AI is shaping our craft
Encouraging leaders to care about the craft of software
Original post: https://www.linkedin.com/posts/asher-cassetto_ai-software-activity-7372180923917877248-TmzN
AI copilots and vibe coding sessions are changing the way teams build software. Code can now be generated at speed, and outcomes reached faster than ever. For senior leaders, this looks like efficiency, but the cultural cost is less obvious.
Historically, developer growth came from the loop of build → iterate → fix → learn. Wrestling with complexity built intuition: why this pattern works, when to refactor, how to balance trade-offs. Those lessons compound into craftsmanship.
Copilots flatten that process. With a single prompt, developers can bypass the struggle and land at “working” code. The risk: a generation of engineers without scars, without judgment, and without the depth needed for long-term quality.
That said, the right path opens new paradigms. When code is cheap to produce (though not cost-free—tokens carry hidden expenses), possibilities expand:
Faster prototyping, more experiments, more shots on goal.
Novel ways to deliver outcomes to users, freed from old constraints of velocity.
Teams focusing less on syntax and more on problem framing, design, and value delivery.
If we agree that users are the most important piece of the puzzle, then some of today’s engineering rituals will become less relevant. But the art of problem solving and critical thinking cannot be outsourced. Those skills must still be forged “on bare metal”—through real debugging, tough trade-offs, and deliberate practice.
The leadership challenge is balance:
Encourage copilots as accelerators of creativity, not replacements for thought.
Preserve pathways for developers to learn the hard lessons, not just the shortcuts.
Redefine craftsmanship around judgment, design, and user focus—while keeping the discipline that sustains quality.
The AI era won’t erase the art of software. It shifts where that art lives. Leaders who manage this shift well will build teams that are both faster and deeper. Those who don’t may find themselves with engineers who can generate code but cannot solve problems.
#ai #software