NASA’s new dark energy space telescope can also detect killer asteroids
technologyreview.com·3h ago
Cloud & Infrastructure
The New Stack·19d ago·5 min read
AI hasn’t shifted the bottleneck from coding to code review
TL;DR
The perception that AI has shifted the bottleneck from coding to code review is misleading; coding was never the true bottleneck. Instead, the real constraint lies in the deployment process, where many changes remain unlaunched.
✦ Why It Matters
Engineers should analyze their deployment processes to identify and eliminate delays in launching code changes.
Key Takeaways
- Assess how many changes are waiting for deployment after code review.
- Identify manual steps in your deployment process that slow down delivery.
- Avoid increasing code review speed without addressing downstream bottlenecks.
- Focus on reducing batch sizes to improve overall software delivery.
Full Summary
Many believe that AI tools like GitHub Copilot have moved the bottleneck from coding to code review, but this perception overlooks the entire deployment process. Research shows that over 90% of teams ship changes in batches, with many changes sitting in a queue after code review.
This indicates that the real bottleneck is not in coding or code review, but in the subsequent steps like testing and deployment. For instance, GitLab's report highlights that while AI adoption increases the number of pull requests, it also leads to longer review times and larger pull request sizes.
Consequently, simply speeding up code reviews can exacerbate the underlying issues in the deployment pipeline, increasing risk as unreleased changes accumulate. Addressing the batch problem is crucial for improving software delivery and realizing the full benefits of AI.