Every year it gets more accessible to a wider audience. Soon there will probably be frameworks that hide the complexity completely and you can just say here’s a massive dataset, I want to train it to be a conversation bot or cat pic classifier, go. But we’re not quite there yet.
Synchronizing local and remote code shouldn't take much time, but it's still at least a few seconds on the critical path for the run->fail->fix->rerun loop.
VSCode's remote mode might be a worth a try for people with such a setup.