project
JCU
Jev Computer Use. You talk, the speech-to-text runs on your own machine, and a model works out which app, site or action you meant. Then it does it.
- open source
what it does
You can chain steps in one sentence: "open Chrome and search for TKS Vancouver, then open Notepad and type hello." Every step is planned before any of them run, so it never gets halfway and gives up.
where it is now
Open source, MIT licence. The first full run on camera did seven steps in a row without a miss. The bugs that mattered only showed up when I fed it things I had actually said out loud, not tidy test sentences.