Cultivating Depth in an Era Where AI Writes Code

Introduction

Recently, I read an interesting article. It suggested that AI writing code could be a repeat of the 'lost decade' that frontend development experienced. Frontend development, once a specialized craft, became something 'anyone could do' as frameworks took over. The same trend, they say, is now happening to programming at large because of AI. How do you feel about that? It made me pause and think. Are the layers of skill and the creator's 'depth' really being lost?

What the 'Lost Decade' Taught Us

From the late 2000s into the 2010s, the world of frontend changed dramatically. Artisanal skills like semantic HTML, CSS, accessibility, and understanding the subtle differences between browsers were 'abstracted away' as frameworks like React and Vue emerged. Efficiency went up. Barriers to entry fell, and many people could now build the surface of the web. But in exchange, a culture that undervalued deep knowledge also took root. To add a button, you didn't need to know what HTML was generated under the hood or how it affected network performance. You could just drop in a Shadcn component and it would work. As someone who has worked at the intersection of hardware and software, I have complicated feelings about this neglect of what lies beneath.

Deskilling: A Double-Edged Sword

In sociology, there's a term: 'deskilling.' It describes how skilled labor gets replaced with semi-skilled or unskilled labor through technological innovation. For companies, it reduces costs and weakens workers' bargaining power. The history of frontend development is a textbook case, and now with AI generating code, the same pattern is spreading to programming more broadly. But I can't bring myself to dismiss this simply as 'efficiency equals evil.' Take my beloved Revox tape recorder, for instance. In the old days, editing reels involved scissors and splicing tape, watching the sound waves while listening intently for every tiny noise. Today, you cut and paste waveforms on a DAW. That might be a form of deskilling too, yet the expressive possibilities have certainly widened. What matters, I think, is whether the sensitivity to 'what you want to hear' stays honed, even as the tools change.

Leaky Abstraction and Lessons from the Garden

Software abstractions are often said to 'leak.' While React hides the intricacies of the browser, on slow networks or underpowered devices, those leaks become glaring. AI code generation also has its own non-deterministic fragility: slight prompt variations can change the output. It's like a home garden that relies solely on chemical fertilizers, ignoring the soil's pH or the balance of microbes. On weekends, I tend vegetables in a small field. The timing of seeding, drainage, the threat of replant failure: the ability to hear the 'voice of the land' that takes years to develop is precisely analogous to the 'front of the frontend' in web development. Even if frameworks and AI smooth over the surface, whether you can feel the underlying 'natural principles' determines how truly fruitful the outcome will be.

Where Does Technical Depth Reside?

When I was involved in developing centimeter-level GPS positioning, I saw time and again how satellite signals were scrambled by atmospheric disturbances and how multipath reflections threw off accuracy. The engineers who could intuitively sense, beyond what algorithms could correct, that 'this error is a reflection from that building to the west' were true artisans. AI can process massive data sets, but it can't feel the breeze at the worksite. This 'embodiment' may be the last stronghold of a creator's depth. When we refactor code that AI has written, we read the design thoughts behind it and fill in the gaps. It's an act like catching a phrase from a fellow musician in a jazz improvisation and adding your own voice on top.

Between Bauhaus and Mass Production

A hundred years ago, the Bauhaus movement met the wave of industrialization by championing the fusion of art and technology. They tried to overcome the loss of handcraft through a new aesthetic language. Today, we stand before the same question: how to engage with the mass-produced code that AI provides. Some argue that using AI-generated code is no different from copying and pasting from Stack Overflow in the past. It's true that the skill of finding the right code snippet with the right search terms resembles the skill of refining a prompt. Yet those who, through repeated copy-pasting, begin to wonder 'Why does this work?' eventually reach original insights. I want to believe that AI can likewise serve as a spark for such learning.

Conclusion: What Remains in Your Hands

In the age of AI-written code, the depth of the maker endures in a different form. It resides in the curiosity to probe the layers beneath invisible abstractions, the sensitivity to catch the faintest gaps, and a sense of beauty cultivated over long years. We must not be swept away by the wave of efficiency; instead, even as we ride it, we must remember to feel the seafloor with our own feet. Does the touch of the skills you've accumulated still linger in your hands? Whenever I turn the knob of my old Revox, I decide once again to trust my fingertips.