The Agent Haunted World

In 1995, Carl Sagan looked forward to the horizon of the 21st century - a technological watershed that was being greeted with near-universal optimism. But in that vision, moving just beneath the glossy surface, Sagan saw a shadow.

In 2005, I read what has become, to me, tantamount to a religious text: The Demon-Haunted World and saw, in my minds-eye, the rough shape of this specter.

In 2025, the shape snapped into focus. As deep-learning and foundational models were perfused into more facets of daily life, I began to grow anxious - not in the technology, but in the response. Sagan spoke of science as a “candle in the dark,” a small hard-won light held up into the vastness of the unknown. But our world is now flooded with a veritable sun of informational glow, and we are mistaking that artifice for enlightenment.

By 2026, that ‘demon’ is no longer a lack of information, it is instead the quiet, banal choice to stop being the ones to process it.

At its cold, mathematical heart, a Large Language Model is an exercise in probability. It is defined by the function:

Which is the likelihood of a token appearing based on the preceding sequence. Yet, in a blink, we transition from viewing these models as software to treating them as some kind of totem. What’s most jarring to me is not that the view is broadly following this kind of modern animism, but it is that a sizeable contingent of engineers and scientists - those tasked precisely with demystifying the world - are following suit. In overly-personifying the math, we seek to be excused from the labor of auditing and shaping its logic.

A note on Materialism: This isn’t a plea for human exceptionalism. I am, at heart, a staunch materialist in the tradition of Julien Offray De La Mettrie. I view the world as a vast, interconnected machine. But even La Mettrie’s materialism was a call to peer further into the gears, to understand ourselves in the context of the physical levers of existence. My point is that we are using the veneer of “sentience” to escape the burden of understanding those gears. It is, in fact, the same indictment I have of some parts of psychological science, but that is better saved for another time.

When the builders themselves, whether authentically or purely for marketing reasons, are seen “praying” to or “worshipping” the machinery, they provide cover for a broader culture already defined by a growing anti-intellectual vigor. In this socio-political landscape where “truth” is treated as a matter of team loyalty rather than based in observable evidence, this shift from “what” to “who” in technological attribution is a dangerous psychological surrender. And experts specifically ought not lead us into the dark, trading their light for fleeting adulation for shadow-play they no longer care to explain.

This psychological surrender leads inevitably to the ultimate end of intellectual erosion: the ceding of knowledge itself. Traditionally, true knowledge has been at least partially a product of friction. It is the grit found in a failed experiment, hours lost to research, and the cognitive strain required to synthesize disparate ideas into a unified truth. It is, at its heart, a process of labor.

When we uncritically accept truth from any external authority without the slightest curiosity, we are decommissioning our own mental machinery. By choosing the path of least resistance, we trade the depth of a realized thought for the convenience of a predicted one - definitionally average. The curiosity itself, the spark that took our species across the Bering Land Bridge, inspired us to refine the bark of willow trees to soothe pain, and to make machinery that carried us into the sky, and further to interstellar space, is the most precious facet of our experience, and most at risk in this decommissioning.

We are at a crossroads where efficiency has become a mask for intellectual slothfulness. We must realize that while this bright age of information can provide us with visibility, that does not automatically grant us vision. If we are, as Mettrie suggested, magnificent machines, then we must remember that a machine’s value lies in its function, not its idleness. To cede curiosity is to purposefully rust. It is a dereliction in the purpose hard-fought by ancestors who suffered the friction of the unknown so that we might have the luxury of the candle.