TDD approach could play the RL role.
But what makes you think the ai generated tests will correctly represent the problem at hand?
But what makes you think the ai generated tests will correctly represent the problem at hand?