Is AI Causing a Repeat of Frontend’s Lost Decade?

Mauro Bieg, writing on the Mastro blog on 23 May 2026, argues that what AI is doing to programmers' jobs should feel familiar to frontend developers, because the same process happened to them over the past decade. Alex Russell's name for it was “Frontend’s Lost Decade”.

Deskilling, twice

Deskilling, as Wikipedia defines it, is the process by which skilled labour in an industry is eliminated by technologies operated by semi- or unskilled workers: it cuts costs, lowers barriers to entry and weakens workers’ bargaining power. Frontend development was once a specialised craft — semantic HTML, CSS, browser differences, accessibility and performance. Frameworks that treat the browser as just another compilation target replaced much of that with generalists who can wire up a component library. AI is now deskilling programming in the same way, and the author notes both the sense of loss and the suspicion that the new process produces lower-quality work.

A tower of leaky abstractions

Another framing is that automation simply moves work to a higher level of abstraction. The problem, Bieg writes, is that details leak through: heavy client-side JavaScript frameworks let developers avoid thinking about accessibility and performance on low-end phones and slow networks. Agentic coding is a leakier abstraction still — unlike a compiler it is not deterministic, and small changes to the input or the model give very different results. The comparison to “junior engineers” fails in one respect: people learn, without endless tweaking of AGENTS.md or SKILL.md files.

LLMs as Stack Overflow copy-pasta

The closest analogy the author has found is how Google search used to behave: choosing the right keywords so that the right forum post surfaced, a skill later applied to Stack Overflow. LLMs continue that trend — making people who know what they are doing slightly faster, and letting people who do not arrive at something that often kind of works. But the abstraction eventually leaks, and someone has to understand the system deeply enough to fix it, just as juniors were once taught to read the answer before pasting it.

Bauhaus, quality and tradeoffs

The Bauhaus answered industrial mass production by having factory workers and craftspeople work together, with designers returning to the workshops and the end user always in mind; Dieter Rams and Jony Ive trace back to it. Software, Bieg writes, sits between craft and industrial design, and the need to write code by hand remains: a web designer should know HTML and CSS as intimately as an industrial designer knows materials. When the hype dies down, LLMs will be one more tool; until then the industry will produce a lot of ugly things — ugly code, broken communication and layoffs under the guise of AI.