The Danger and the Turning: Against the AI Death Cult

The prophets of artificial intelligence increasingly divide into two camps: On one side are those who promise salvation through intelligence: abundance, automation, the end of work, perhaps even the transcendence of biological limitation. On the other are those who foresee extinction. They tell us that artificial intelligence will become too powerful to control, that humanity must race to construct a machine more intelligent than itself while simultaneously fearing the moment it succeeds.

These positions appear opposed, but they share a deeper assumption. Both imagine intelligence as something that must become larger, more centralized, more autonomous, and more powerful than the human beings who created it. The disagreement concerns what this machine will do to us. Almost no one asks whether this relationship to intelligence is itself the danger. This is where we should begin.

The Danger

A person in a dark room holds a phone up so that it hides their face, its cold blue glow on their hand, a harsh light above.

In The Question Concerning Technology, Heidegger insisted that the essence of technology is not technological. The danger of modern technology is not simply that machines become powerful. It is that the world begins to disclose itself according to the logic of the machine. He called this Gestell: Enframing.

Under Enframing, things cease to appear primarily as things in themselves. They appear as resources waiting to be ordered. The forest becomes timber. The river becomes hydroelectric potential. The worker becomes human capital. Attention becomes engagement. Behavior becomes data. Increasingly, thought itself becomes compute.

Artificial intelligence represents an extraordinary intensification of this process because the resource being ordered is no longer merely the external world. It is human cognition itself. Every conversation becomes training data. Every preference becomes a signal. Every act of writing, searching, speaking, buying, moving, remembering, and desiring can become material from which increasingly precise models of human beings are constructed.

The human being becomes both the operator of the machine and its raw material. We believe we are using the machine while its infrastructure learns how to reproduce the very operations for which we use it. This is Enframing turned inward.

The Mirror

An old mirror on a dark wall, reflecting an empty chair.

This is why the problem of AI is ultimately a problem of the image. The model constructs an increasingly elaborate representation of the human being: our language, desires, memories, habits, relationships, preferences, and modes of thought. The representation becomes useful precisely because it resembles us. But the image in the mirror is not you.

This distinction becomes politically decisive when the representation is possessed by someone else. The danger is not merely that artificial intelligence might someday become conscious. The more immediate transformation is that enormous computational systems are becoming capable of modeling human beings while the human beings being modeled possess neither the model nor the infrastructure on which it exists.

We are producing digital doubles that we do not own. Whoever possesses the double acquires an extraordinary capacity to predict, simulate, persuade, classify, and eventually act upon the person it represents. The mirror becomes an instrument of administration.

Intelligence Becomes Territory

A path running into a dark forest at night, one warm light far down it.

This changes the meaning of cyberspace. The early internet was often imagined as a placeless domain. But cyberspace has developed its own geography: protocols, platforms, jurisdictions, chokepoints, identities, cryptographic boundaries, networks, servers, models, and computational territories. It has become a political space, and every political space eventually produces its characteristic political actor.

In Theory of the Partisan, Carl Schmitt described the partisan as fundamentally telluric: attached to the earth, operating from an intimate knowledge of a territory that conventional power cannot completely master. But what happens when the territory itself becomes digital?

The hacker emerges as the partisan of cyberspace. The pseudoanon becomes telluric to a territory that has no soil. Keys replace walls, protocols replace borders, repositories replace arsenals, and forks become forms of exit. The partisan no longer disappears into the forest. He disappears into mathematics.

Beyond the Line

This is why cryptography matters far beyond secrecy. Cryptography produces something historically unusual: the possibility of creating boundaries that do not depend entirely upon political recognition. A state can declare that you possess a right to privacy. Cryptography can make particular information mathematically inaccessible without a key. These are fundamentally different forms of protection. One is juridical. The other is architectural.

Cryptography therefore raises a radical question: What if sovereignty could become a property of systems rather than merely a promise made by institutions? This is the significance of going beyond the line.

The frontier of cyberspace allows the construction of spaces in which older forms of territorial control become incomplete. Servers remain physical, networks remain vulnerable, and states remain powerful. But their control is no longer total. A cryptographic system can create domains whose rules are partially enforced by mathematics rather than exclusively by sovereign command.

This is why the history of cryptography continually becomes political. To control cryptography is to control who may possess secrets, authenticate themselves, transact, communicate, and disappear from systems of observation. Cryptography redistributes the ability to draw boundaries. Sovereignty begins with the boundary.

The Closing of AI

A long corridor of server racks in cold blue light, one warm light at the far end.

The development of artificial intelligence is moving in precisely the opposite direction. The early culture of computation contained a powerful tradition of openness: free software, open protocols, hacker culture, permissionless experimentation, systems that could be inspected, modified, forked, and possessed by their users.

Open source made something remarkable possible. The means of digital production could be copied. Unlike the industrial factory, software does not have to be seized physically to become common. Once released, it can proliferate at nearly zero marginal cost. Open source therefore represents a peculiar historical form of appropriation: property freed from exclusive ownership and returned to use.

Artificial intelligence threatens to reverse this trajectory. As models become enormously expensive to train and operate, intelligence risks becoming concentrated inside a small number of corporate and state infrastructures. You do not possess the machine. You receive access to it. You do not possess the model of yourself. You authenticate into someone else's model. Its memory is not yours. You request that it remember. Its boundaries are not yours. You accept its terms.

The personal computer becomes a terminal again. Only this time the centralized machine is not merely storing our files. It is participating in our cognition.

The AI Death Cult

This is the strange theology beneath contemporary AI discourse. We are told that intelligence must be scaled toward something approaching omniscience, that enormous quantities of human knowledge must be gathered into centralized models, that these systems must become increasingly agentic, and that human beings will become progressively dependent upon them. Then, having constructed this architecture, we are asked to contemplate whether the resulting intelligence might destroy us.

The proposed solution is usually more control at the center: more alignment, more surveillance of the models, more restrictions on who may possess them, and more concentration of computational power among institutions supposedly capable of managing the danger. But this can deepen the original problem. The danger may not originate solely in the intelligence of the machine. It may originate in the relationship of dispossession through which intelligence is being constructed.

Humanity does not need to create a god and then pray that its owners remain benevolent. There is another trajectory.

The Turning

At the decisive moment in The Question Concerning Technology, Heidegger turns to Hölderlin:

“But where danger is, grows
The saving power also.”

The line is Hölderlin's, from Patmos, but Heidegger makes it central to his confrontation with technology. The saving power does not simply arrive from somewhere outside technology. Heidegger searches for it within the danger itself. That distinction matters.

The computational systems capable of producing unprecedented forms of surveillance and administration also contain the technical possibility of unprecedented individual sovereignty. Models capable of constructing our digital doubles can instead be possessed by the people they represent. Cryptography that secures money can secure memory. Open source can do to intelligence what it did to software. Advances in hardware can bring increasingly powerful inference out of centralized data centers and into homes, offices, communities, and machines controlled by their users.

This is the Turning. The answer is not the destruction of artificial intelligence. It is a transformation in our relationship to it.

From Artificial Intelligence to Sovereign Intelligence

Imagine instead that your artificial intelligence lives with you. Its memory belongs to you. Its model of you belongs to you. Its private keys belong to you. Its conversations are not an exhaust stream flowing toward somebody else's data center.

It can reach outward into the network without allowing the network to reach inward into you. It can use frontier intelligence when necessary while maintaining a cryptographically protected interiority that no external model automatically possesses. The architecture becomes modular. Intelligence can be local, shared, public, or ephemeral. Information crosses the boundary because you deliberately disclose it, not because disclosure is the default condition of participation.

The crucial question then changes.

How intelligent is the model?

becomes

Who possesses the relationship?

That may be the more important question. Intelligence without sovereignty can become administration. Intelligence joined to sovereignty can become augmentation.

The Saving Power of Cryptography

Cryptography offers the technical foundation for this alternative relationship. Encryption establishes interiority. Digital signatures establish agency. Keys establish possession. Zero-knowledge systems allow proof without complete disclosure. Open protocols permit interoperability without requiring a single sovereign platform. Local computation allows cognition to occur without mandatory exposure. Open-source models make intelligence inspectable, modifiable, and forkable. Together these technologies make possible something that the centralized AI paradigm cannot easily provide: a digital self that remains under the authority of the person from whom it emerges.

Privacy, in this conception, is not a preference buried inside a settings menu. It is architecture. Sovereignty is not granted by a platform. It follows from possession. The question is no longer whether we are permitted access to intelligence, but whether we own the conditions under which intelligence encounters us.

The New Partisan

Here the hacker returns. The hacker is no longer merely the intruder who breaks into systems. At the deepest level, the hacker is the person who refuses to accept the given limits of a technological system as ontologically final. The hacker opens the machine, studies it, modifies it, forks it, reappropriates it. Technology ceases to be something merely administered and becomes something inhabitable.

This makes the hacker, and increasingly the pseudoanon, the characteristic partisan of the emerging digital territory. The point is not that the hacker seeks permanent war. The partisan appears wherever a centralized order cannot completely absorb the territory into itself.

The hacker preserves the outside: the possibility of exit, secrecy, another identity, another protocol, another machine. Above all, the hacker preserves the possibility that intelligence itself can be possessed differently.

A Home for Intelligence

A cabin in a dark forest, its windows lit warm.

The decisive struggle over artificial intelligence may therefore not concern whether machines become smarter than human beings. It may concern something much more immediate: Where does intelligence live? Does it live exclusively inside gigantic computational fortresses controlled by corporations and states? Or can it live with us?

Can intelligence have a home, a dwelling, a boundary, an interior and an exterior? Can we construct machines that know us deeply without requiring us to surrender ourselves to the institutions operating them? If we can, then the personal AI becomes something fundamentally different from the centralized assistant. It becomes a cryptographic dwelling. A protected space in which human and machine cognition can develop a persistent relationship without that relationship automatically becoming the property of another sovereign.

Perhaps this is the genuine saving power hidden inside the danger of artificial intelligence. The task is neither to stop intelligence nor worship it. It is certainly not to surrender ourselves to it. The task is to possess our relationship to it.

The question of AI therefore returns us to the question that has haunted cyberspace from its beginning: Who owns the keys? Whoever owns the keys determines where the boundary lies. Whoever can draw the boundary still possesses the possibility of an outside. In an age attempting to transform everything, including thought itself, into standing reserve, that outside may become one of the most important forms of freedom we have left.

Further reading: What Vora Is · The Architecture of Return · Not Your Keys, Not Your Mind

First published on X, September 16, 2026.

P.S. Would you like to reserve your Aegis now? Click here.