I mean they are text apps basically, they should be flying on modern PCs that can do close to a trillion ops per second. Inference is done elsewhere, so what's the deal?
Virtualization missing, that's commonly a difficult task to nail on a messages list.
another thing is mounting and unmounting complex elements efficiently, that was part of why switching threads or sessions felt heavy.
btw, native apps can feel heavy and janky too if not properly optimized for large lists. it's just a matter of shipping a good app that's battle tested with large data sets
