Snap Launches CodePal: AI-Powered Code Review Assistant Handling 90% of Internal Pull Requests
Key Takeaways
- ▸AI-assisted coding tools increased Snap's PR merge rate by 60%, but created a code review bottleneck that threatened developer velocity
- ▸CodePal reviews 90% of Snap's pull requests before human review, providing context-aware feedback tailored to Snap's engineering practices
- ▸The tool reduces review queues and merge times while maintaining code quality through intelligent automation
Summary
Snap has built CodePal, an internal AI-powered code review assistant designed to address a critical bottleneck in its development workflow. As Snap engineers adopted AI coding tools like Cursor and Claude Code, the company achieved a 60% increase in merged pull requests year-to-date. However, faster code creation shifted the bottleneck downstream—code review processes became overwhelmed, PR queues backed up, and merge times grew.
CodePal was designed with a specific mission: provide highly intelligent, context-aware feedback that understands Snap's engineering practices and culture. The system reviews pull requests before human reviewers see them, offering immediate, valuable feedback to PR authors while dramatically reducing the review burden on engineering teams. Since deployment, CodePal now handles the initial review of 90% of all pull requests at Snap.
The tool embodies a collaborative vision where AI augments rather than replaces human judgment. By automating the high-volume initial review phase, CodePal frees human reviewers to focus on complex architectural decisions, design trade-offs, and nuanced feedback that require human expertise.
- CodePal demonstrates how AI can solve downstream bottlenecks created by earlier AI productivity gains
Editorial Opinion
CodePal represents mature AI adoption in software development—identifying a real operational constraint and building appropriate tooling rather than blindly accelerating upstream processes. As AI coding assistants become ubiquitous, other engineering organizations will likely face the same code review bottleneck. Snap's solution shows how AI can augment human expertise rather than replace it, handling high-volume tasks so humans focus on decisions requiring judgment and context.
