Very interesting.

Many feel a model should be good at all things, and when it isn't, they poo poo that model.

I think that even if MD's model isn't perfect at everything, if it can indeed drag multiple needles out of a vastly larger haystack, then that's cause for celebration.

If we end up with a range of models with a range of architectures to solve a range of real world problems, then laughing. One uses a hammer when a hammer is needed. If we only have hammers, we have a problem.

Beer all round.

And it is still 2024.