Known Limitations

As we continue to develop tapioka.ai, there are certain scenarios and patterns that the platform currently handles with limitations.

1. Fast-Changing Views (eg. Video Players)

Tests involving views that change significantly in real-time—such as active video players—may be unreliable. By the time the AI decides which element to interact with (e.g., clicking the player overlay), the UI state might have already changed or the overlay may have disappeared.

2. Single Expected Result

Currently, each test case supports exactly one Expected Result check at the very end.

• Manual Tester Analogy: Unlike a manual tester who checks results at every step, the AI focuses on reaching the final destination successfully.
• Current Workaround: If you need to verify multiple intermediate states, split the journey into several smaller test cases.

3. Dynamic Backend State

Testing actions that change the global backend state (like deleting a unique user profile) can make tests non-repeatable. We recommend focusing on "read-only" explorations or transient states (like logging in/out) until further support for state resets is implemented.