Agents are evolving from code-completion tools into genuine engineering participants in software development. This raises a key question: does the engineering process from idea to launch need to be redesigned? In practice we found that what determines agent efficiency is not only model capability, but whether the engineering system can provide clear, structured, verifiable context. Following the complete zero-to-one build of a real product, this talk shows how agents participate in requirements, design, development, testing and deployment, and summarizes the agent-oriented product engineering methodology we have developed.