Harvard study finds AI coding agents boost code output but not software output
A Harvard University study using data from Jellyfish and GitHub found that AI coding agents increased lines of code by 30 percent, commits by 20 percent, and pull requests by 23 percent across over 700 firms from 2021 to March 2026. However, the study found no statistically significant change in issue or epic resolution rates, suggesting the extra code did not improve overall software output.…
Key points
- AI coding agents increased lines of code by 30 percent, commits by 20 percent, and pull requests by 23 percent
- No statistically significant change in issue or epic resolution rates after AI tool introduction
- Increased code review effort absorbed efficiency gains from faster code generation
The story so far
7 episodes →- Harvard study finds AI coding agents boost code output but not software outputthis story
AI coding agents generate more code, but not more software
arstechnica.com · 12 October 2026
Loading the full article…
This text was published by arstechnica.com and written by Kyle Orland. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Hacker News discussion · 350 pointsnews.ycombinator.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Enterprise & Industry
All →- Salesforce acquires AI analytics startup Loops for Tableau integration · 1 src
- TechCrunch Disrupt 2026 begins with 10,000+ attendees at Moscone West · 1 src
- French companies deploy agentic computing with NVIDIA across space, finance and public sector · 1 src
- LegalOn halves costs while keeping development speed · 1 src
- Anthropic launches Cyber Mission with Critical Infrastructure Defense Program · 1 src
Comments
via GitHub Discussions