Can LLMs "invent" life hacks?
If given a problem with everyday objects are they capable of at least rarely responding with useful novel solutions? Even rare anecdotal evidence might be interesting if it gives better chances than the monkey typing
They are "capable", but only by chance. If you generate millions of random sentences in a human language then some of them would express some "useful new solutions". LLMs would have a higher chance of this happening as they generate sentences that are statistically more probable and therefore more likely to represent a coherent thought, but this would still be a chance.