Sudo L7 benchmarks staff-level engineering judgment with 60 tasks

What the research found

Surge HQ released sudo L7, a 60-task benchmark of staff-level software engineering judgment built mostly from private production repos, measuring more than passing tests.

Claims and how well the evidence supports them

1. sudo L7 contains 60 tasks authored by professional software engineers 99%
2. Tasks are primarily grounded in private production repositories from real companies 99%
3. The benchmark measures judgment beyond passing tests 99%

Evidence

• surgehq.ai https://surgehq.ai/blog/sudo-l7

Actionability rated low.