@existentialcomics I am pretty satisfied myself, after trying to engage Grok in a couple in-depth conversations. These things seem good at what you'd expect them to be good at: amassing a lot of training information into a kind of encyclopaedic summary good enough for answering simple questions where little genuine follow-up is expected, at least not of an intellectual sort. But start asking the machine specific follow-up questions about the material it's just regurgitated and...it just can't. All of its talk circles back round to reiterating the regurgitated information. If you catch the device out in a specific error or fallacy, it just starts apologizing profusely and tries to change the subject.
These things are stupid. They are very obviously extremely stupid and lacking in self-awareness, and yet there's been very little widespread willingness in the technology press to cast doubt upon the supposed miracle. I suppose tech reporters have been selected for their "optimism", generally--i.e. their willingness to allow tech execs to lie freely without being troubled by searching questions.