Cosmo aims to give AI agents a desktop UI, launched by Shiyuan on X
In an X thread, Shiyuan says Cosmo lets you type or speak from the desktop while the interface renders live; some demo changes are not live yet.
By Ryan Merket · Published
Why it matters
Agent tech has raced ahead of practical interfaces. A usable, real-time UI layer could turn agent demos into workflows operators actually adopt on the desktop.
Shiyuan Liu (.@shiyuan_l) unveiled Cosmo, a UI layer for AI agents that generates interfaces in real time from text or speech input, in a thread on X.
Framing the problem as human, not machine, Shiyuan wrote: "AI needs an interface. Actually... no. AI doesn't. It's us, humans, we need visual interfaces to interact with computers. And that's what AI agents are lacking." The pitch: "Cosmo adds the UI layer to AI agents. Just type or speak from your desktop, and UI generates in real time."
https://x.com/shiyuan_l/status/2057585183902310470/video/1
A short video in the thread shows the concept in action, and Shiyuan shared the product site at buildcosmo.com. In a reply to Robert Scoble (@Scobleizer), @shiyuan_l added, "Some of the changes shown here are not live yet but we're working fast on shipping them." Shiyuan also tagged Founders Inc (@fdotinc) in the thread.
The focus is a thin, responsive UI layer for agent workflows rather than another back-end model or orchestration framework. If Cosmo can make agent interactions feel native on the desktop, it could broaden where agents fit in everyday work and who is comfortable using them.