
we spent the last few months building our own vmm for macos on apple silicon hoy vms boot 20% faster than apple's virtualization framework, and run local models on the mac's gpu at close to native speed it's free
This tweet is unavailable

the CUA wizard ʕ•ᴥ•ʔ @trycua

we spent the last few months building our own vmm for macos on apple silicon hoy vms boot 20% faster than apple's virtualization framework, and run local models on the mac's gpu at close to native speed it's free
This tweet is unavailable


If you use macos you need to see this asap
This tweet is unavailable

And it runs local models with GPU acceleration.
This tweet is unavailable

Apple's macOS 27.3 now warns you before you give an AI app full control of your Mac. DHH's answer: get a new OS. 🖱️ The numbers back him up. @trycua and Scale AI had an agent use the computer while a person kept typing, the same 7 tasks on 4 systems. How often the agent left your keyboard, mouse and windows alone: Omarchy: 97% macOS: 83% Windows: 41% Ubuntu: 40% Why 𝗢𝗺𝗮𝗿𝗰𝗵𝘆 wins: the agent gets its own second cursor and keyboard, drawn by Hyprland itself. A pop-up can't steal your typing either. Cua built it with DHH. Where it's still rough: drag and drop, and Chromium/Electron apps (a fix is in review). Would you let an agent work on your desktop while you do?
If you're excited about the age of agents, Apple has a message for you: "Agents? Not even once!" Best get you a new OS, bro.



this is great work. apple needs to rip the bandaid and make the density economics work.
This tweet is unavailable




Cua in @antigravity free student edition is way better than @GitHubCopilot It works especially well with apps that require authentication. Browser integration with CUA is very disappointing. Maybe Microsoft should consider proposing an acquisition of @trycua

CUA 做了针对 agentic 场景优化的 vmm,各种性能非常好,还有快速 fork 的能力,也有 cloud 服务,我要试试给 ducter
This tweet is unavailable

20% faster local MacOS VMs
we spent the last few months building our own vmm for macos on apple silicon hoy vms boot 20% faster than apple's virtualization framework, and run local models on the mac's gpu at close to native speed it's free

