Your work stays yours.
Privacy is a route. In gOS, the question is not only what the system can do. It is where the work should happen, what context is needed, and who gets to decide.
The Mac is the center of the workspace. Local capability is used when it fits. Hosted capability can be used when it is allowed and useful.
01
Local work stays on the Mac when the task allows it.
Many tasks do not need to leave your computer. A local model can help with private drafts, file search, lightweight reasoning, and workspace actions where speed and privacy matter most. This keeps sensitive work close to the place you already trust.
- Use local models for private context.
- Keep files inside the Space.
- Avoid remote calls when local work is enough.
02
Cloud capability stays explicit.
Some work benefits from a larger or more specialized hosted model. gOS is designed to make that route a choice, not a surprise. Routing should balance privacy, capability, cost, and speed in plain language.
- Know when a task uses hosted capability.
- Send only what the task needs.
- Keep provider paths understandable.
03
Memory should be inspectable.
A personal AI system should remember useful context without turning memory into a black box. Mind is the durable surface for beliefs, preferences, project facts, and learning events. The user should be able to inspect and govern what the system learns.
- Review saved context.
- Correct stale assumptions.
- Approve important learning before it changes behavior.
04
Ownership is the default shape.
Spaces, files, flows, and outputs are designed around user ownership. gOS should make the workspace more capable without making it less yours. That principle shapes routing, memory, approvals, and the way reusable work is saved.
- Own the source files.
- Own the output.
- Own the workflow that produced it.
Privacy is not a toggle at the end. It is a decision made throughout the work.