I think we will see more frameworks that uses only ES modules and don't new a bundler/compile step, improving a lot the dev experience.
I hope in the future we have more of flutter and flutter like frameworks and less of the web, it'd be a little more cost effective as it'd reduce server costs and it'd lessen the effect of cloud companies on climate change.
The next big step in the web will be economic. It has gotten far too expensive.
Back when I started learning this stuff the idea was for ignorant moms and pops to quickly markup some text to create web pages. Now you need a mountain of technology thousands of times larger and more complex than people putting people on the moon just to put text on screen using layers of unnecessary bullshit like Next and React. WTF?
It’s only a matter of time before absolutely everyone, except for web developers, start looking at alternatives. Seriously, if the only goals are putting text on screen and handling user interactions both employers, hobbyists, and moms/pops going to consider anything else that costs less a million per year, minimum, for a technology team who delivers so incredibly slowly.
The next big step is dependent on JS language/runtime features. Specifically the Record-Tuple proposal (https://github.com/tc39/proposal-record-tuple)
TL;DR: It's a way to designate data as being immutable.
JS doesn't currently have a good way to do this, so everything built for it tries to assume that data hasn't changed since it was last set. This leads to headaches for developers and wasted cycles for runtimes.
When JS has a proper way to make data immutable, it makes it easier for libraries/frameworks to make guarantees about the underlying data and makes it possible for runtimes to better optimize code.
I wish Web App could be built in one single language as I despise JS. But for that, browsers need to be created differently.
The next thing in user interfaces is combining LLMs with GUIs to create a flexible interface that isn't fully defined in advance but works with the user to adapt to their needs. This is a total departure from the pixel-copying design methods we have today and there is no tooling yet for designing such an interface.
In a few years time, we'll expect all applications to have LLM inputs and full natural language understanding in addition to their menus and buttons, just like we expect all software today to make use of the internet where before, none did.