EyeR, first of all, I agree that "the less power-centric, the less power can do when it turns evil". So yes, I have been a proponent of "small government", "less power concentration" principle.
However, we also have to admit 2 facts about human nature:
1) there are individual differences -- actually the whole concept of individual freedom is based on the assumption that individuals are different - otherwise (if everyone is exactly the same), there would be a single "best" approach for everyone to live, and inidividual freedom would mean nothing because "the best" is a single instance that universally applicable to everyone. (Now if you don't agree with this principle, it may lead to a much deeper discussion, about human nature, but then, the fact that we would disagree on such fundamental issues about human nature would "prove" that we are indeed different).
2) because of #1), we also know not every individual in a society (or nation) would be capable/interested/moral/trusted to perform every tasks.
Is direct democracy the answer to 2)? first of all, I support direct democracy to a large extent and I do believe democracy will evolve towards direct democracy in most jurisdictions; secondly, I don't think it's the silver bullet that can solve everything. There are some innate issues, for example, to your arguement about it is "cost ineffective" to bribe 600K people, actually it might be easier than you expected.... Someone make Universal Income as a campaign promise, campaign for $50,000/person/year is effectively bribing the citizens. Now this won't work day one because most people would know there was no budget. This will however, become a viable "bribe" once the "initial" portion is passed.
We also know from history - populist are not always best for people - in many cases, people can chose path that were against their best interest - either because of lack of knowledge, interest, information, or simply because we don't understand people enough - that what we think their best interest may actually not be what they wanted most.
So no, I don't consider direct democracy the magic that solves all the problems "simply".
Another area of concern is (due to #1) above), there are going to be cases where you still need concentrated power - whether it is building a large particle collider, or explore the space, they are all well beyond single person's capability, and will need "collaboration" which means using resources that are from many people (mostly through tax $$) to focusing on a few topics. Such projects can't manage itself. You will need a way to allow such concentration of resources to happen, and such resources to be managed. Such things won't magically happen by itself.
That's why despite I agree with you, the power needs to be closely watched, monitored, and restricted, I do not think it is possible, or practical to completely do away with power. Even in Science, there are different oppinions, research direction, what is the "best" area to get the funding? is direct democracy the best way? I'll give you a real world example, deep learning (neural network) was "scientifically discredited" back in the 80s - it nearly lost all funding in most part of the world .... other than in Canada where the Canadian Institute of Advanced Reseach funded such study for over a decade. It turns out that this was the area that is NOW the most hopeful approach of AI. The fact that Canadian government fund went "against" scientfic community consensus was actually saving us all from decades of delays in AI development. Coincidence? Maybe.
The other philosophical argument against "mob rule" had been established by most smartest mind from Socrates down. . .
The main reason that I can't "give up all previous human systems which all failed" was I don't believe our generation is that smart that we can ignore some of the smartest mind in the world, what they have done, tried, proposed, analyzed, and simply come up with something brilliantly flawless without drawbacks of all the systems we tried in previous thousands of years. Yes, if you are talking about python scripts to teach computers to learn, I can do better than even Einstein, but for human which took hundreds of millions of years to evolve, I don't believe I (or anyone I know) can simply disregard what have been tried before us, and invent something suddenly works better against the same human nature that has ruled us all for millenniums.