> no amount of UI design can really fix it.

This jives with something I've been saying and feeling recently but which I hadn't identified: that widespread adoption of cryptography needs to be predicated on including a basic understanding of async crypto in secondary education curricula. It's not a hard concept, but it's like referring to the unit circle around someone who has never taken trigonometry.

> Software should generate dedicated keys for every relationship, connection or context.

I agree with this, but it becomes tricky to map software to relationships. This is the next step in adopting cryptographic identities, but I don't know how you solve it without adding a ton of additional cognitive load for users.

The technical description itself sounds quite good, although I would be interested to see how you solve the networking layer, especially in terms of when to use direct connections, p2p technology and centralized intermediaries (like nostr relays).

You're a very good writer, I enjoyed reading this.