5. The Agentic SDLC Framework: AI Autonomy and Responsibility

1. The article analyzes the evolution of artificial intelligence in software engineering, pointing to the shift from generative assistants to autonomous AI agents. The author emphasizes that the key change is not about the volume of code produced, but about agency and the ability to achieve goals within a complex environment. Using the SWE-bench benchmark and the Agent-Computer Interface (ACI) concept, it is demonstrated that AI efficiency depends not only on the model but on tools designed specifically for machines. The text redefines the division of cognitive labor in the SDLC process, shifting the focus from simple prompting to managing the agent's trajectory of actions and human oversight over final deployment.