Review Is the Bottleneck
🌳 Evergreen ·
AI agents produce code faster than humans can review it, making review the structural bottleneck. Seven independent sources converge: Xu et al. (2025) show core developers review 6.5% more code but produce 19% less after Copilot adoption. SonarSource (2026) reports 42% of committed code is AI-generated with 38% saying review takes more effort. Google reports 75% of new code is AI-generated. Dao et al. (2026) find the field now needs predictive triage to determine which AI PRs need careful review. When you need triage, comprehensive review is already dead. The bottleneck is structural, not a tooling gap.