AI Won’t Take Your Career, but Blind Coding Might

Summary

Andrej Acevski argues that AI is changing software development without making engineering judgment obsolete. He uses AI every day for code review, code generation, and building complete systems, but says engineers still have to read the resulting code; in practice, AI shifts some effort from writing code toward understanding it. He acknowledges that coding agents make it easy to create clones and turn weekend projects into viable small MVPs, making software cheaper to produce. The harder test is whether those systems remain readable, extensible, and scalable after two years or under much larger data and user loads. Acevski warns that blindly accepting generated code can create defects that remain hidden until later prompts or feature changes build on them. He recommends using agents for repetitive work through multiple iterations, clear rules, and skeptical review, while keeping architecture decisions under human oversight. In his view, engineers develop the judgment to distinguish good from bad code by learning, experimenting, and making mistakes deliberately. AI remains a powerful tool, but experience and the ability to evaluate code cannot be bought as a monthly subscription.