The Horse Job Cope
The Horse Job Cope
Nate, former Senior Director of Procurement at BigCo. 2 kids, one a senior in high school, the other a sophomore at State U. Helen, his wife, has MS, she can’t really work. Anyhow they’ve made the whole shebang: house in the burbs, two cars, kids looking like they’re on the right path for takeoff velocity.
Then the RIF.
Nate’s out. He’s not really surprised, the department was shrinking, even though the work wasn’t. First 8 people, then it was down to 4. Then 3. A lot of automation. At the end, Nate and his little team were mostly translating outside stakeholders’ interest to internal requirements, fed into Procure.AI.* They mostly approved the moves that the automation was recommending.
Anyhow, RIF. So the obligatory LinkedIn post. The post is graceful, professional, and humiliating in the way LinkedIn demands you humiliate yourself: “Excited for the next chapter!” Nate lists skills. Nate thanks BigCo. Nate does not mention that the VP who cut his team is getting a promotion for this “cost containment”.
The comments are a master class in performative reassurance. “New doors will open!” “The best talent always lands.” And the evergreen, the one you’ve heard so many times it sounds like scripture: “Technology has always created more jobs than it destroys.”
[* Note from your tour guide / author: Ok LOL when I wrote this I made up “procure.ai” off the cuff but OMG it’s real. Somehow, that’s not comforting. AT ALL.]
The automobile eliminated the stable hand but created the mechanic, the driver, the road builder, the gas station attendant. The PC eliminated the typing pool but created the IT department, the web designer, database admins, desktop publishing designers, etc. Same for the Internet. The pattern is so reliable it became an axiom. Technology destroys jobs. Technology creates jobs. Be patient.
This “law of productivity physics” has been correct for two hundred years. I predict it is about to stop being correct, and the reason is structural.
Previously, every displacement technology was narrow and domain-specific. The power loom replaced hand weaving. It could do nothing to the dye maker or the accountant. The automobile replaced the horse but still required a driver. The ATM handled cash but still routed the complicated transactions to a human. Each technology eliminated one category of work and, in doing so, created adjacent categories that required human labor, because the technology had narrow boundaries. The loom needed a human operator. The car needed a human driver. The creation-destruction cycle worked because each generation of automation could do one thing. Humans did “everything else.”
From what I know, EverythingElse is a super great place, and we should visit. Start that visa application soon.
Artificial good-enough intelligence (AGEI, ok yeah, that’s an ugly-stick acronym, but ok) has no domain boundary. It processes language, analyzes data, writes code, drafts legal arguments, generates marketing copy, reads medical images, manages compliance workflows, and synthesizes research. I could keep going for five paragraphs, which is precisely the point. Every one of those tasks is pattern recognition over structured and semi-structured information. The tool that kills the old job can do the new one too.
When the automobile displaced the horse, the resistance of driving to mechanization is what created the driver’s job. The new job lives in the whitespace between what the machine could do and what remained beyond its reach. When AGEI displaces your cousin’s customer success team, the equivalent gap vanishes, because the tool that eliminated her role is also capable of performing whatever new role the reorg dreams up. The loom could weave but had no capacity for thought. The automobile could move but required a human to steer. AGEI operates on cognition itself, and cognition is what the knowledge economy is made of. I wrote about this before.
Here’s where the optimists’ math gets ugly.
Automation hit individual jobs. The loom replaced the weaver. The ATM replaced the teller. Job for job, task for task. Contrast: AGEI replaces the compliance department, from analyst to manager. The whole org unit that previously required a team coordinating across roles collapses into a single person directing a suite of models, or into no person at all. One person with agents replaces a team. The arithmetic of creation-destruction which held at the task level breaks at the function level.
Evidence, hot off the press… Meta’s CEO announced in early 2026 that “projects that required big teams are now accomplished by a single person” and began flattening organizational layers accordingly. Y Combinator’s 2026 request for startups explicitly calls for AI-native agencies that sell finished product rather than software tools: replacing the firm, at scale. Goldman Sachs embedded AI engineers for six months to automate entire compliance and accounting workflows. The creation side is producing one new role (the person who directs the agents) at a ratio of roughly one to many.
The many are gone.
Now comes the counterargument you’re already composing: creativity. Human creativity will remain irreplaceable. I hear this from smart people, people I respect, and it misidentifies the market question so badly it makes my teeth hurt.
The market does not ask whether human creativity is beautiful. The market asks whether human creativity commands a premium when a machine produces output that is good enough at a fraction of the cost. And the answer, amigo, the answer we already know from two centuries of industrial history, is: nope. Markets choose “good enough and cheap” over “better and expensive” every frickin’ time. The automobile was worse than the horse: less maneuverable, polluting, dependent on fuel infrastructure, useless in mud. It won because it was cheaper per mile at scale.
AGEI doesn’t need to be ASI. Not even AGI. It needs to be good enough to clear the CFO’s bar. That bar is lower than the one the optimists are defending. Every quarter it drops.
(We’ve seen this movie before. Financialization, offshoring, now automation. Each round the gains concentrated further, the socialized costs got larger, and the state’s capacity to absorb them got weaker. I’m no labor economist, but the pattern is not subtle. You don’t need a PhD to read the chart when the line only goes one direction.)
Every previous round of automation hit this same dynamic at the task level: the machine output was worse but cheaper, and the market accepted it. What has changed is that AGEI’s domain of “good enough” expands with each model generation and each agentic framework layered on top of it. The new jobs that the optimists promise must exist beyond the boundary of machine competence, and that boundary moves. Continuously. In one direction.
The horse-to-automobile analogy describes the past accurately. It describes the present with breathtaking and unjustified optimism, because it assumes the pattern holds. Trust me: it ain’t gonna hold. Every previous displacement technology left a gap where human labor was necessary. AGEI closes the gap: cheaply, at scale, which is all the market has ever required.
Here’s where capital comes unglued from labor. God help us all.
Nate’s LinkedIn post will get 200 likes and 47 comments telling him the future is bright. Those people are not lying, they’re being kind. They’re repeating a myth that was true for two centuries. They’re the horse breeders of 1908, assuring each other that demand for quality horses has never been higher. They’re right about the past.
This is the future now.
The horse job cope is the belief that the cycle of job emergence will keep pace with the cycle of job destruction, when the tool driving destruction is the same damn tool that would need to be incapable of doing the new work. The argument requires AGEI to be simultaneously powerful enough to eliminate existing jobs and too limited to perform the jobs that replace them.
I love the smell of contradiction in the morning.
Colin Steele writes to you from a sunny beach in EverythingElse, where he sips Mai Tais, writes about things he has no business writing about on colinsteele.org, and waits patiently for gig work from an AI looking to hire.