SWE-CI: Evaluating Agent Capabilities in Maintaining Codebases via CI

[maxpla3]

Hi Dex,

you write " There are no good benchmarks for a model's ability to maintain codebase quality. "

This might be interesting for you:

SWE-CI: Evaluating Agent Capabilities in Maintaining Codebases via CI
https://arxiv.org/html/2603.03823v3

I'm half through reading " Why Software Factories Fail " and so far it absolutely aligns with my experience.

Thanks for sharing your knowledge!