AI Coding Tools Increase Code Volume but Reduce Productivity Through High Revision Rates

Software developers using AI coding agents like Claude Code, Cursor, and Codex are generating significantly more accepted code, but productivity gains are undermined by high revision rates. Engineering analytics companies report that while code acceptance rates reach 80-90%, real-world acceptance drops to 10-30% due to subsequent revisions. Waydev, founded in 2017, recently overhauled its platform to track AI-generated metadata, providing insights into code quality and cost. GitClear's January report found AI users experience 9.4 times higher code churn than non-users, while Faros AI's March 2026 data showed an 861% increase in code churn with high AI adoption. Jellyfish's Q1 2026 study of 7,548 engineers revealed that those with large token budgets produce more pull requests but achieve only double the throughput at ten times the token cost. Atlassian acquired engineering intelligence startup DX for $1 billion last year to help customers assess ROI on coding agents. Developers report accumulating technical debt and increased code review workloads, with junior engineers accepting more AI-generated code and facing more rewriting. Despite these challenges, the industry views this as a permanent shift in software development practices.