After the big waves in the last two decades of leveraging AI and ML-based technologies for pushing test automation, in the latest couple of years the new big thing is undoubtedly LLM-assisted testing. LLMs are increasingly used for test case generation at differing levels of the testing process, for oracle derivation, for improving test code, as well as for debugging and code repairing. Several studies already show that LLMs can be an effective collaborator in different testing activities, but also warn about possible latent issues. Beyond the specific technique or tool, in this keynote I will talk on how LLMs are already largely transforming testing research. Questions that I would like to put on the discussion table are: can LLMs really substitute human testers? And to what extent can they solve the longstanding challenge of test automation? on what problems should software testing research focus? The talk will be based on a critical reflection of white and grey literature.