It’s necessary to lean on something like xDai for now because the compute and storage for Circles have to come from somewhere. Whatever that source is, it must be reliable and resilient to attacks. Circles is too small to have powerful enemies right now, but if it’s successful we can expect it to attract the attention of people who would harm it.
Traditional economic incentives are currently the best way to get access to such resources: We pay for our compute and storage, and in return the people who keep xDai up and running make sure that we’re hard to attack. It’s very expensive to attack Ethereum, and therefore it is very expensive to attack Circles. (I’ve been meaning to look into how xDai achieves this at such a low cost, but for now I just trust that it does).
At least in my community, the circles web of trust doesn’t mean too much. It’ll take some practice, strategy, and determination to make it into something that I trust to provide me with important things like food. But imagine a world where circles trust was at least as reliable as bank-issued money for ensuring that your needs are met. With such a resilient trust network available, it should be possible to turn circles into a layer 1 protocol (i.e. it would not depend on any other token). This would require a certain saturation of trusted individuals who are willing to run nodes for their communities, and communities that are willing to support that activity.
Proof-of-work (like Bitcoin) and proof-of-stake (like xDai) are only necessary in a trustless context. If you’re already trusting your neighbor the farmer to grow the food that you’re going to eat, I don’t think that it’s so much of a leap to trust your neighbor the techie to run a node that supports the currency that you’re paying to the farmer.
Probably you would want to trust multiple techies that don’t trust each other, that way there’s a reduced likelihood that several of them conspire to cheat. (Luckily, it’s easy to find two techies that don’t trust each other. Just pick an emacs user and a vi user )