Coding Agents Increase Code Volume but Not Software Output
October 12, 2026
A study of hundreds of firms finds that AI coding agents increase the volume of generated code but do not increase overall software output. The efficiency gains are absorbed by a bottleneck in the human-led code review process.
HOW THIS AFFECTS YOU
●
builderYou should focus on tools that assist in code review and verification, not just generation.
●
founderThis indicates a market opportunity for AI agents that can handle the full SDLC, including review and testing.