Harvard Working Paper: AI Coding Agents Shift the Bottleneck to Code Review
AI summary
The working paper by Harvard doctoral students Fiona Chen and James Stratton, based on data from 718 companies and 725,938 employees, finds that companies adopting AI coding agents saw about a 30% increase in code volume, but there was no measurable rise in completed work, and code review time increased by 49%. The paper also finds that by March 2026, nearly 80% of companies had used AI review tools, but AI generated only 23.3% of review comments, and the bottleneck remains.
Why read this
AI score 75
The paper uses data from 718 companies to show that after AI makes writing code faster, the review stage instead becomes the new bottleneck.
Topics
Body
Only a summary is authorized here. Continue at the validated source link.
AI Trends