Blog: notes from the evaluation lab.

What we're measuring, how we measure it, and what the results do — and don't — tell you.

9 Nov 2026Benchmarks
12 min read

Introducing ROBO SWE: can coding agents repair the software robots run on?

Real issues, real repositories, the developers' own tests. Why robotics software is its own engineering problem, how the benchmark is built, and what the first results show.

Read the post →