Let AI Handle the Tests

Working Code

Start Claude Code in a project with functions to test:

> Write unit tests for all functions in src/utils/math.ts.
> Use Jest and include edge cases.

Claude generates a test file. Run it right away:

> Run npm test

Any failing tests? Let Claude fix them:

> Check the failing tests and fix the code

Try It Yourself

Now let's try the test-first TDD approach:

> Write tests for an email validation function first.
> Test cases: valid email, email without @, email without domain, empty string.
> Don't create the function itself yet.

Run the tests — they'll obviously fail:

> Run npm test

Now request the implementation:

> Implement a validateEmail function that passes these tests

> Run npm test again

Confirm that all tests pass.

"Why?" — Why Testing Matters More in the AI Era

In a world where AI writes code for you, tests are the safety net that validates AI's output.

Claude Code + Testing Workflow

1. Ask Claude to generate tests
2. Run tests → confirm failures
3. Ask Claude to implement
4. Run tests → confirm they pass
5. Request additional edge cases
6. Repeat

Useful Prompts for Test Generation

> Write tests for this function. Include normal cases, error cases, and edge cases.

> Write integration tests for this API endpoint. Include success, failure, and auth error cases.

> Find files with low test coverage and add tests.

Deep Dive