Relevant today as I unpick some unit tests I let AI write and turn out to be very plausible-looking at first and second glance, but turned out to test nothing of value when properly examined.