But looking closer, there seems to be restrictions. First, instead of combining all computers together, it just duplicates computations on each single machine. So, instead of a super computer, we end up having millions of slow computers.
Second, I don't now if the term Turing complete is that meaningful in practice. To guarantee deterministic results, the Turing complete machine of ETH can't generate random numbers, can't accept inputs from 3rd party apis, can't have threading....
I don't know what realistic large scale problem it can solve.
Accurate descriptions need more jargon: for example, you could call it a peer-to-peer settlement network with autonomous business logic, or something like that.
Here's what I wrote on a reddit thread to pitch Ethereum in a simple way:
> If you want to operate some mechanism that involves binding agreements and reliable bookkeeping -- like a crowdfunding campaign, a remote employment contract, or even a digital currency system -- then you currently need to set up a highly secure central server, and every participant needs to trust you with the admin password.
> With Ethereum, you can upload the exact rules of your "game" to an enormous peer-to-peer network where each and every node verifies the integrity of every action, where forgery is made so difficult that even a well-funded state actor probably can't do it, and where all history is backed up across the internet.