How Meituan’s AI Now Writes Over Half of Its Code – What It Means for Developers

On May 26, Meituan CEO Wang Xing announced that 52% of the new code written at the company is generated by AI.

What does “52% AI code” actually mean?

It doesn’t mean AI has taken over programmers’ jobs. Meituan’s AI code generation is based on its own large model “LongCat”. The AI handles repetitive, “heavy‑lifting” tasks such as scaffolding, simple CRUD, format conversion, and can produce hundreds of lines of code from a few keywords, leaving developers to fine‑tune details.

Meituan calculates the proportion by dividing the number of lines generated by AI each day by the total lines of code submitted, and even tracks AI‑written and human‑written code separately.

However, AI is not the main driver for core business logic or complex architecture; those still require human insight. Think of AI as a diligent intern that can build the walls, while engineers still draw the blueprints.

Programmers’ “crisis” and “opportunity”

While the figure sounds alarming, AI is neither a threat nor a magic solution. It behaves like a smart intern that can do work but needs supervision.

It understands syntax, not intent. AI excels at simple CRUD or format conversion, but struggles with complex business logic, often requiring multiple revisions.

Technical debt can grow. AI tends to copy‑paste code; small changes can lead to duplicated functions that become painful to maintain later.

Relying blindly on AI will backfire; the key is to use it wisely.

In the near future the role of developers is expected to shift from pure coding to “prompt engineering + code review”.

Three practical tips

Offload the “dirty work” to AI. Let AI handle scaffolding, format conversion, and simple components so you can focus on higher‑value tasks.

Guard the core logic yourself. Business logic, data models, and architecture remain your competitive edge; AI‑generated code must be reviewed and owned by you.

Treat AI like an intern. Give clear instructions, monitor its output, and perform thorough reviews. Properly guided, AI can save you up to half of the development time.

Final thoughts

AI development is accelerating, but developers should stay grounded. Writing code is more than typing; it involves understanding requirements, designing solutions, and collaborating.

AI can boost efficiency, but the quality of the product and user satisfaction still depend on human creativity and insight.