I’ve had some luck at using AI to get over the hump of the first “does this component work” test - it’s easy to detect stuff that needs to be mocked and put in stub mocks using GPT. GPT is horrible at writing good tests, but often it’s harder to write that first one than the meaningful tests.