The article addresses the impact of artificial intelligence on the IT industry and profession, drawing on the perspective of Pedro Samorrinha, an alumnus of Instituto Superior Técnico with approximately twenty years of career in Computer Engineering, currently holding responsibilities in technical leadership of teams, system architecture, and driving the software development lifecycle.
Samorrinha considers that the scenario frequently described in AI communications — in which a technical lead arrives in the morning, checks the work produced overnight by agents, corrects errors, and assigns new tasks — does not match his current reality, although it may come to be. He prefers shorter validation cycles and emphasizes that this form of producing software will only be truly productive if it rests on a rigorous development methodology, with precise requirements engineering, good planning, and testing capability.
Samorrinha's team deliberately avoided so-called "vibe-coding," choosing to maintain a classic development process now "agentified," with verifiable stages and human validation in each cycle. The methodology is considered the critical point: through well-defined specifications and human verification at each step, the space for hallucinations and unexpected results is reduced. Despite the risks, the quality of what agents produce has improved significantly in recent months, and data collected by the team points to productivity improvements of 30% or more.
The article concludes that if code has become easier to produce, software engineering — the discipline of deciding well what to build and verifying it was well built — is today even more critical. The training of younger engineers is also changing: the center of gravity is shifting toward functional and technical specification, where the experience of seniors has greater value. A junior who follows rigorous specification and validation processes from the start may learn faster than they would through traditional programming practice.
