At the "Global C++ and System Software Technology Conference," Huang Ning, the leader of product development at Sili Xin Technology (aiXcoder), gave a presentation entitled "Beyond Vibe Coding: Constructing a Human-Centered, Dependable Development Process." During his speech, he delved into the present landscape and anticipated future directions of software development propelled by large-scale models.

Huang Ning highlighted that the Vibe Coding model falls short in addressing the development demands of intricate enterprise-level projects. He stressed that for reliable research and development (R&D) processes to be established, AI must be seamlessly woven into the fabric of software engineering. He underscored that within enterprise application contexts, the Vibe Coding model is riddled with deficiencies, largely due to AI's absence of the extensive experience held by human engineers.

To tackle this issue, aiXcoder has put forth three guiding principles for AI software engineering and, consequently, has shaped an "AI + Software Engineering" development paradigm. This paradigm encompasses the following:

• Dissecting complex tasks to delineate the capability boundaries between humans and AI.
• Constructing verifiable systems to uphold enterprise-level safety benchmarks.
• Extracting implicit enterprise knowledge to furnish AI with enriched contextual insights.

This paradigm has already proven its worth in real-world projects. Looking to the future, Huang Ning envisions the industry transitioning from "defining software" to "defining software development models." aiXcoder is striving in both "horizontal and vertical" domains to render enterprise-level AI development more dependable, manageable, and streamlined.