🇳🇱
Yea with modern computers using an 4096 asymmetric key on any reasonable amount of data is not a problem at all. Though keep in mind if you were encrypting directly off the asymmetric keys for chat it is not the most secure approach, nor the most efficient, you can still use the asymmetric keys but you should be using them a type of key exchange generally referred to as "Perfect forward security" or just "forward security".. for example specifically the Diffie-hellman algorithm.
From the user perspective you appear to still be using your asymmetric keys, and those are where it tarts, but no actual encryption takes place using the asymmetric keys directly. In short it has all the speed of symmetric encryption, the convenience of asymmetric encryption, and is, in fact, more secure than either of these options directly
Anyway with that said on to trust levels... They are kind of an open problem, and it was the intent of keybase to address it. The problem is while you can decide who you trust if you wish to infer trust via friend-of-a-friend it can be rather unreliable and breaks down very quickly more than one degree out. Afterall you dont really know how rigorous any one person in the chain was, or if someone is compromised. Keybase hoped to link your keys to your public facing identity so as to make the process of identifying someone as who they claim to be (which is the hard step to do when it comes to assigning a trust level reliably).
But yea if you could find a good way to calculate trust levels of FoaF, and generally the trust levels set are reliable, then I could see this being very useful for any sort of interaction that requires cryptographic trust. As I think I said this is somewhat of an open problem in general.
