Building Jev voice control inside a cloud agent session
The author describes using a Devin-style cloud agent workflow to iterate on projects like jev-voice-control in a plain Ubuntu session with no repo connected.
Working with Cloud Agents is just so cool, I don't know how to define that better. Having a cloud agent like devin where you just have all the tools you need, you have all the VMs with all the OS you need, your github account connected, vercel or other hosting connected. The way this works really sparks creativity, more than a desktop app or CLI, somehow this way of working is very easy and frictionless so you just chat with the agent and ideate and things just naturally come together. If I play around, I just open a plain ubuntu session with no repo connected, ideate, I might come with few ideas, I just ask the agent to create separate parallel session one per idea, switch the tab to X or my actual work and come back later. I've built Jev-voice control this way https://github.com/chris-wozniczek/jev-voice-control Or the Opus 5.5 animations/apps like Handel, Vertigo, Knuckles, or Leapshot Sometimes I have something I want to ask, I just open a side conversation and chat with the agent in the same session without interrupting his work. In the end I might quote it in the session chat and ask to add these features, remarks, tasks to his list. In the end all gets pushed to github, I one click merge the stuff etc. It is as of now the most fun way to build stuff. Plus when you are on the go, you pick your phone and build stuff for fun via PWA (Mobile app soon!) I still use local agents where it makes sense but the cloud agent thing is just more fun for some reason.