Harvard researchers find AI coding agents shifted the bottleneck to code review (complexdiscovery.com)

🤖 AI Summary
A recent study by Harvard Ph.D. candidates Fiona Chen and James Stratton reveals that the introduction of AI coding agents in 718 firms led to a notable increase in code production—30% more lines of code added per engineer—but did not correspond to an increase in completed work. In fact, the time taken for code review surged by 49%, resulting in longer queues and a doubling of formal change requests. This suggests that while AI tools enhance coding speed, they inadvertently create bottlenecks at the code review stage, where quality control measures seem to demand more time and scrutiny. This research has significant implications for the AI/ML community, highlighting the need for a careful evaluation of AI's role across different stages of software development. The findings indicate that faster drafting does not equate to greater productivity, as the review process becomes the critical choke point. The study employed an extensive dataset from Jellyfish analytics, analyzing millions of coding activities to arrive at these conclusions. Such insights call for a re-examination of how organizations integrate AI tools, emphasizing the importance of maintaining robust quality checks even as automation accelerates initial workflows.
Loading comments...
loading comments...