After years of experimentation with generative AI, a seasoned developer argues that AI coding agents have no place in professional production code. Despite their impressive capabilities, the author identifies four critical issues, with skill atrophy being most concerning. Engineers supervising AI-generated code will gradually lose their technical proficiency.

The author contends that even if AI produces acceptable code 95% of the time, human reviewers will inevitably become complacent when examining hundreds of AI-generated submissions daily. Unlike human colleagues who can be consulted about implementation decisions, AI agents fabricate plausible explanations without actual understanding.

Beyond skill degradation, the artificially low cost of AI development raises red flags. Major tech companies continue pouring resources into unprofitable AI technologies despite clear financial realities. The author suggests this resembles a Silicon Valley bubble, with companies treating AI coding agents as proof their massive investments will pay off.