Blog
Writing on what we are building.
Notes on Starlight, edge AI, infrastructure, and the federal-grade software we ship.

Why AI needs a new isolation boundary
AI workloads decide what to do while they are already running. Containers and conventional VMs were built for software that did not. MicroVMs give agents and Kubernetes nodes a hardware boundary with a lifecycle that fits modern infrastructure, and confidential computing extends that boundary to the host itself.
→Kubernetes without the legwork: x0 in five demos
We have written about how x0 works. Now we want to show it. Five short terminal recordings take x0 from creating a Kubernetes cluster to provisioning s0 persistent storage, switching contexts, deploying and scaling an application, and inspecting it all in the x0 TUI.
Announcement·September 13, 2026·Mainsail engineering·8 min
→Confidential compute, ready on arrival
Mainsail and HIPER Global are bringing together Starlight and the HG-5610-2UF rugged edge server as one integrated appliance, with Intel TDX confidential computing configured before shipment. The confidential compute foundation is ready when the system reaches the site.
Announcement·August 12, 2026·Mainsail Industries·6 min
→A lighter foundation for Kubernetes
Bare metal ties Kubernetes to physical servers. Traditional VMs put another platform underneath it. x0, Starlight's Kubernetes microVM orchestrator, makes the node itself lightweight: isolated, schedulable infrastructure without a heavyweight virtualization layer.
Engineering·August 8, 2026·Mainsail engineering·10 min
→The agent is the computer
Why microVMs are the right isolation boundary for AI agents, why the model driving them belongs on your own hardware, and what we are building underneath Starlight.
Engineering·August 7, 2026·Mainsail engineering·11 min
→Grace Blackwell at the edge: adding the Dell Pro Max GB10 to Starlight
Starlight now treats the Dell Pro Max with GB10, NVIDIA's Grace Blackwell developer system, as a first-class cluster node. The Blackwell GPU and the local models that run on it are now part of the platform — managed, governed, and secured the same way every other Starlight node is.
Announcement·June 25, 2026·Mainsail engineering·4 min