>>527630318
I feel like for the discerning eye it's a perpetual cycle of:
>spend hours trying to get things just right
>it finally works but only for shallow context
>give in and have dumb fun with the slop for longer context chats
>get tired of the slop
>return to square one
It's a sisyphean struggle of "there's a glint of the technology I want, but it's just not quite there". Every big model release is one step forward, one step back toward the ideal.