Today I was thinking about some of the posts I made previously on bemoaning the web browser as becoming the ubiquitous user interface. I complained the technology was so complicated that it couldn’t be recreated easily–and that’s why there are only a couple of browser implementations. I (probably?) complained that many of the apps we use on computers do rendering with some sort of embedded web component. It’s easy, cross platform, and mature technology.
I wrote all that before AI agents were writing C compilers. I wrote all that before I stopped using IDEs and primarily programming through the terminal using tools like Claude Code.
Maybe it’s time now that someone attempts to use these new powers to build a new UI framework from scratch, in a safe/GC-free language like Rust, without extra dependencies. In the meantime, I will enjoy how coding agents have put more life into TUIs.
If I was making a new UI framework today, what might it look like? I am sort of spit balling as I type, but maybe something like an event loop over a socket. The language you write the program in wouldn’t need to be the same as the UI toolkit. Maybe you could send reduced form of HTML to render to display the UI toolkit components. The events could come like JSON RPC. That way your embeddable web view replacement wouldn’t need to be a full browser–it could be a lot simpler, but you could get most of the functionality.