i think this is the more important question: the 100 million PhD thought experiment. how much economically valuable work actually needs more intelligence than cheap models will soon provide?
that matters more to me than which subscription offers the best value at a particular point in time.
if what i currently pay $200 for eventually costs $20, will smarter models unlock enough new valuable work to keep me paying $200?
one anecdotal observation: every time a new model arrives, i struggle to imagine how it could get better or what else i’d use it for. but i always end up finding more things to do with it.