I'm glad you called this out. p-hacking can be useful to generate hypotheses, which ought to be then tested (rather than thinking the p-hacked conclusions are just that)