Topic · 2 articles

Evals

Guides 6 min

AI code review's absence blindness: measure what it misses

Teams reviewing a growing volume of AI-generated code need to score what a reviewer fails to notice, not only what it flags. A measurable protocol.

September 28, 2026

Guides 2 min

How Martian's Code Review Bench separates reproducible fixed-dataset evals from streaming real-world evals, and the tradeoffs hidden in each.

August 20, 2026