You could have a command that recommends commands and then you select them on a drop-down list.

Alternatively if the dataset is verified you wouldn’t need to worry about it running dangerous commands, since it doesn’t know any. Or you could have a list of verified commands that run automatically and any command not on that list requires confirmation.

But this is missing the point that most of the time I know exactly what command I want to run so adding a LLM Is quite useless. The reason so much of linux is still relying on commands is because for a lot of people (myself included) commands are quick and efficient.

Ideology. every cent a capitalist doesn’t make because of me is a small victory.

All murders happen because of emotional (killing someone in anger), economical (Theft gone wrong) or psychological (Doesn’t realize it’s wrong) reasons. none of these is prevented by sticking the murderer in a box after the murder.

All of these are prevented by building strong social network to manage any harmful impulses before something happens, which is something any reasonable anarchist would agree with.

Also If you think the list is incomplete then feel free to give another example.

Oh yeah also political assassinations and wars. But your comment already addresses those.

I think a better wording is that anarchy is naive. And I’d rather be naive than accept that this is the best we can come up with, because that’s depressing.

matrix.org is basically IRC 2.0. It’s federated and has a lot of cool clients.

there should be an instances button at the bottom of the page that shows it.