"History doesn't repeat itself, but it often rhymes.", Mark Twain
The present moment in software looks less like a sudden break and more like the same tune played on new instruments. The sewing machine didn't kill tailoring; the tractor didn't kill farming. They made a few people vastly more productive and reshuffled who captured the value. AI agents are doing the same thing to software right now.
Andrej Karpathy put it more bluntly: programming has changed fast, not gradually, in the last few months. The new craft is explaining what you want, breaking work into tasks, and supervising systems that execute them.[1] Jack Dorsey recently halved Block's headcount citing exactly this leverage.[2] The AI-2027 scenario sketches out how this accelerates: agents moving from helpful assistants to employee-like systems wired into long workflows, compounding what small teams can do.[3]
My prediction is that the short-term pain is real. Some jobs, particularly those built around writing boilerplate, translating requirements into code, or doing repetitive analysis, will shrink or disappear faster than most people expect. That is not a scare story; it is just the same thing that happened to weavers, telegraph operators, and data entry clerks. The disruption is concentrated and fast at first.
But the longer arc bends toward more work, not less. The tractor eliminated farm laborer jobs and created agronomists, equipment engineers, logistics networks, and an entire food processing industry. Software will do the same. As agents commoditize execution, the demand for people who can define what to build, evaluate whether it worked, and connect it to real business problems will grow. The jobs shift upward and outward, into new roles that don't have names yet.
The rhyme is not a guarantee of a happy ending. It is just a reminder that we have been here before, and that the people who fared worst were the ones who waited for things to go back to the way they were.