AI turns China’s coders into “tech janitors” as cheap, buggy code floods the market

The AI boom that promised to turbocharge software development is instead hollowing out margins and leaving programmers to mop up faulty machine-generated code. Who fixes the mess when a client’s AI prototype is an empty shell? For many independent developers in China, the answer is: them — for far less pay.

"Tech janitors" and collapsing fees

Qi Lei, a 35‑year‑old programmer in Changsha, told Sixth Tone he once charged as much as 300,000 yuan for custom mini‑programs; now he’s lucky to get 20,000. Clients increasingly present AI‑generated frameworks and demand the same price. Qi says he had to cut a recent quote nearly in half after a client insisted the AI output made the work “free.” Smaller shops and laid‑off engineers have flooded the market with bargain offers, driving down rates across the board. One sales employee in Xuchang reports his firm — which markets itself as “AI free” — has halved prices since 2023 and can build an app “visually similar” to Xianyu (闲鱼) in 20 working days for about 12,000 yuan. Reportedly, some clients now expect to pay only a few thousand yuan for projects that historically required months of skilled labor.

Bugs, technical debt and a geopolitical backdrop

AI speeds scaffolding but creates brittle systems that require experienced engineers to untangle. “Ask for Chinese noodles and the AI will serve spaghetti,” one Guangzhou freelancer joked, saying a small flaw can force reworking an entire module. Developers report heavier workloads and sharply lower pay; one freelancer estimated salaries have fallen roughly 35%. The problem isn’t just economics — it’s technical debt: quick AI fixes introduce cascading bugs that demand human intervention. Against this domestic picture sits a broader geopolitical frame: it has been reported that U.S. export controls on advanced AI chips have accelerated China’s push for homegrown models and tooling, which analysts say can be uneven in quality and may be contributing to the surge of imperfect, hard‑to‑maintain code. Can the industry retrain clients and reprice expertise before the next wave of automation arrives? The answer will determine whether developers remain builders — or permanent janitors.