The AI Revolution in Software Engineering: From Craftsmanship to Mass Production

The software engineering profession stands at a crossroads. What was once a craft built on elegance and artistry is rapidly transforming into something resembling mass production. This shift, driven by the accelerating capabilities of large language models and AI coding agents, has left many engineers grappling with an existential question: if AI can write most of our code, what's left for us to do?

The End of Code as Identity

The resistance to AI coding tools initially made sense. Early models hallucinated, produced buggy code, and struggled with complex logic. But somewhere in the past year, holding onto these objections became increasingly untenable. Today, engineers at leading AI labs routinely work without writing a single line of code themselves. Startups are following this pattern, and enterprises won't be far behind.

This transformation is painful for those who found identity in the artistry of programming. The romantic notion of the craftsman carefully sculpting elegant solutions is giving way to a reality where typing code is becoming optional. If your daily routine as a software engineer doesn't already look vastly different than it did in 2022, it soon will.

The Constants Beneath the Change

Despite this upheaval, some fundamentals remain reassuringly stable—at least for now. Software engineering has always been about outcomes, not code. The best engineers have always spent significant time thinking about productivity and team coordination rather than just writing functions and classes.

This is fortunate because the principles that make human teams operate efficiently also make AI agents more effective:

• Small, stacked diffs: Easier for humans to review and understand, and perfect for swarms of agents making concurrent changes
• Continuous deployment: Already a good practice, now essential when shipping code at unprecedented volumes
• Automated testing: Critical when you can't manually review every line of AI-generated code
• Easy rollbacks: Necessary when deployment frequency increases by orders of magnitude

The Evolution of Engineering Taste

What makes a good software engineer? Historically, it's been a combination of taste and intuition—qualities built through years of experience wrestling with code. This intuition is evolving. Rather than operating at the level of individual lines of code, it increasingly manifests as architectural judgment.

Junior engineers now face a compressed learning curve. They must develop architectural taste immediately, largely bypassing the traditional apprenticeship of code craftsmanship. Frontier models are writing increasingly clean code, especially when guided by well-crafted AGENTS.md files that provide context and constraints.

However, AI still struggles with understanding the nuanced constraints that define much of our work—both technical and social. Understanding why a particular approach won't work with the existing team culture, or why a seemingly elegant solution would create maintenance nightmares down the road, remains a distinctly human skill.

The Deep Blue Feeling

Adam Leventhal and Simon Willison have aptly named the pervasive dread many engineers feel: "Deep Blue," after the chess computer that defeated Garry Kasparov. This feeling is real and valid. There are days when the implications of this transformation feel overwhelming.

Yet there's another side to this revolution. When you lean into this new way of building, there's an undeniable thrill in the insanely fast feedback loop and the feeling of expansiveness that comes from orchestrating multiple agents working concurrently toward a shared goal. It's a different kind of creativity—less about the craft of individual lines and more about the architecture of systems and the coordination of multiple AI workers.

The Coming Creative Destruction

Not everyone will enjoy this new mode of work. Many engineers, especially those earlier in their careers, won't have the experience or professional networks to cushion the transition. We're living through what economists call "creative destruction"—the process by which new innovations disrupt existing industries and ways of working.

There's real excitement in this transformation, but also genuine grief. The ability to hold both emotions simultaneously is perhaps the defining challenge of this moment in software engineering. The craft we loved isn't disappearing entirely, but it is fundamentally changing. The question isn't whether to adapt, but how to adapt while preserving what made us passionate about this field in the first place.

For now, the principles of good engineering—clarity, simplicity, maintainability—remain our anchors. They guide both human teams and AI agents toward better outcomes. Whether these principles will hold true beyond the next five years is uncertain. But in this moment of transformation, they offer a foundation to build upon as we navigate the shift from craftsmen to architects, from code writers to system orchestrators.