I feel like at this point when people make some claim about LLM they need to actually include the model they are using. So many "LLMs can do / cant do X", without reference to the model, which I think is relevant.

I am hoping that the LLM approach will face increasingly diminished returns however. So I am biased toward Sabine's griping. I don't want LLM to go all the way to "AGI".