What's the use in picking a proof of work function that is easy to optimise in the future though? Isn't this like a new Bitcoin coming out that picks a proof of work function that is easily adaptable to ASICs? I thought the whole point of proof of work functions is that they're meant to be difficult?