> Say I want to find a way to "crack" X software? done. Say I want to find a vulnerability in Y website? easy peasy.

OTOH, such a powerful model might be able to do formal verification practical, too. So your model might say "Can't get RCE on that software because the model that developed it also created a proof that that's impossible".

> Once everything can be created automagically... what happens?

The people who can come up with novel concepts would matter a lot more then than they do now. Like how image models can create pictures in the style of van Gogh, but they wouldn't if van Gogh hadn't existed. So we'd need "van Goghs" to come up with the new things that the models would create a thousand varieties of.

> Somehow, Hashcash was conceived to deal with spam/automation type of attacks, so I assume a version of that will have to be used to "structure" Cyberspace in the future.

"Proof of N" doesn't need to involve spending something. It could be "proof of citizenship" (or some other "proof of offline identity") combined with rate-limiting per identity. The right cryptographic protocols could do it without revealing your identity to the service you're interacting with.