[zozbot123]

You could do this quite simply by requesting trusted apps to be identified by Named Information (ni:// or nih://) URIs (see https://tools.ietf.org/html/rfc6920 ) using a digest algorithm of sufficient strength. But the ability to "seamlessly" replace web apps is something that many websites would insist on, I think. Of course ni:// and nih:// can be applied to documents as well. They work on the IPFS model, where you enter some digest of the desired content as your URI (it's actually a URN, not a URL!) and then it's the user agent's job to fetch it from wherever, perhaps in a decentralized way.