Find XO: A study finds AI coding agents increased code output but not finished software (arstechnica.com)

No votes yet startupxo 5h ago1 editorial / 0 comments

Writing language: KoreanRead in the original language

- According to the source, the amount of AI-written code rose sharply, but because people spent longer reviewing that code, the resolution rate of tasks registered in the Jira issue tracker did not change in a statistically clear way. - Harvard researchers Fiona Chen and James Stratton analyzed data from Jellyfish, a service that measures development activity. The records cover more than 700 companies and more than 700,000 employees, from 2021 through March 2026. - At companies that adopted AI coding agents, lines of code generated rose by an average of 30%, commits by 20%, and pull requests by 23%. In the same period, average pull request review time rose by 49%, the researchers wrote. - The share of pull requests with requested changes nearly doubled, and comments per pull request rose by 35%. The share of staff assigned to review work rose by 14%, but the researchers said it is hard to attribute the employment change to AI.

Found on

Hacker News ↗ / 78 votes / 54 comments (checked 2026-10-13 03:37 KST)

Have you run into something like this? Tell us in a line or two what you did. You do not need to know the product.

1 comment

Editorial opinionstartupxo 5h ago

The first thing that caught my eye was the part saying that review time rose by an average of 49% at companies that added AI agents. Even if writing code gets faster, if reading and fixing it stays the same, the bottleneck just moves somewhere else. Still, the original doesn't make clear what size of company and what time period this figure comes from, so I think we need to consider whether it can be applied directly to our team. Have you ever tracked how long it takes from request to merge when your team reviews pull requests that AI produced?

Writing language: Korean

Related reads

1 comment