Perhaps it's kind of inevitable to have some bloat. For example apps these days handle most of the languages just fine including emoji, LTR/RTL and stuff. Some have pretty decent accessibility support. They can render pretty complicated interface at 8k screen reasonably fast. (often accelerated in some way) There is a ton of functionality baked in - your editor can render your html or markdown side by side with source code as you edit it. You have version control, terminal emulator, language servers, etc..

But then there's Electron, which just takes engine capable of rendering anything and uses it to render UI, so as a result there's not much optimization you can do. Button is actually a bunch of DOM elements wrapped in CSS.. Etc.. It's just good enough for the "hardware is cheap" approach.

I think Emacs is a good example to look at. It has a ton of built in functionality and with many plugins (either custom configuration or something like Doom Emacs) you can have very capable editor very comparable to the likes of VS Code. Decades back Emacs had this reputation of being bloated, because it used Megabytes of RAM. These days it's even more "bloated" due to all the stuff that was added since. But in absolute numbers it does not need as much resources as its Electron based peers. The difference can easily be order of magnitude or more depending on configuration.

I have a few suggestions:

1. Better education. Don't scare people who're learning programming away from the lower-level stuff, especially as people are even getting scared to use type declarations, not just the pointers (of which I was fearmongered with in college, as they told me Java is the future).
2. Better portable APIs. Thanks to WebAssembly, one could easily have both something portable in a web browser and as a native desktop app, except instead we get browsers running said applications. I had some thinking about such a project, but then I remembered my iota project (a D-native replacement of SDL/SFML/GLFW, but without bloat by including standard library features), and then stopped thinking about it immediately, since a much smaller project already causes me too much headache. (Someone has a handy guide on win32 API? I have issues on getting certain messages produced, like input language change, and I don't know if I glimpsed over some functions that enable them and just weren't included in the documentation of the input language change event codes.)

nope. The bloat is there mainly because it makes the job easier for the devs.

It's not really inevitable, it's just a consequence that develops can get away with being lazy because the hardware can cope with it.