Their grass really is greener, and our local vendors should be trying harder.
My response to this is always the same: Our software is FOSS, if you don't like some parts of it, go fork and improve it yourself, instead of demanding that somebody else does the work for you.
The dual root partitions we described in Deepin 20.5 are gone, but version 23 still sets up a moderately complex partitioning scheme, including an EFI system partition, a 1.5 GB /boot partition, a swap partition, and a 15 GB root partition, and the rest of the disk given to a partition labeled _dde_data. All are in plain old ext4 format, but there's some magic being done with the data partition that we didn't have time to trace. It appears to be mounted at multiple places, including /home, /var, /opt, and a mount point called /persistent beneath them all. We're not sure exactly how it's been done, but the distro has some kind of atomic installation facility with rollback.
Lack of proper documentation by Deepins Devs is enough of a red flag for me to never consider trying it.
If any KDE developers read this, we seriously and strongly recommend that the KDE team takes a long, close look at Deepin. DDE shows both how pretty and how functional a Windows-style desktop can be, without needing Plasma's multiple overlapping options. This is what we'd hoped to see in Plasma 6, but didn't get. It's probably fair to say that Deepin is the most sophisticated implementation of a Windows-like interface on any OS these days. It's considerably more flexible and configurable than Windows 11 itself, and it makes Cinnamon, MATE, Budgie, and the others look clunky and old-fashioned. Don't get us wrong; this grumpy old vulture is happier with the relatively austere and Spartan Xfce, but for those who want something that's shinier out of the box, DDE shows how it ought to be done. Even Microsoft could learn from the Chinese developers.
I am not apprehensive towards AI/ML but I don't want LLMs in my system. I have no need for it and it will just unnecessarily slow down my system. Instead, I want useful ML stuff, some autocorrct or autocomplete or better image processing in my camera app, you know the basic ML stuff that smartphones do since like 2017.