Manager’s Push for AI Coding Leaves Developer Fighting Bad Code and Design

Summary

A software developer says his company’s push to use AI in development has produced over-engineered code, defects, and technical debt because teammates often accept generated output without understanding or adequately reviewing it. He describes finding that roughly half of a lengthy technical design document consisted of AI-generated nonsense, while some coworkers could not explain code produced for their own tasks. During a review of his own pull request, his manager asked him to use AI for another review. The developer separated useful suggestions, such as documentation and error handling, from hallucinated recommendations and changes caused by a flawed design. After a call established which proposals were unnecessary, the manager later returned with another AI-generated branch that reintroduced the rejected changes and made the structure more complicated. The author argues that AI also reproduces poor patterns from old parts of the codebase, rather than judging whether those patterns remain appropriate. He says the resulting cleanup is frequently assigned to him despite his junior level and lower pay, leaving him discouraged about the team’s process and incentives.