Local agents need visible authority.

Local is a location, not a guarantee

Running a model on the user's machine can reduce exposure, latency, and dependency. It does not answer what the model may read, what it may change, which tools it may invoke, or how the user understands the resulting action.

Privacy-first computing needs an authority model. The important question is not only where intelligence runs, but under which capability, through which broker, with which durable record.

The operating system is part of the product thesis

Ulori explores a Rust-first userspace where supervision, manifests, capabilities, and audit are not background plumbing. They are the conditions under which local agents become understandable enough to trust.

Bootable, early, honest

The current build is research, not a daily-driver claim. That distinction matters. Serious systems work should show its current boundary as clearly as its ambition.