OpenYabby

What if your computer didn't just respond to commands, but actually understood how to get work done? That's the idea behind @OpenYabby, and honestly it's what got me curious enough to dive in and make a video on it. Most AI products still feel like smarter chatbots. You ask, they answer, and then you're the one copying, pasting, switching tabs, opening files, running commands and finishing the workflow yourself. But real work doesn't live inside one chat window. It moves across your terminal, browser, filesystem, APIs, local apps and all the small decisions in between. OpenYabby is trying to solve exactly that. It's a voice first operating layer for Mac. You say "Yabby, create and ship this landing page" and it starts coordinating the work. It creates a lead agent, breaks the request into phases, launches specialized sub agents, runs CLI tools, reads and edits files, uses the browser, performs QA, and manages the execution loop from start to finish. What stood out to me the most wasn't just the automation. It was the orchestration. Agents working in parallel within each phase. Tasks moving sequentially across phases. Memory carrying context across sessions. Real system access through terminal, AppleScript, Playwright and the browser DOM. It's not a staged AI demo. It's actual work happening live on the machine. There's also a voice layer built around how people naturally talk. Wake word detection. WebRTC voice sessions. Bidirectional audio. Speaker verification. So instead of typing prompts or clicking through tools, you just talk to your computer and delegate. It's still early stage, and reliability and execution quality are clearly things being improved on. But watching it work made me feel like this is where personal computing is actually heading. From using software to managing intelligent collaborators. I made a full video showing the entire process, from giving a command to the final result, watch below 👇🏻