I haven’t read the entire specification, but I think there is a fundamental design problem. Why would someone put an encrypted message on a public feed that is completely useless to everybody other than the one recipient? This doesn’t make sense to me. It of course depends on the threat model, but wouldn’t one also want to minimize the publicly visible metadata (who is communicating with whom and when) when privately messaging? I feel there are better ways to accomplish this. Sorry, if I miss the obvious use case, please let me know. :-)