Meta tried to shrink engineering teams around AI
20 days ago
- Meta's AI 'pods' boosted code changes 220%, but shipped features rose only 36% while incidents jumped 40% and firefighting time increased 70%.
- Across 500+ organizations, median weekly pull request throughput rose 37% but developer experience fell and time spent building new capabilities remained flat.
- Recommended team size is five to eight people; shrinking teams should be the last step in AI transformation, not the first.
- AI amplifies existing issues like code review bottlenecks; product engineers and platform engineers become more critical.
- Leaders should focus on outcome metrics rather than output numbers to avoid producing more stuff without useful results.