A/B testing usually asks which variant should change. Far less often do we ask whether the experiment itself should change: the number of arms, allocation policy, stopping rule, promotion threshold, replay strategy, or quality controls that determine what the organization is able to learn.
That becomes especially important in adaptive and AI-driven systems, where the method can evolve almost as quickly as the product. Recursive experimentation treats the testing method as another governed object: measurable, replayable, challengeable, and reversible.
The goal is not only better experiments, but a system that can explain why its next experiment deserves to be different.
https://www.linkedin.com/pulse/your-ab-test-should-testing-itself-rogerio-figurelli-2vfbf