Ziggy Stardust and Hypernormalization in the AI Era

- 8 minutes read - 1560 words

A few posts back, I referenced Bowie’s opening track, “Five Years,” from “The Rise and Fall of Ziggy Stardust and the Spiders from Mars (1972).”

In this song, humanity has been told Earth has 5 years left. This catastrophic message and the song’s minor-key, elegiac quality lay a foundation upon which the narrator describes ice cream parlors (hedonic commitment after the weighty news) and public weeping. Against the looming Big End, people keep doing what they do in the most bourgeois and pedestrian way because, well, in the words of Bowie’s fellow Englishmen Pulp: “there’s nothing else to do.”

I really feel like that sentiment connects to the current AI moment. Sometimes I’m elated and sometimes I’m horrified. After whipsawing back and forth I too want a milkshake like Bowie described.

This idea has been called “hypernormalization:” the way we all keep going through the motions we’ve always gone through, because this is the only way we’ve ever known how to conceptualize life, even though the ground is actually falling out from under our feet.

We’ve already lived through a smaller-scale version of this. The internet still looks like the internet — the same tabs, the same feeds, the same muscle memory — while quietly having been gutted into a commercial hellhole underneath3 . Nobody declared the old internet dead; we just kept clicking through the motions of a thing that had already changed shape beneath us.

If AI pessimists are correct, we have fewer than “Five Years” before superintelligence remakes the world in a way we will either not be a part of, or not recognize.

The Mania Beneath the Motions

Here’s the first extreme to the whipsaw motion: “AI is bunk and leaders are in a mass psychosis triggered by a coordination problem.” There’s a line in “Five Years” for this: “and I felt like an actor.” Not devastated, not resigned — performing acts of fealty and agreement that we have to race for the cliff’s edge too just as hard as the next team/country/organization.

Nikhil Suresh, who runs a data consultancy and has logged roughly 300 conversations with professionals across industries over the life of his blog, argues that corporate AI adoption has become a mass psychosis rather than a rational strategy. His core claims:

  • Every AI project his team has observed in a year and a half — whether they were hired to work on it or just watched it happen — has failed. Internal chatbots go unused because the underlying documentation is bad; customer-facing chatbots are broadly unpleasant to use; and success metrics are either not tracked or are easily gamed.
  • Raising doubts about AI strategy inside a company has become career-ending. Executives and employees alike are pressured into public professions of faith in AI’s transformative power, even when they privately admit their org uses it for nothing. Engineers now do the work manually and simply claim an LLM did it — “AI-washing” — to satisfy management. Some game “token leaderboards” by looping prompts to nobody while the real output goes unused.
  • Executives are stuck in a coordination problem: none of them can admit publicly that AI gains are overstated without implicitly calling their peers, customers, and vendors liars, risking their jobs or their companies’ contracts. So the exaggeration compounds indefinitely, because no single actor can break rank first.
  • Many “AI projects” are actually ordinary projects with an AI label bolted on to survive internal politics. Suresh gives an example of a database migration where an LLM translation step failed, was quietly redone by hand, and then reported upward as an AI success.

This leads to a cynical addling for the rank and file. They know the execs are hyping that which isn’t true, yet they see that the emperor has no clothes. They are living with the constant gaslighting against their actual lived experience and whipsaw to “this is bullshit.”

The Terror That Isn’t Close Yet

Now here’s the other extreme: dislocation is assured, just not close like an inferno. It’s close like a climate catastrophe. Bowie’s song’s five years work the same way — a deadline and a permission slip at once, a countdown far enough out that you can still go on living inside it, but you’d also be forgiven for asking “Why fucking bother?“1

I got the same feeling watching Kyiv women talking about dating in wartime under actual, occasional bombardment. The terror is all around and death is random and stalking (like it always is), but not abstract: not the “abstract, Sylvia Plath way” the narrator from Fight Club means when he says it.

The terror comes from considering the moment that the flubbed projects stop and the shop starts improving itself recursively to stop all possible failed projects and recursively improve itself, tirelessly, all hours of the day unto a level of capability unassailable by human effort.

And that cynical knowledge worker who’s walking around with eyes rolled in their head as the CXO talks about innovating paperweights with AI, also knows that something is happening, a claim I’ve made here before. Maybe organizations don’t know to execute well. Maybe some projects are vaporized bullshit. But somewhere people are doing things.

Your Identity Has Been Deleted

If you’ve whipsawed from cynicism to catastrophizing, as I’ve described thus far, get ready for sorrow. For a number of us, doing programming or doing knowledge work has formed an integral piece of who we are. That next whipsaw threatens our livelihoods and our self-conception and a favored hobby, the open-source tinkering that fills the hours in between.

If machines are able to do knowledge and/or creative work, what will we the knowledge/creative workers do? Who will we be? Even when what I made wasn’t the best or most impressive thing out there, at work or in hobbyist contributions, earning a smile or a nod of appreciation for it was profoundly rewarding to me. That’s going away, too? Any insight, any intimacy, any care that I found can be stolen, reduced to something bolted-on, and quite-likely exceeded in seconds.

I credit Jack Maguire for calling it grief. I might call it anticipatory grief. This grief is structurally suppressed, because layoffs are framed as routine business decisions that leave no socially-sanctioned room for mourning. The grief is sharpened by the way the roles themselves are dissolving rather than simply shrinking. Workers aren’t taking a grief-blow (like a death), mourning, and then moving forward. They’re dying a little bit every day and being told to not see it. Kenneth Doka calls this disenfranchised grief: a loss that is not acknowledged or socially supported.

Nolan Lawson gets at the same thing from inside the craft itself, mourning the loss of “holding code in our hands and molding it like clay"2 — not grief over a job disappearing in the abstract, but over the specific, tactile satisfaction of the work itself going away out from under people who are still, for now, doing it.

That grief carries a clinical signal now, too. In September 2025, psychiatrists Stephanie McNamara and Joseph E. Thornton proposed a construct in the journal Cureus called Artificial Intelligence Replacement Dysfunction, or AIRD: elevated anxiety, insomnia, and depressive symptoms among workers facing AI displacement. Health outcomes, with the downstream costs in care, productivity, and lives that health outcomes carry.

The “impacted” are told to adapt and move on. But if the entire class of creative knowledge work goes away, if the entire fumbling scientific effort of making the bumbling become the state of the art is consumed by the AI labor army, all that’s left to us will be a desert we call progress.

The Acceptance Problem: Where the Model Breaks

The Kübler-Ross framework for how humans process grief assumes that acceptance is reachable, because the loss it was built to describe is finite. When a person dies, the absence becomes permanent. The bereaved adjusts to a stable, if painful, new reality. Acceptance is possible because there is something fixed to accept.

AI displacement does not offer a fixed endpoint. The process is ongoing and accelerating, with no stable post-AI equilibrium to adapt to. A worker who retrains into this year’s safe role may find that role automated within two years. There is no permanent absence to grieve, only a moving frontier and the anxiety about how one’s grief will be deepened again. Workers are being asked to accept a process rather than an outcome, and the process keeps advancing through fields of misery.

This is where the proposed solutions tend to falter. The common advice is to anchor identity in adaptability itself, to stop being a data scientist and become, in effect, a professional adapter. But capable AI models will affect all fields, all at once, irrevocably.

There is no safe harbor.

And thus we get back to “Five Years” and “hypernormalization.”

Our acceptance is Bowie’s five years stretched into an open sentence: no five-year mark to reach, no verdict to receive, only the countdown resetting each time a role gets automated. Hypernormalization, as a coping mechanism.

We never get to stop the hollow performance; the center does not hold; the workers perform competence and faith until they’re found out and made to scurry to their next hiding place. It’s a dignity-denying life. In this light, enjoy the milkshake.

Footnotes

  1. See also Camus, Albert: The Rebel
  2. We Mourn Our Craft
  3. The Boring Internet