The Backend Engineer Who Discovered AI Devalued His Eight-Year Career Moat
Coding proficiency is being repriced across the industry: AI and junior engineers now cover most of what senior developers used to sell. The story shows the moat is no longer framework fluency but the ability to own a product loop and build things people use, which is why developers everywhere are rethinking what they spend their hours learning.
Eight years of Spring Boot, microservices, Docker, Kubernetes, message queues, and database sharding didn't keep a 31-year-old backend engineer employed. The company he'd been at since his early career shrank from over 300 people to about 80 in under two years. His team went from 12 to 4. On a Friday afternoon, HR asked him to a meeting room, and he left an hour later with an n+1 severance package and a hand that was steady when he signed.
The real shock wasn't the layoff itself, which he saw coming from industry headlines. It was the realization that the technical depth he'd spent eight years building was now matched, at about seven out of ten, by AI tools and recent graduates. With Cursor and Claude Code, an API endpoint that used to take two days and multiple rounds of frontend alignment now comes out in an afternoon, tests and docs included. The scarcity used to be writing code; now it's thinking through requirements, judging AI output, and taking an idea from zero to users.
His response has been pragmatic: he built a small tool for his wife's cross-border e-commerce business that calculates freight and tariffs, using AI. It took two weeks, now has 200+ users, and makes under 300 RMB a month. He's not claiming a comeback. He's stopped chasing new tech checklists, started learning only what solves immediate problems, and redefined the skills he needs to carry out of a building. The phrase he repeated: "AI doesn't equal making money. Writing something nobody uses is the same as not writing it."
The author's framing that coding itself is losing scarcity is a direct consequence of tools like Cursor and Claude Code, and it reframes experience: familiarity with a specific framework or internal system is less transferable than the ability to reason about requirements and product outcomes.
The n+1 severance and the three-month self-check show how Chinese tech culture is adapting to layoffs, but the underlying shift applies broadly: company-specific knowledge is a liability when the company itself is doing the cutting.
The small tool's trajectory (200 users, less than 300 RMB/month) is a realistic counterweight to AI side-project hype. Distribution and user retention, not just code generation, remain the bottleneck.