AI coding agents write 30% more code but resolve zero extra issues
AI coding agents are generating more code. They are not generating more working software. A new Harvard working paper makes that distinction with hard data for the first time.
What the Study Found
Researchers Fiona Chen and James Stratton analyzed data from the Jellyfish engineering analytics platform, covering 718 firms and over 725,000 workers between 2021 and 2026. Tools like Claude Code, Cursor, and Devin increased code production by 30%, adding roughly 1,500 lines per worker per month. Workers pushed 20% more commits and 23% more pull requests per month.
Resolved Jira issues and completed epics: no statistically significant change.
The Bottleneck Is Code Review
Faster code generation pushes more work into the review queue. Review times jumped 49%, up an average of 3.45 days. The share of pull requests requiring revision nearly doubled. Comments per pull request rose 35%. One senior engineer quoted in the study put it directly:
“We were able to write code faster in some instances, but there is only one bottleneck at a time. If you speed up coding, the bottleneck shifts somewhere else.”
14% more workers are now being pulled into review. That load falls hardest on senior engineers, who have to vet the increased volume of AI-generated updates.
The Measurement Problem
AI agent adoption climbed from near zero in late 2024 to over 95% of firms by early 2026. 80% of those firms use AI-assisted review tools, yet only about 23% of review comments are actually AI-generated. Human reviewers are still doing most of the quality control work, on a larger pile than before.
The paper’s conclusion is direct:
“Realizing the benefits of AI adoption requires complementary investments in downstream review capacity, rather than simply increasing code output.”
For operators running engineering teams, the takeaway is that the line item to watch is not code volume. It’s review throughput. Speeding up one stage of a system without relieving the next constraint does not increase output. It just relocates the pile.
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
