A new study from Wharton and MIT researchers reveals that while AI coding tools significantly boost developer productivity, human bottlenecks in review and integration limit finished software output.

Key Points

• Researchers analyzed over 100,000 GitHub developers using AI coding tools between 2022 and 2026.
• AI coding agents increased individual coding activity by up to 180%, yet software releases only rose by 30%.
• The study identified post-coding tasks like reviewing, integrating, and distributing software as the primary production constraints.
• Data from the Apple App Store, Google Play, Chrome Web Store, and SourceForge shows a surge in new applications since 2025 without a corresponding increase in user adoption.
• Monthly new releases on the Apple App Store climbed from 30,000 to 100,000 by April 2026, while total app usage remained flat or declined.

Why it Matters

These findings suggest that current AI productivity gains are hitting a ceiling because they fail to address the complex, non-coding stages of the software development lifecycle. Businesses must recognize that simply accelerating code generation will not automatically translate into higher-quality software or increased market success.