AI coding dependence reaches 80% of developers

A survey of 305 developers found that 80% felt their AI use had become more dependence than advantage; 43% kept coding after hours despite intending to stop, and 32% postponed sleep to continue. ZDNET’s report cites the underlying Coddy survey, while noting that fast output can create a slower verification burden when generated code is almost—but not quite—right. Engineering leaders should read the finding with GitHub’s agent control-plane push and measure review load and work-hour creep, not just lines shipped.