web-applets

That's specifically what I'm working on at Unternet [1], based on observing the same issue while working at Adept. It seems absurd that in the future we'll have developers building full GUI apps that users never see, because they're being used by GPU-crunching vision models, which then in turn create their own interfaces for end-users. Instead we need apps that have a human interface for users, and a machine interface for models. I've been building web applets [2] as an lightweight protocol on top of the web to achieve this. It's in early stages, but I'm inviting the first projects to start building with it & accepting contributions. [1]: https://unternet.co/ [2]: https://github.com/unternet-co/web-applets/

For those interested, I've been working on something related to this, Web Applets – which is a spec for creating AI-enabled components that can receive actions & respond with state: https://github.com/unternet-co/web-applets/