AI-Generated Code Passes Review but Hides Subtle Bugs, Developers Warn

A developer survey of 7,258 programmers found that AI now writes an average of 54% of their code, up from 28% a year earlier. The article argues that most developers were trained to write code rather than review it, a skill now central to their jobs. It notes that AI-generated code often looks clean and confident while containing subtle errors that trigger no warnings. Examples include async calls inside forEach that never get awaited, which can silently return empty results. The piece walks through several common JavaScript and React bugs AI tends to produce, each with a concrete fix and a review rule.