The software development industry stands at a pivotal moment as AI agents transform programming from a manual craft into an automated process, mirroring historical technological disruptions like the Jacquard Loom that eliminated traditional weaving jobs while creating new opportunities for those who adapted.
The software development industry stands at a pivotal moment as AI agents transform programming from a manual craft into an automated process, mirroring historical technological disruptions like the Jacquard Loom that eliminated traditional weaving jobs while creating new opportunities for those who adapted.
The Jacquard Loom: A Historical Parallel
In 1801, Joseph-Marie Jacquard unveiled his revolutionary weaving machine in Lyon, France. This device, which could be considered an early computer, used punch cards to automatically create complex fabric patterns. The impact was staggering: productivity soared from 2 inches of fabric per day to 2 feet daily—a 1,100% increase.
But this technological leap came at a devastating human cost. Skilled weavers who once earned 6 shillings per day (enough for a comfortable middle-class life) saw their wages plummet by 80%. The specialized craft that defined their careers essentially vanished over two decades. The resulting social upheaval was so severe that it spawned riots, including the infamous Luddite movement and even "hunger riots" where starvation drove people to protest.
Yet history reveals a nuanced truth: not everyone lost their jobs. France eventually became a textile powerhouse, with new roles emerging for those who could design patterns, optimize production processes, and explore new markets. The disruption, while painful, ultimately created opportunities for the creative and technically adept.
The Programming Loom Has Arrived
The parallels to today's software development landscape are unmistakable. On a single Friday, I expanded my uuidv7 library from Java/Kotlin to implementations in Ruby, JavaScript, Swift, and Go—without opening an IDE or typing a single line of code. Instead, I orchestrated AI agents, providing clear specifications, reviewing their output, and ensuring quality standards.
This wasn't programming in the traditional sense. I was running a factory of programmers, powering multiple "looms" with my software engineering expertise while doing zero manual coding. The work required considerable skill: defining library standards, selecting appropriate algorithms, and maintaining quality through iterative feedback loops.
This transformation extends beyond simple code generation. I've successfully deployed AI agents in critical, complex backend systems at Square, using techniques like Reinforcement Policy Improvement (RPI) and breaking work into clear, agreed-upon steps. When agents struggle, I refine my prompting, update documentation, or create custom skills—treating the AI as a collaborator rather than a replacement.
The End of Traditional Programming
The implications are profound and unsettling. The expectation that developers will receive clean tickets and requirements, then spend two weeks completing them, is becoming obsolete. The 2 inches to 2 feet productivity shift that transformed weaving is happening in software development right now.
This isn't a distant future scenario—it's happening today. The tools will only improve, making the current limitations temporary. Engineers who cling to the belief that "AI can't do it well" are in denial, much like the weavers who couldn't imagine a machine replacing their craft.
From Construction Workers to Engineers
The distinction between "programmers" and "software engineers" has never been more critical. In traditional engineering disciplines, professionals rarely perform the manual labor of construction. Bridge engineers don't weld steel, but they understand welding techniques, weight distribution, material properties, and structural loads. They design visions that require thousands of construction workers to execute.
Similarly, software engineers must evolve from manual coders to system architects and orchestrators. The 80 hours an engineer spends designing a system translates to hundreds of thousands of hours of "construction work" performed by AI agents. Engineers still have careers; programmers—the construction workers of code—do not.
The Path Forward
This transformation demands a fundamental shift in mindset and practice. Traditional development methodologies—6-month plans, 2-week sprints, backlogs—may no longer align with the new reality of AI-augmented development. The granularity of planning needs rethinking.
Success in this new landscape requires:
- Adaptation: Embracing AI tools rather than resisting them
- Orchestration: Learning to direct and refine AI agents effectively
- Architecture: Focusing on system design rather than manual implementation
- Quality Assurance: Developing skills to review and validate AI-generated code
- Innovation: Exploring new possibilities that AI enables
The historical pattern is clear: technological disruption eliminates certain roles while creating others for those who adapt. The Jacquard Loom didn't end textile production; it transformed it. Similarly, AI won't end software development, but it will fundamentally change what it means to be a software professional.
For those willing to evolve from programmers to engineers, the future holds unprecedented opportunities. For those who resist, the path forward becomes increasingly difficult. The choice, as always, is ours to make.
Comments
Please log in or register to join the discussion