More AI Code, Less Production-Ready

AI tools are generating 110% more code. Lead times are up 480%. Code churn is up 861%. Something is very wrong.

This is a commentary on the Faros AI Engineering Report 2026 (“The Acceleration Whiplash”), surfaced by Kent Beck’s LinkedIn post. The data covers 22,000 developers across 4,000 teams.

Kent Beck’s post: https://www.linkedin.com/posts/kentbeck_the-ai-engineering-report-2026-the-ai-acceleration-share-7471011014738841600-JErm/
Faros report: https://www.faros.ai/blog/ai-acceleration-whiplash-takeaways

What this video covers:

• What the report’s headline numbers actually mean (and what they don’t)
• Why theory of constraints and queuing theory predicted exactly this outcome — and why the report’s authors miss it
• The 861% code churn stat that matters more than any productivity metric
• Why 31% more PRs are merging without any review — and why that’s a security problem
• Why the report’s 10 recommendations treat symptoms, not causes
• What I think needs to happen instead: small teams, pilot-copilot model, TDD at agent level, end-to-end ownership

The bottleneck didn’t disappear when we sped up code generation. It moved downstream — and it’s building pressure at every stage.