A 30-Something Full-Stack Dev Gets Blunt Career Advice: Promotion Is a Quota Game, Not a Meritocracy
AI coding costs are being squeezed hard enough that a Chinese company is swapping Claude for domestic models, a trade-off that will reach Western budgets as domestic models improve. For individual developers, the real constraint on promotion is headcount, not skill, and the overseas programmer job market is already saturated with applicants.
The ninth installment of a Chinese developer interview series follows a married, 30-year-old full-stack engineer at a mid-sized internet company's branch office. He owns a house and car, earns decent money, and has a stable job, but the company has run several rounds of layoffs, hiring only backend Java/Go developers for Agent work. Front-end roles have effectively disappeared in internet companies, and the developer is now exploring a management track or an overseas position.
The company's budget tightening extends to AI tools: it previously paid a few hundred dollars per developer per month for Claude Code with the Claude model, but now it's cutting that allowance and actively encouraging domestic Chinese models. The interviewer reads this as both a cost-saving move and proof that domestic models have become good enough for routine development. Meanwhile, the company is trialing AI-driven code review and automated testing, which could replace some testers alongside developers.
On AI applications, the interviewer observes that since spring, tools across life, office, and production have stabilized with no next-generation leap. He dismisses the overseas "goal and loop" agent paradigm as unfit for wide adoption and predicts that AI coding plus full-stack development will settle into best practices next year, much like Vue and Springboot did a decade ago. On the promotion question, he tells the developer outright that ambition doesn't create openings; company headcount and growth do. In a no-growth firm, even excellence won't promote you unless your boss leaves. The practical advice is to keep building value and wait for luck. For the overseas job idea, he recommends sending resumes first before investing in spoken English, noting that foreign markets are months ahead on AI and equally flooded with unemployed programmers.
Companies now treat AI tokens as a negotiable line item, not a strategic investment, and will switch models as soon as a cheaper alternative clears the quality bar.
Domestic Chinese foundation models have reached a viability threshold for routine development work, which reshapes the AI vendor landscape for employers watching costs.
The plateau in AI tool functions suggests the industry is entering a consolidation phase where process and workflow best practices matter more than model features.
In shrinking organizations, a management promotion is closer to a lottery whose odds are set by headcount plans, not by individual performance.
Advising developers to signal interest in overseas jobs via resume cold-calls before learning the language assumes a buyer's market where effort should follow market response, not optimism.