I want to share an idea that I had that I have not had time to implement, and I believe would be a neat feature.
Edit: Leaving in tact here, but actually the next few paragraphs are wrong. rshares efficiency is the same, no matter what your voting power! Only thing that changes is your max weight
How many of you have the problem that your VP is constantly low? For example, if you are consistently below 75% power, you are basically voting at 75% of the level that you could be voting at in terms of your stake. Sure, if your vote isn't worth much, it isn't a big deal, but over time it's something that adds up. And the thing is, in many cases simply re-timing your existing votes, and possibly tweaking the scaling slightly would allow you to place the same values on the respective posts while having room for more votes in the day.
Remember the optimal timing to make use of your full stake is 10 100% votes spread evenly per day. Or 20 50% votes spaced evenly, etc. Or more generally, only voting when your VP is 100%.
Wouldn't it be great if you had a robotic assistant that could arrange the votes for you?
I imagine you would tell your assistant, "Hey, I want to vote person X for about $Y".
The most primitive form of the assistant would simply take it in, wait for your VP to be 100%, and cast the desired vote.
The power here is simply the ability to queue up multiple votes that would slowly cycle through, and you would be responsible for filling the queue up. The assistant will estimate how long it has before it has exhausted your queue, as well as if there are votes that area in danger of being too late (likely not a problem depending on how much you fill things up).
More advanced features would also be about taking curation into consideration. Depending on what votes are already on the post on question, it could decide whether it needs to react more quickly or slowly to take out a good chunk of future potential votes. But I think in most cases, one only decides how much they want to pitch in and not think about timing considerations. And this becomes even more true when HF20 rolls around, where really the only consideration becomes how much of the curation do you want to get.
Anyway, this is what I'm calling the efficient rshares cyborg. Cyborg because it's all human input for what to vote and at what value, and the bot part simply times the vote just right. And this can also mix with auto-voting as well, though if you're auto-voting on popular authors in a way that you want to maximize your curation for, then this won't be as good.
Does anyone else want this thing? shrug. Perhaps I also have the wrong idea about the question of efficiency too, in which case I would welcome some corrections. I suspect 
Edit: As pointed out by
If that's the case, then that's great news, all I really care about is whether or not I need to give heavier votes or not, and usually it's not the case. I just want to spread votes around. Well, I think I'm going to pretty much throw the cyborg out the window now after I verify it.
Second Edit: Oh. In the computation, both the rshares corresponding to the vote and the regeneration time for the amount of VP used are proportional to the VP used on a given vote, and hence the rshares assigned per unit of time needed to recover from the vote is kept constant. No rshares efficiency loss! The only thing to watch out for is that if you're at lower VP, your maximum vote/rshares assignment becomes lower. Now I understand that bit. Wellp.
destroys the robotic part of the cyborg and crawls out of the machine.