1 comments

  • frabonacci 4 hours ago ago

    hi HN - i'm Francesco, founder of Cua. Since we released Lume almost 2 years ago (https://news.ycombinator.com/item?id=42908061) we've used it to run agents and CI on macOS.

    In order to meet our customer demands for evals, and RL at scale, we decided to build our own vmm on Hypervisor.framework. We call it Hoy, and it provides faster boot time to ssh and faster resetting of VMs between runs (on the same M4 Pro: boot 5.3 s vs 6.6 s, restore 1.2 s vs 3.0 s with Apple's Virtualization framework), as well as better compatibility with graphic design software requiring gpu, and a gpu renderer path for running local models at close to native speed.

    We're releasing it for free, including commercial use. closed-source to start with. some limits: two macOS VMs per Mac, enforced to meet apple's macOS license, and no iCloud sign-in yet.

    If you meet any bugs or have questions, please post here or at https://github.com/trycua/cua/issues