The new parameter models work well but the idea of one single large composite plan is a bit confusing. This also goes against the plans for Agamemnon 2.0, where the large instruction coming from Agamemnon will be split into many smaller instructions. We should have a discussion, probably with analysis too, about what we would like the Agamemnon -> Hyperion interface to look like and what this means for the internal structure of parameters and plans.
The new parameter models work well but the idea of one single large composite plan is a bit confusing. This also goes against the plans for Agamemnon 2.0, where the large instruction coming from Agamemnon will be split into many smaller instructions. We should have a discussion, probably with analysis too, about what we would like the Agamemnon -> Hyperion interface to look like and what this means for the internal structure of parameters and plans.