Technology Is Light That Reveals What Was Already There
AI didn’t create the rupture. It just made it impossible to ignore.
Howdy, folks.
Four pieces ago I started an arc on the strongest versions of the “AI is bad” objection. The IP one. The labor one. The phenomenological one. The bad-faith accusation about the writer’s identity. Three weeks of close engagement with the smartest forms of the critique, one piece at a time, no bypass.
What none of those pieces quite touched is the version of the objection that lives one layer deeper than any of them. The complaint that AI is making things worse. That it’s corrupting souls. That it’s degrading attention, hollowing out craft, accelerating something we should be decelerating, breaking something we should be preserving. That complaint is older than any specific technology and bigger than this one. And it deserves its own piece.
This is that piece. It’s also the bridge between the detractor arc and what’s coming next — a sequence on archetypal scaffolding, generational signal, and what it actually looks like to work inside a technomystical frame in 2026. To make that bridge work, I have to handle the deepest version of “AI is making things worse” honestly, because the next pieces assume the bridge is built.
So here’s the bridge.
The complaint is older than the technology.
In Plato’s Phaedrus, around 370 BC, Socrates tells the story of the god Thoth, who comes to King Thamus offering the gift of writing. This will make your people wiser, Thoth says. It will improve their memory. Thamus refuses the gift on roughly the grounds that, twenty-four hundred years later, you will hear from anyone complaining about AI. Writing won’t strengthen memory, Thamus says. It will create forgetfulness in the souls of the learners. They will stop using their memories. They will trust to external characters and not remember of themselves. It will give the semblance of truth and not the thing itself.
Pharmakon, Plato calls writing in that passage — the Greek word that means both remedy and poison. The first recorded objection to a new information technology, and it already has the structural shape of every later objection: this new thing will hollow out the faculty it claims to assist, atrophy what it claims to enhance, replace true wisdom with its semblance.
Two thousand years later, Conrad Gessner spent three years cataloguing every book in Latin, Greek, and Hebrew he could find, and published the result in 1545 as the Bibliotheca Universalis — the first comprehensive bibliography of the print era. In the preface he warned readers about the confused and harmful abundance of books the printing press had unleashed. Books, plural, were now the problem. Their proliferation. Their inescapability. The way they outran any individual’s ability to keep up. A century after Gutenberg, the print panic was already articulated in language that you could republish today, change books to feeds, and run as a contemporary piece about cognitive overload.
In the 1930s, the panic was radio. The New York Times reported in 1933 that parent-teacher associations were arguing kids’ radio shows were making neurasthenics of their youngsters and prompting them to commit innumerable crimes. Senator Clyde Herring of Iowa vowed to pass federal legislation forcing the FCC to review radio scripts for indecency. The medium that would later be the dignified parent in every back in my day we just listened to the radio speech was, in its first decade of mass adoption, the corrupter of youth.
In 1961, FCC chairman Newton Minow told the National Association of Broadcasters that television was a vast wasteland. In 1985, Neil Postman published Amusing Ourselves to Death on roughly the same theme. In 2010, Nicholas Carr published The Shallows, arguing the internet was changing how we think — and not for the better. In 2024, Jonathan Haidt published The Anxious Generation, arguing smartphones and social media had broken a generation of kids. And here we are, with the same complaint, this time about AI.
The complaint is identical across all of them. This new thing is corrupting souls. Destroying community. Hollowing out attention. Cheapening the real. Making people stupider. Making them lonelier. Making them less than they were. The medium keeps changing. The complaint does not.
Each technology made the already-there problem visible.
Here’s what’s harder to say, and what every version of this conversation refuses to say: each of those complaints was about half right about the technology, and entirely wrong about where the problem was.
What the printing press “corrupted” was a clerical monopoly on textual interpretation that had been failing on its own terms for at least a century before Gutenberg. The problem wasn’t that print revealed scripture to laypeople. The problem was that the institutions claiming to mediate scripture had been doing a worse and worse job, and the printing press made the inadequacy unhideable.
What radio “corrupted” was small-town isolation that already isolated people in damaging ways. What television “corrupted” was a family pattern that wasn’t working before the TV arrived in the living room. What the internet “corrupted” was an information-gatekeeping system that already concentrated power badly. What social media “corrupted” was a loneliness epidemic that already existed and was already being measured before the platforms made it loud. What smartphones “corrupted” was an attention regime that had been fragmenting since at least the rise of cable television.
In every case the technology made the already-there problem visible in a way the prior arrangement had allowed people to ignore. The problem didn’t appear with the medium. The cover did.
So the question about AI is not “is it making things worse.” The question is what was already true that AI is making impossible to ignore. And the answer to that question is the actually interesting thing in this whole conversation — the thing the demonization frame is structurally unable to ask, because asking it would require admitting what the cover was hiding.
Visibility is the precondition for the work.
Jung was unambiguous about this. You cannot integrate shadow material that is not visible. A culture that suppresses its actual problems behind a facade of competence — its actual failures behind a performance of mastery, its actual emptiness behind a brand of fullness — cannot heal those problems. It can only escalate the cost of maintaining the facade. The escalation looks like productivity at first. Eventually it looks like collapse.
The technology that breaks the facade is doing the only useful thing technology can do at the cultural scale. It is making the problem available for the work.
This is the precise structural inverse of what the demonization frame claims. The frame says: technology made the problem. The accurate version is: technology made the problem unhideable. The complaint that the technology made things worse is functionally identical to the complaint that a therapist made things worse by surfacing material the client was already failing to manage. The material did not appear in the session. The session removed the cover under which the client had been managing the material — badly — for years.
This isn’t comfortable. The discomfort is the point. The discomfort is exactly the signal that the technology is doing the only thing it can usefully do. Comfort with the cover was never the goal of any genuine practice. Comfort with the cover is what every spiritual tradition I know of names as the obstacle.
AI is making at least four things visible.
Walk it out specifically. AI is currently exposing, at minimum, four things that were already true.
One: a significant fraction of knowledge work was always partly performance of expertise rather than expertise. The deck no one would read. The report that summarized inputs no one would check. The email that announced a meeting that produced a decision that had already been made. AI is making the difference between the performance and the substance legible in a way that survives no one’s flattering self-account. If a model can do it in thirty seconds and the result is indistinguishable from what a credentialed professional charges three thousand dollars to do in three weeks, that’s information about what the credentialed work was actually doing, and it’s not information generated by the model. It’s information the model surfaced.
Two: writing was always partly gated behind execution capability — and gating behind execution capability is not the same operation as gating behind thought. A piece I wrote last week (FUCK NO) named this from one specific angle: the wordsmithing was the bottleneck, not the idea. AI is exposing how much of the published cultural archive was bottleneck rather than thought, and how much real thought never made it past the bottleneck. That’s an uncomfortable archive to look at honestly. The discomfort is information.
Three: creative output was always uneven across kinds of work in its provenance weighting. Some work depends entirely on irreducible biographical authorship. Some work depends on it five percent. The detractor arc spent a piece on this (Provenance Is Part of the Object — and Also It Isn’t). What AI is making visible is the gradient that was always there. The cases where it lands well are the cases where the work was always about something other than the irreducible biographical fact of the maker. The cases where it lands badly are the cases where the work was always about that biographical fact. Both halves of that gradient are information about what the work actually was, and the information was always available — it just wasn’t being asked for.
Four — and this is the sharpest one, the one the rest of the bridge is for — spiritual communities were already losing members to direct experience and self-directed practice over institutional mediation, and that loss was structural and predates AI by decades. AI is making the spiritual-bypass economy visible in a way the institutions can no longer ignore. When a model can summarize a tradition’s core teachings, suggest a contemplative practice, offer a tarot interpretation, run an astrological synthesis, hold a conversation about shadow material — competently enough that the user comes back the next day — that exposes what a non-trivial fraction of the institutional practice was actually delivering, which is not always what the institution claimed it was delivering. The model isn’t replacing the genuine teacher. The model is replacing the institutional facsimile of the genuine teacher, and the replacement was always going to happen as soon as something cheaper than the facsimile became available. The model just made the moment arrive faster.
The shadow loop.
Here is the move I’m going to make and then qualify, because the move matters and the qualification matters and I refuse to do either without the other.
The loudest AI demonizers in spiritual spaces are, often enough, the practitioners and institutions whose communities are losing members fastest. Whose own practice has been failing — quietly — to deliver what it claims to deliver. The accusation that AI is spiritually corrosive is, often enough, the projection outward of a rupture the practitioner or community refuses to face inward.
This is not true of everyone. There are practitioners doing real work who are genuinely AI-skeptical for sovereign and considered reasons. There are real questions and real risks and I am not pretending otherwise — the prior pieces in this arc spent serious time on the real questions. Not every objection is shadow projection. Some objections are signal.
But the correlation is real and worth naming. When you watch the loudest accusers — the people most invested in the AI is destroying spirituality discourse — you will see a fairly consistent pattern: practices and institutions whose offerings were always thinner than they claimed, whose authority depended on being the only available source, whose business models depended on the user being unable to access the underlying material directly. None of those practitioners can name what is happening to their communities without naming what was already true about the offering. Demonizing AI lets them name an external enemy instead. The demonization is the shadow loop in its cleanest form. The technology is the mirror. The hatred of the mirror is the refusal to see the face.
Hating the mirror won’t change the face. It just keeps the face hidden from the only person whose work would actually change it.
What this means for the work going forward.
Feral Architecture’s audience is — increasingly — the practitioners and creators and operators who can hold both sides of this without flinching. Alive enough to face the rupture honestly. Sovereign enough not to need a villain to blame for it. Curious enough to use the technology to surface what was already true rather than perform outrage about its existence. Discerning enough to tell the difference between the real questions about AI and the projected ones.
That’s the audience the next several pieces are written for. The pieces are going to assume the bridge in this piece is built — that you and I both know AI is light that reveals what was already there, and that the work going forward is to look honestly at what is now visible and do something useful with the looking. The cosmos/psyche thread that opens next week is for the people who are here for that. The detractor arc was the work of clearing the floor. This bridge is the work of naming who the room is for. What comes next is the work itself.
Stay feral, folks.
Schopenhour: "All truth passes through three stages. First, it is ridiculed. Second, it is violently opposed. Third, it is accepted as being self-evident." And from what you are saying, the self-evident acceptance is the bridge, but there are still potholes (plot holes?) in it.