@despotic_machine thank you. This sounds interesting!

@fediverse I’ve read that this is called an overlay network. Unfortunately many of the ones I’ve seen documented focus on keeping things in their own private networks which is okay but not fun.

ULA addresses require no permission and were designed precisely to knit together private networks. We can just use domain names and convert them via checksum into a static ULA /48 prefix. DNS can be used to announce routes, or eventually something more BGP-like given that ownership of a domain can be verified and thus authorization to announce routes.

If domains ever become a bottleneck one could use private TLDs with some consensus mechanism and even create multi-layer networks this way where packmates.layer.1 and packmates.layer.2 are two different networks even though they might have the same address range.

Anyways, I’ll go out and touch some grass now.

@nysepho @fediverse there would be routing without being peered directly by delegating your endpoint to another peer you trust (this can create an infinitely long routing chain depending on where you latch on so to speak, but you would be in control)

@yote_zip @HappyFrog RIP gnome 2.

@baronvonj I don’t think that’s currently implemented

@BrooklynMan while I realize that it sometimes doesn’t make sense to care too much for how mastodon users see things, maybe there’s some way top posts and comments can be distinguished so that we can filter them out. But I do understand that there’s limitations and it’s a reasonable compromise.

@csolisr @lemmy the post is the title. It’s one and the same, I believe. It also added a weird link at the beginning as seen from lemmy.ml but maybe I should have added the mention at the end instead of at the beginning.

@BrooklynMan yeah, this is pretty awesome. The only big thing I’m struggling with is that the community account, as seen from Mastodon, boosts absolutely everything… all comments, in any order, not just top level posts.

@gzrrt @dl007

The biggest problem with GPT-4 is that it’s severely restricted.

I believe that AI’s next big change will be model optimization and open source models will be able to catch up.