which technical feasibility is not enough to establish legitimacy. On this view the techno-optimist enterprise is not merely doomed, it is categorically misdirected: it treats as a breakdown what is in fact a condition.

I should flag a tension with a position I have defended elsewhere. I have argued that the idea of a finite world is an illusion and that growth remains our only way out, which is a resolutely anti-limitational stance. The contradiction is only apparent: resources and human ingenuity have no known ceiling, the human condition does, and conflating the two is precisely the error I am examining here.

But if I stop the analysis here, I do exactly what I am about to accuse the other side of doing. One camp would be diagnosed, the other merely stated. So let us put the counter-position under the same scanner.

It too has its structures. A revelation, not derived from experience. An authority, not subject to revision. A community, dividing those who know from those who have not yet understood. And above all, an unavailability of proof that mirrors the one on the other side.

The symmetry is easy to state. The theological position predicts a failure that no intermediate result can disconfirm: if senolytics add fifteen healthy years to life expectancy, it will say that death remains, and therefore that nothing has changed. The techno-optimist position predicts a success that no failure can dent: if a major clinical trial collapses, it will say the horizon has moved out by three years.

Two systems of conviction whose common feature is that nothing that happens can get through to them.

An objection presents itself here, and it is better handled than left to the reader. The symmetry is not complete, and the asymmetry runs squarely against the theological camp. Laboratories fund experiments that can fail, and that do fail. They publish negative results, imperfectly but genuinely. They abandon lines of inquiry, redirect programs worth hundreds of millions, change their hypotheses. The theological camp has, by construction, no equivalent: nothing in its apparatus can go wrong, because nothing in it is put to the test. On that ground the comparison is crushing, and it would be dishonest to hide it.

But it does not bear on the same object. The research enterprise is falsifiable; the discourse of certainty wrapped around it is not, and that is the one I am talking about. A laboratory testing a compound is doing science. An executive asserting that aging will be solved within the decade, whatever that compound turns out to do, is doing something else. The symmetry is not between theology and biology, which would be absurd. It is between two ways of knowing in advance.

That is what deserves attention, far more than the question of which of the two camps will turn out to be right.

The diagnosis, and its weak point

The diagnosis makes itself as soon as you listen to radical techno-optimist discourse with a reasonably trained ear.

There is a promise of redemption: the end of suffering, of disease, and eventually of aging, which is to say the precise lifting of the curse that every theology concerns itself with. There is a saving figure, not a man but an intelligence, external to us, more powerful than us, whose arrival is awaited and dated. There is a teleology, a direction of history toward singularity or abundance, which gives meaning to the present and turns every passing month into an approach. And there is a division between those who have seen and those who have not yet seen, complete with the vocabulary of conversion, of waking up, of coming to realize.

One recognizes in passing the structure Huxley described better than Orwell: a servitude desired rather than imposed, a promise of technical happiness to which people assent out of relief. And one finds the same ideological closure Arendt identified in the great systems of the twentieth century: a logic that seals itself shut and absorbs every objection as further confirmation.

The word is available, and many people stop there. They are wrong, and it is better to say so in the body of the essay than to wait for the comments.

Because structural resemblance is not an argument. Exactly this was said of Marxism, with its fall, its redeeming proletariat, and its end of history. It was said of Enlightenment progressivism. And the fact is that the Enlightenment also promised the end of a vast share of human suffering, in a thoroughly messianic tone, and life expectancy did in fact double. We forget too easily, besides, that medieval abbeys were engines of progress rather than seats of obscurantism: religion and technical advance are not natural adversaries, and history is full of counterexamples for anyone inclined to make a rule of it. A belief can have the form of a religion and still turn out to be correct. To say that something is a religion and consider the matter closed is a genetic fallacy: judging a claim by the psychological origin of the conviction behind it rather than by what establishes it.

The criterion therefore has to move, and it is that move which rescues the diagnosis.

What is religious in this discourse is not the content of the promise. It is the relationship to evidence. A scientific claim states what would refute it: here is the experiment which, if it came out this way, would oblige me to abandon my position. That is Popper’s demarcation criterion, and it remains the most useful one we have. Put the question to a radical techno-optimist and watch the answer. What result, what failure, what barren decade would make you give this up? The question usually goes unanswered, or receives an answer that dissolves on contact. Every failure converts into a delay. Every setback becomes a scheduling problem, never a problem with the thesis.

This criterion also handles the most serious objection, which comes from Amodei himself. His essay opens with an explicit warning against grandiosity: he says he is put off by people who speak of the post-AGI world like a prophet leading his people to salvation, and he calls it dangerous to frame technological goals in essentially religious terms. The disavowal reads as sincere, and it does him credit. It changes nothing. It is not the author who decides whether his discourse is messianic; its relationship to evidence decides that. And the same text, a few pages later, suggests that once life expectancy reaches a hundred and fifty, most of those alive today might live as long as they wish. That is an address to the currently living, which is to say an exhortation, in a vocabulary that refuses to be one.

That is the criterion to hold to. Not resemblance in form, which proves nothing. Unfalsifiability, which proves something.

Why now

That leaves the question the diagnosis alone does not exhaust: why this faith, in this form, at this moment.

Three factors seem to me sufficient.

Secularization, first, from which the wrong conclusion was drawn. The retreat of religious supply never made the demand for eschatology go away. The need to know where history is going, to believe that suffering is not final, to hope that the dead are not simply lost, survived intact when the institutions that used to handle it emptied out. It was redirected. It had to go somewhere.

Stephen Cave has shown that these narratives are as old as the species and reduce to four patterns that recur from one civilization to the next. So it is not the narrative that is new. It is the budget.

The concentration of capital, second, which is the factor least often mentioned and the one that explains the most. The people funding the promise are precisely the people who would benefit from it first, and they are rich enough to buy it for themselves personally. The age of the founders is not a biographical detail: a generation of men who succeeded young and spectacularly is arriving at the point in life where finitude stops being an abstraction. They have, and this is historically unprecedented, the means to throw billions at it. The demand for eschatology has met a solvent supply. And the same actors turn up everywhere, including when the task is selling us universal basic income as emancipation: Silicon Valley does not merely produce tools, it produces the story that accompanies them, and the claim that technology is neutral is the first piece of that story.

A genuine jump in capability, third, and this is what rules out filing the whole affair under delusion. Something happened over the past five years whose scale surprised even the people working in the field, to the point that the builders themselves cannot fully explain what they have built. Ordinary millenarianism rests on a text. This one rests on a curve, and the curve is real. That is what makes the phenomenon vastly more interesting than its predecessors, and vastly harder to assess.

What is actually at stake

I am not settling the question. I do not know whether aging will be defeated, and I observe that neither those who assert it nor those who deny it know any better: both camps are describing a future that nothing currently in existence allows us to establish.

What I do observe, and what is verifiable, is that one of the great forms of faith of our era is taking shape in front of us, with everything that implies. Its founding texts, published and dated. Its institutions, capitalized. Its clergy, credentialed. Its faithful. And its exhortation to the dying, which is the surest sign that we are dealing with something other than a research program.

And since one has to end on the single thing I hold a firm position about: whoever writes the eschatology writes the agenda. A civilization that decides what it is waiting for thereby decides what it funds, what it regulates, what it teaches, and what it neglects. It is the same mechanism I described when I wrote about Musk buying not so much code as the right to say what is true. And this eschatology is being written entirely in California, in English, with American money, within a circle of a few hundred people who all know one another. Meanwhile France would rather manage its decline than build its future, and the only serious answer within our reach remains open, local infrastructure, which has at least the merit of not delegating to others the job of defining our horizon.

This “we” is not a geographic camp, and the point is worth making to head off the misreading. It includes the overwhelming majority of Americans. It designates everyone whose finitude is at stake and whose opinion is not being solicited, which comes to roughly eight billion people, minus the circle.

We have no voice in it. Not even the voice of doubt.