Scaling Mobile Chaos Testing with AI-Driven Test Execution
Uber's AI system found 23 hidden app risks, preventing crashes and failed orders.
Deep Dive
Researchers at Uber built an AI system to automatically stress-test mobile apps by simulating real-world backend failures. It combines an AI that navigates apps with a tool that injects faults. Since early 2024, it has run over 180,000 tests, saving an estimated 39,000 hours of manual work. It uncovered 23 resilience issues, including two app crashes and twelve severe problems that could block ride or food orders.
Why It Matters
This makes apps more reliable for billions of users by automatically finding and fixing hidden weaknesses before they cause problems.