How to Test Your Code. Even When AI Writes It.

Letting AI write your code and then using another LLM to generate the test cases is an architectural echo chamber. You are not validating correctness; you are just cross-compiling the same systemic hallucinations into your test suites. How do you guarantee integration safety when your AI-generated mock objects fail to replicate complex state mutations under production load?

@sahnlam When AI writes the code, I move test effort to the boundaries: contracts between services, data migrations and failure paths. Unit tests are cheap to generate now, so they also tell you less. A contract test catches the break that a tired reviewer skims past.

@sahnlam nice video man