GoKawiilStudy: AI coding agents boost code volume but not software output
Harvard researchers Fiona Chen and James Stratton analyzed Jellyfish analytics covering 300 million work events across more than 700 software firms and 700,000 employees from 2021 to March 2026. They found that while AI coding assistants and agents increase the volume of code generated, firms show little evidence of increased software output or reduced employment, because human code review becomes a bottleneck, with longer reviews, more revision requests, and more reviewer comments.
GoKawiil's interpretation of the reporting above, not reported fact.
The findings suggest that gains from AI-generated code may be offset by downstream review and quality-control costs, meaning raw code output is a poor proxy for actual productivity. This could temper expectations among companies and investors who assume AI coding tools will directly translate into fewer engineers or faster shipping, since the data implies organizational bottlenecks may persist regardless of coding speed.
- AI tools increase code volume but not measurable software output or employment reductions, per the study.
- Code review time and revision requests rise significantly after AI tool adoption, acting as a bottleneck.
- The analysis draws on data from over 700 firms and 700,000 employees via Jellyfish analytics, spanning 2021–2026.
JetBrains All Products Pack — If code review and quality are the real bottleneck in AI-assisted development, a strong IDE suite with deep static analysis and refactoring tools can help engineers catch issues before they ever reach a pull request. JetBrains tools are widely used by professional developers precisely because they surface bugs and style problems early, reducing the downstream review burden the article describes.
See JetBrains All Products Pack on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: arstechnica.com, 2026-10-09
Published there as: “AI coding agents generate more code, but not more software”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.