From Abstraction to Artificial Intelligence: Why Programming Faster Does Not Mean Knowing Less

This position article examines generative artificial intelligence as a new layer of abstraction in software development. Through a structured narrative review, it connects the historical evolution of programming abstractions—from machine code and assembly language to high-level languages—with evidence on productivity, correctness, security, and AI-assisted learning. It argues that generative AI reduces part of the mechanical effort required to express a solution without eliminating foundational knowledge or professional responsibility. Instead, engineering complexity shifts toward problem formulation, architecture, validation, security, integration, maintenance, and human oversight. The article proposes the ability to specify, understand, and verify AI-generated results—not manual difficulty alone—as a criterion of professional competence.