跪拜 Guibai
← All articles
Frontend · Full-stack · AI Programming

Pure Frontend Is Over: A 9-Year Developer Gets the Full-Stack Ultimatum

By 双越AI_club ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

The full-stack, AI-assisted baseline that is wiping out pure frontend roles in China's private sector is spreading to Western companies as they consolidate teams; developers who stay narrowly specialized risk being replaced by generalists who can ship end-to-end.

Summary

In the tenth installment of a one-on-one interview series, a nine-year frontend developer from a mid-sized construction company in Shenzhen asks how to keep his career sustainable. The answer is blunt: pure frontend positions no longer exist in private companies, and clinging to the frontend-backend separation model of his semi-state-owned employer is a dead end. The recommendation is to learn the company's Java backend stack and build a personal project that runs end-to-end, then match the backend team's daily development needs.

On AI programming, the advice distinguishes between using AI tools and governing them. Colleagues who race through features with AI produce unmaintainable code, but managers don't penalize them, so the author suggests using skill tools, MCP, and harness processes to constrain AI output on your own. He also tells the developer not to fight the reality that bad code wins; the job belongs to the boss, but skills belong to the developer.

On the second question, the author is equally direct: management promotion is unrealistic when teams are shrinking, and the only realistic path is to keep working as a frontline developer as long as the company keeps you, even if it means overtime. He dismisses the idea of quitting for an easier job, noting that no job is truly relaxed, and points out that a nine-year career is not exceptional; many have worked fifteen years and still clock in.

Takeaways
— Pure frontend positions have disappeared from most private Chinese internet companies, with teams merged into full-stack roles.
— A developer in a traditional industry still working with frontend-backend separation is at risk; the author advises learning the company's Java backend stack.
— To transition, a developer should self-study to build a full project that runs through all processes, then match the backend team's daily needs.
— AI programming code quality can be controlled via skill tools, MCP, and harness workflows, but an individual cannot force company-wide adoption.
— The author's stance is to accept that bad code may win in the short term; prioritize personal skill growth over the boss's projects.
— Management promotion is unlikely during layoffs and shrinking teams; keeping a job and working overtime is the realistic expectation.
Conclusions

"There is no frontend" is an overstatement, but it captures a real shift: in agile private companies, the frontend-backend boundary has blurred into full-stack ownership, and specialization alone no longer protects a role.

The advice to learn Java specifically because the company uses it, rather than the latest stack, shows that career pivots depend on fitting existing systems, not on chasing trends.

The acceptance of overtime and the "don't care about code quality" attitude reflect a survival mindset in a shrinking job market, where individual developers are told to insure themselves through skills, not through organizational outcomes.

Concepts & terms
MCP
Model Context Protocol (MCP) is an open standard that lets AI models access external tools and data sources through a uniform interface, allowing agents to perform actions like file reads, API calls, or database queries. In AI programming, MCP helps constrain what an AI assistant can do and keeps its outputs within project-defined boundaries.
Harness
In AI programming, a harness is a structured workflow or framework that defines how an AI agent operates: how it receives a task, which tools it may use, how it verifies outputs, and where it stops. Harnesses enforce code quality and stability by adding checks, tests, and review steps around raw AI generation.
From the discussion

The panic over pure frontend's death meets a blunt dismissal. Layoffs are real, with backend staff using AI to patch frontend and testers also cut. The counter view calls that narrative nonsense, noting career changers survive and easy jobs exist, and blames the author's rigid programmer mindset. A third reply is just a consulting pitch.

— Pure frontend roles are being eliminated, with layoffs and backend engineers using AI to modify frontend code; testers are also affected.
— The doom narrative is false: many who have switched fields have not starved, and easy jobs are abundant after unemployment.
— The author's panic stems from a typical programmer's fixed mindset rather than actual market reality.
Featured comments
excellent

Pure frontend is indeed difficult. Now it's basically all layoffs; only backend remains, using AI to modify frontend code. Some testers were laid off too.

enjoywindstorm

Nonsense talk. Plenty of people switch careers and none of them starved. If you actually become unemployed, you'll find that easy jobs abound. The author's programmer fixed mindset is way too strong.

See top comments, translated →
Source: juejin.cn ↗ Google Translate ↗ Backup ↗