How AI Interviewers and Vibe Coding Are Redefining Software Hiring

Hello, I'm Weiwei. Recently I noticed a job posting that seemed to be from Meituan, which this year introduced an AI interview bot for campus recruitment.

According to Meituan CEO Wang Xing, about 52% of new code is generated by AI, over 90% of engineers use AI coding tools, and the company will continue investing in large language models.

In June, Meituan released an AI product called NoCode, allowing users with no coding background to generate websites and software through conversation, essentially turning everyone into a developer.

Meituan's AI interviewers have already been used in spring and autumn campus recruitment, presenting candidates with a wide range of questions—including rote, scenario, and system‑design problems. Feedback indicates that the AI‑driven interview experience is surprisingly satisfactory.

While AI interviewers can handle over 60% of the interview workload, their question banks are static and do not adapt to a candidate's résumé, which limits personalization.

Nevertheless, allowing AI in interviews reflects a broader trend: developers now routinely use AI, so evaluating how they collaborate with AI is becoming essential.

"Using AI even in second‑round interviews is now allowed – Vibe Coding"

New Trend

Previously, using AI during an interview was considered cheating, but Meituan's "Vibe Coding" concept suggests that AI should be openly integrated into the interview process.

Vibe Coding, a term coined by an OpenAI co‑founder in early 2025, describes a development style where developers guide AI tools with natural‑language prompts to generate, refine, and debug code, enabling rapid iteration and even allowing non‑programmers to describe desired functionality.

In the future work model, programmers will inevitably rely on AI; Vibe Coding is inevitable.

Consequently, interview formats should evolve to assess developers' ability to work with AI:

Junior developers: must be able to use common AI coding tools to complete routine tasks and demonstrate basic Vibe Coding skills.

Mid‑level developers: should independently develop complex features with AI, possess strong prompt‑engineering abilities, and use AI for research, design, and feasibility analysis.

Senior developers: need to lead AI‑driven development processes, evaluate and introduce new AI tools, and translate business requirements into precise technical prompts.

Only AI‑permitted interviews can truly evaluate these capabilities.

Even though AI tools boost efficiency, they cannot replace fundamental knowledge. Without a solid grasp of algorithms, frameworks, or database principles, developers cannot verify AI‑generated code or designs.

The future interview will not eliminate traditional questions but will transform them into a "new eight‑leg" format that tests both core principles and AI‑augmented problem solving.

Ultimately, Meituan's experiment signals a shift from testing "what you know" to testing "how you collaborate with AI to achieve results".