Working in traditional document-driven software with a good autosave implementation is like using v2.5 of a really bad source control system.
In v1, before autosave, the system only kept one commit, but you could choose when to do it.
V2 let you commit when you want but also committed for you automatically, which is nice, but hopefully you don't want to revert changes.
V2.5 is like v2 except that it keeps a few of the older commits. The usefulness of this feature is almost completely eliminated by the fact that the software decides which commits to keep, not the user.
When you think about it that way, it's insane to think that it's the normal mode of operation. The technology and concepts to fix it already exist; the only reason we don't have anything better is because no one has figured out the right user experience.
In v1, before autosave, the system only kept one commit, but you could choose when to do it.
V2 let you commit when you want but also committed for you automatically, which is nice, but hopefully you don't want to revert changes.
V2.5 is like v2 except that it keeps a few of the older commits. The usefulness of this feature is almost completely eliminated by the fact that the software decides which commits to keep, not the user.
When you think about it that way, it's insane to think that it's the normal mode of operation. The technology and concepts to fix it already exist; the only reason we don't have anything better is because no one has figured out the right user experience.