This is the first thing I “vibecoded” a couple of days ago. I really like the game in general and now find myself playing a few times per day. I might try to add some non-classical modes to the game in the future, or I might just leave it as it is, not sure.

I've written a NEAT implementation in Rust a few years back. It was able to train XOR and pole balancing.

https://sgolem.com/blog/neat-simple-neuroevolution-framework...

Can this be run with llama.cpp?

yep. this guy keeps up to date with quantised models: https://huggingface.co/TheBloke

As long as users need to think about homeservers we can't expect a substantial migration.

As far as I'm aware there is work being done on the fully p2p version of Matrix where every device is a server. If that allows abstracting away homeservers, and if they solve multi device identities in a clean way, only then Matrix might be able to compete with an "ordinary" chat app for the majority of users.