and relying mostly on Kbin's existing code as its base
This one most certainly not. We actually stopped porting kbin code a few months into the project, because it just was too much work and it was obvious that Ernest didn't want us to. So everything which changed on mbin in the about 8-10 months since, was purely our own work. Of course the basis will always be kbin, but the form will most likely change
We've been keeping the UI mostly as is, because we all like it, however on the backend site of it a lot has changed. The biggest problems kbin had were compatibility wise (federation) and scaling wise. These were the points where we made huge changes. The federation compatibility has improved a lot (yes there is still a lot to do) and scaling/performance has also improved a ton.
The biggest UI changes we made are:
new filter designs that work for threads as well as microblogs
a subscription panel
a usable instance wide modlog
a cake day display
and more stuff that I am forgetting at the moment (it's been a while since I looked at kbin and I am mostly a backend dev)
The backend changes we improved are (imo) more impactful:
(next release) direct messages are federating
(next release) pins federate
deleting users federate
magazine descriptions are federating correctly
mods federate
reports federate
incoming likes are working
the "hot" sort actually makes sense with lemmy content because it also looks at upvotes and not just at boosts
completely redone the hashtag system so it scales at all
completely redone the background worker system so it scales better (partly next release)
And these are only the changes I could think of in 5 minutes. We likely changed a lot more things, which I just forgot.