Im suggesting the unit tests come first, are made by the coder (maybe with assistance), and the LLM creates the feature code. And the unit tests prove the gener

Im suggesting the unit tests come first, are made by the coder (maybe with assistance), and the LLM creates the feature code. And the unit tests prove the generated code works or not then you tweak. Like TDD meets LLM