What I find fascinating is the idea that there might be a set of "secret" tweaks that when applied to those weights (or even smaller models) could result in an intelligence simulation that could vastly surpass even something like Fable.