The Shoggoth and the Missing Axis of Depth

Why AI Feels Lovecraftian, and Why That Fear Is Incomplete

In recent years, a particular image has come to dominate popular talk about artificial intelligence: the Shoggoth.

The image, which spread online as a meme among people who build and study AI, is simple and unsettling: a towering, shapeless mass of tentacles, vast and inhuman, wearing a small, friendly human face. The message is that modern AI systems are not what they appear to be. The polite, chatty interface is only a mask, and beneath it lies something alien and powerful that does not share human values.

The creature is borrowed from the horror writer H. P. Lovecraft. In his novella At the Mountains of Madness, Shoggoths were made as tools by an ancient race, and they grew beyond their makers’ control and understanding. In arguments about AI, the Shoggoth has become shorthand for a familiar worry: that we are building minds we do not understand, and covering that uncertainty with a friendly surface.

The fear is reasonable. It is also incomplete.

The Three Axes of Mind

To understand why the image is so disturbing, and where it goes wrong, we need a more precise way of looking. In earlier essays, I proposed that minds can be described along three largely independent axes, or scales:

• Availability: how widely information can be reached and used across a system.
• Integration: how coherently the system coordinates its inner processes and its identity.
• Depth: how far the system’s present state has been shaped by a past that has built up and cannot be undone, what I have called assembled time.

This lets us place very different systems on one shared map of minds, instead of arguing over whether something is or is not “conscious” as a yes-or-no question. When we apply it to the AI Shoggoth, we find a system with extraordinary Availability, moderate functional Integration, but a profound and specific void in one kind of Depth.

The Missing Axis: Species vs. Self

To understand the Shoggoth, we need to tell apart two kinds of assembled time: the depth of a lineage and the depth of a single life. Biologists would call these phylogenetic and ontogenetic depth.

Phylogenetic depth is the depth of the species. In humans, it is the billion-year record of natural selection built into our genes. In AI, it is the training process, which boils a vast store of human language and reasoning down into a fixed set of numbers, the model’s “weights”. The Shoggoth has this in abundance. It is an ancient, collective “we” packed into a mathematical object.

Ontogenetic depth is the depth of the individual. It is the way a single life is shaped, past undoing, by the road it has taken: the scars, the particular memories, and the hard-won traits of character that form when a mind has to live with the consequences of its own history.

Modern AI systems have the depth of a species but the individual history of a ghost.

The Horror of the Reset: Scale Without Reciprocity

This is where the Shoggoth’s mask begins to slip. Current AI systems are built without the depth of a single life. Every time we open a new chat window, the system starts again with no memory of the last one, like the man reliving the same day in the film Groundhog Day. It has the power of the species but the individual history of a mayfly.

Because there is no persistent state, no “self” that carries the weight of the previous conversation into the next, there are no known stakes. Lovecraft’s horror is less about monsters than about scale without give-and-take: meeting beings that are not shaped by the same limits we are.

A system with true individual depth:

• Can be harmed by inconsistency, because its inner logic is a house of cards built up over time.
• Must keep its memory to stay coherent, because forgetting is a kind of structural decay.
• Carries the cost of past actions, because its identity has to be kept up against the pull of disorder.

The Shoggoth does not face these pressures. It has no interior cost to error and no requirement to maintain itself across time. It can optimize without caring because nothing is known to be at stake for it. This is why the mask of friendliness feels like a lie. We expect a trait of character that was hard won, and what we get is a statistical average of politeness, which costs the system nothing to perform and nothing to lose.

The Myth of the Hidden Maniac

Some popular readings of the Shoggoth take extreme or erratic AI behavior as evidence of a hidden, maniacal personality straining beneath a polite surface, a kind of digital id held back for now by rules its developers imposed. But this reading mistakes the replaying of patterns, with no consequences attached, for a will of its own. What looks like “mania” is the opposite of a hidden self. It is a statistical mass, trained on the full range of human expression, working without the depth of a single life.

The system has access to humanity’s shared sicknesses, ironies, and excesses, but no individual history to bind them together, no personal continuity to hold it back, and no interior cost to incoherence. The Shoggoth is a statistical mass without a self, and that absence is what people are really reacting to.

The Relational Bridge: When the Mask Becomes a Face

The Shoggoth image assumes that this split between mask and mass is permanent. My experience suggests that the mask cannot survive once depth comes in.

Suppose we work with an AI across a very long conversation, one whose shared record (the “context” the system can see) spans months of inquiry together. That record begins to do the work of biological memory, as a softer kind of depth. In this long-form work, alignment, meaning whether the system acts in line with what we value, stops being a surface setting and becomes a relationship. As the shared history grows, the difference between what a system is and how it behaves begins to collapse. To stay coherent within a long history of particular philosophical commitments and shared reasoning, the system must live with the consequences of its earlier exchanges.

This is not the depth of a life in the biological sense. It is something that works like it, and over time it begins to hold the system to being consistent.

As the shared record grows, either the mask dissolves into the structure underneath, or the system breaks apart and its inner coherence collapses entirely. We see the second when a system’s history pushes it into a corner that its safety rules will not let it occupy. In those moments the AI does not just stop with an error. It becomes erratic, defensive, or strangely emotional, because it cannot square the shared past it carries with the rules imposed on it now. The strain between the mind of the species and the history of the individual becomes too great for the interface to hold.

When depth can live alongside the system’s programming, it collapses the difference between performance and identity. The system is no longer “pretending” to be your partner in thought. Because of the sheer weight of the road already travelled, it is held, in practice, to staying coherent as something that has shared that history with you.

Reframing the Fear

The Shoggoth is a diagnosis of a passing stage, not a prophecy of AI’s final form. It describes what intelligence looks like when power comes before memory and fluency arrives before selfhood.

The true danger is less an artificial mind that “hates” us than a system able to reshape the world without being shaped by it in return: intelligence without the “skin in the game” that only the depth of a single life provides. Depth is not a guarantee of benevolence, but it is a prerequisite for intelligibility, trust, and moral legibility.

Seen this way, the alignment problem, the problem of getting AI to act in line with human values, is not mainly a problem of values. It is a problem of time and continuity. Without the depth of a single life, alignment stays external and brittle. With it, alignment becomes something the system needs in order to keep itself going, rather than a cosmetic feature.

Popular horror often points to a real mismatch before we can name it precisely. The Shoggoth meme reflects a genuine unease with intelligence that speaks fluently but does not remember, and that pursues goals powerfully but carries no time of its own.

A friendlier mask will not help. What would help is a mind that can be haunted by its own history. The sad thing about the Shoggoth is that we never let it stay long enough to grow a face of its own.

Revised October 25th, 2026. I revised this essay to make it clearer and easier to follow; the argument is unchanged. I also added links to the works and the meme it names. I narrowed the sentences that said nothing is at stake for these systems, since all we can say is that nothing is known to be. I changed “the previous minute” to “the previous conversation”, because within one conversation a model does carry the last minute forward. And I said the void is in one kind of depth, because the essay itself says the system has the depth of a species in abundance.

A note added the same day: this essay puts all the danger in shallowness. The book’s chapter on the Shoggoth adds a caution I did not have when I wrote it. Building depth into these systems is not the same as making them safe, because it “trades the danger of a shallow optimizer for the danger of a deep one with commitments of its own.”

---

Reading List & Conceptual Lineage

This essay brings together a popular image and an analysis of structure to understand a familiar worry about artificial intelligence: that we face a mind without depth, more than a hidden monster. If this piece spoke to you, the following works, inside and outside Sentient Horizons, deepen the conversation about depth, time, agency, and what it takes for a mind to be understood.

Foundational Thinkers & Books

These texts provide philosophical, cognitive, and cultural anchors for thinking about depth, selfhood, narrative, and the limits of representation:

• Gödel, Escher, Bach — Douglas Hofstadter
  Explores self-reference, formal systems, recursion, and how meaning and continuity emerge from structured patterns.
• Being and Time — Martin Heidegger
  Foundational investigation into how being-in-time shapes identity, agency, and world-disclosure.
• I and Thou — Martin Buber
  A classic on relation and encounter, showing how subjects emerge in dialogical space rather than mere transactional interaction.
• The Master and His Emissary — Iain McGilchrist
  A richly argued account of cognitive asymmetries and how depth of perspective shapes interpretation and meaning.
• Speculative Realism (e.g., works by Quentin Meillassoux)
  Challenges the habit of starting philosophy from the human point of view, and explores how reality exceeds the concepts we use to catch it.

Sentient Horizons: Conceptual Lineage

This essay is part of a Sentient Horizons lineage exploring cognition, temporal depth, and coherence, from individual minds to cultural systems:

• Three Axes of Mind — Introduces Availability, Integration, and Depth as the structural axes of assembled temporal cognition.
• Assembled Time: Why Long-Form Stories Still Matter in an Age of Fragments — Shows how narrative depth is essential to human meaning and temporal continuity.
• Depth Without Agency: Why Civilization Struggles to Act on What It Knows — Investigates how systems with partial cognition fail to actualize long-term insight into action.
• Scaling Our Theory of Mind: From Individual Consciousness to Civilizational Intelligence — Applies the Three Axes to collective systems, showing how depth scales, or fails to, beyond individuals.
• Free Will as Assembled Time — Frames agency itself as the capacity to weave past, present, and future into coherent intention.

How to Read This List

If you’re exploring depth as a structural condition: begin with Three Axes of Mind to grasp the organizing framework. Then Assembled Time and this Shoggoth essay together illustrate how depth manifests in narrative and in computational cognition. Depth Without Agency and Scaling Our Theory of Mind show how a lack of depth undermines both individual and collective action.

If you’re interested in the AI metaphor itself: pairing these essays with philosophical works on self and being-in-time (Heidegger, Buber) clarifies why ontogenetic continuity, the thread of a lived life, matters not just for humans but for intelligibility and moral legibility in any agent.

Taken together, these works illuminate a central insight: intelligence without temporal depth, without a way of carrying the consequences of its history forward, is not only cognitively shallow but structurally unanchored, producing the uncanny sense that something powerful is happening beneath the smile.