I am a veteran of more than twenty years in the code trenches. I once hand-wrote semantically elegant HTML like lines of poetry and spent whole nights debugging CSS box-model quirks across browsers. Now, standing at the threshold of 2026 and watching AI's magic wand wave, I feel a familiar chill: is history repeating itself? The 'lost decade' that frontend development experienced is now spreading, in a broader form, across the entire programming world. Will the flame of craftsmanship be swallowed once more by the fog of abstraction?
How Frontend Craft Was Lost in the Framework Wave
Over a decade ago, frontend was not a 'drag-and-drop game' anyone could play. To build a good web page, you had to master semantic HTML—choosing the most fitting words for every tag so screen readers could 'read' the content; command CSS's infinite variations for precise layout and responsive adaptation; and navigate the battles between browsers—IE6's quirks, differences between rendering engines. On top of that came accessibility, progressive enhancement, web performance optimization, and genuine care for real user experience. These skills formed the core of 'the frontend before the frontend.'
Then JavaScript frameworks rose like steam looms of an industrial revolution. The browser was no longer an art hall to be carefully served, but a compilable target runtime. Developers could 'quickly' assemble apps with React's magic hooks and ready-made Shadcn components. The underlying details—performance bottlenecks, accessibility traps, cross-device compatibility—were buried under layers of abstraction. Executives applauded: why pay premium salaries for 'fussy specialists'? Hire a few framework-fluent 'generalists' who could pivot to React Native or Electron—maximum flexibility, minimum cost. Workers' bargaining power eroded accordingly.
This is exactly what Alex Russell warned about in his talk *Frontend's Lost Decade*: the de-professionalization of skills floods the market with 'good enough' code that has lost its soul. Like craftsmen sighing before the assembly line, the developers who once fought for pixel perfection and for vulnerable users fell silent.
> Deep note: Deskilling is not mere technical simplification. It replaces skilled artisans with semi-skilled laborers via new tools, achieving cost savings and lower barriers to entry. It has recurred throughout history, from textiles to programming today. It delivers short-term efficiency but can erode long-term innovation and practitioners' dignity. As the Wikipedia definition cited in the source essay states, it comes at the cost of professional depth.
The New Wand: AI's Latest Wave of Deskilling
Today, AI is sweeping over programming in a strikingly similar way. The fine craft of hand-writing code is being diluted by the whispers of 'prompt engineering.' Companies see the cost-saving temptation again: let AI agents generate features and fix old bugs while developers merely describe intent—'semi-skilled' prompters producing 'close enough' results. Why keep a team of senior engineers? Let AI, the 'junior engineer,' do the work with a few supervisors watching. Bargaining power, once again, is diluted.
This brings a profound sense of loss. Like craftsmen facing the roar of machinery, developers who fought for elegant architecture and edge cases now watch AI-generated 'code spaghetti' with mixed feelings. The flame of craft seems to dim bit by bit. AI has no 'apprenticeship'—it does not truly grow from your feedback.
The Maze of Abstraction: Leaky Truths Behind the Efficiency Illusion
Some will argue this is just 'a higher level of abstraction,' freeing us from trivia to focus on the big picture. Abstraction does improve efficiency. But as Joel Spolsky famously warned: *all abstractions leak*. The veil is never perfect—it hides everything in calm weather, but under pressure the thorns beneath pierce through.
In modern frontend, this tower of abstraction is riddled with holes. React and similar frameworks abstracted away DOM manipulation, letting developers ignore performance disasters and accessibility failures on low-end devices. When users arrive on cheap phones and weak networks, the 'hidden' details flood out—stuttering pages, unclickable buttons, screen readers losing their way. Developers choose not to think or care because the tools have 'handled everything.' But leaky abstractions never truly disappear. They are just waiting for the next storm.
The Agent-Coding Gamble: The Ghost of Non-Determinism
More troubling is today's fashionable 'agent coding.' You describe requirements in natural language, and the AI 'fills in' details from its vast training data. This is nothing like a compiler's determinism—tiny input changes or random model fluctuation can produce wildly different output. It is an assistant permanently stuck at the 'junior' stage, yet unlike a human, it cannot truly learn from feedback. There is no config file you can edit to make it 'grow.' This goes further than the Stack Overflow era's copy-paste—and is more dangerous.
Early Google searches demanded precise keywords; Stack Overflow demanded that you understand an answer before using it. LLMs make everything 'faster'—and make understanding optional. Just as we once taught junior developers to read search results before using them, we must now teach people to read AI-generated code before using it. Otherwise bad code spreads exponentially, and team communication fractures in endless arguments over 'this part was written by AI.'
Quality's Cold Eye: What Is Code's Warmth Worth in Business?
Many developers never truly bothered to understand Stack Overflow answers—'whatever, it runs.' Companies tolerated it. The embrace of AI is even more open and less scrutinized. New ways of producing bad code multiply, and disagreements over AI integration tear at team workflows.
Yet the harsh reality: software quality rarely correlates directly with business success. Projects are treated as black boxes, expected to fail, and 'de-risked' by restarts. Frontend failures—slow sites, annoying cookie banners—barely dent brand loyalty or pricing strategy. Competitors' sites are just as bad. Who gets fired for choosing React? 'Nobody gets fired for using React' became industry slang.
That does not mean quality doesn't matter. It just tends to be outweighed on the business scales. AI makes 'slop'泛滥 everywhere—which is precisely the reminder: we still need people who truly know and truly care.
The Bauhaus Lesson: Reconciling Craft and Industry
When industrialization flooded markets with cheap reproductions, some chose retro imitation; others charted a new path. The Bauhaus movement did not reject the machine. Instead, designers went deep into the nature of their 'materials'—wood, metal, glass—seeking balance between function and form, aiming to make good design affordable to the masses without losing craftsmanship. That spirit influenced Dieter Rams, Jonathan Ive, and others after them.
Software is the same. It is both craft (code as product) and industry (mass distribution). Facing AI tools, we should not blindly let them swallow everything, but like the Bauhaus masters, deeply understand our 'materials'—HTML semantics, CSS layout philosophy, how browsers work. Even with frameworks and AI, let them serve users, not the reverse. AI can generate mountains of slop, but it can never replace those who know what they are doing—and care.
Fission and Rebirth: How Will the Future Workshop Take Shape?
So how does this end? High-quality work may become a small niche, but cheaper tools will grow the whole 'pie.' Whether absolute demand rises or falls remains unsettled. Some niches may be harder to make a living in, but great work will still shine.
For prototyping, a fast MVP is fine; but performance and architecture debt is hard to repay later. A better strategy: start simple and add complexity gradually. AI is just one item in the toolbox—alongside build, buy, and open source. Until the hype recedes, we may still endure ugly code, broken communication, and layoff waves justified in AI's name.
But remember: care for quality and for users is always the core. However tools evolve, that reverence for 'materials' and empathy for users is the undying ember of the code craftsman. Perhaps the future of programming will be a feast with more participants—and the fine works will still quietly shine in the hands of those who understand the materials and guard the users.
References
1. Mauro Bieg, "Is AI causing a repeat of Frontend's Lost Decade?", Mastro Blog, May 23, 2026. 2. Alex Russell, "Frontend's Lost Decade" (talk video). 3. Joel Spolsky, "The Law of Leaky Abstractions", Joel on Software, 2002. 4. Baldur Bjarnason, "The deskilling of web dev is harming us all" (blog post). 5. Bauhaus movement historical materials and the Wikipedia "Deskilling" entry (based on citations in the source essay and historical context).