Research on how coding agents work together.

We measure what happens when AI agents change the same code, on real public projects, and describe the method of every result so anyone can repeat it.