Most AI interfaces expose capability by adding chrome. There is a prompt box, an activity stream, a context inspector, a tool log, and sometimes a second navigation system for the assistant itself.
This can be useful for operators. It is exhausting for everyone else.
Fylune takes a different route. Its document editing features keep the file central, show review state only when it matters, and leave permanent AI control panels out of the writing surface.
Keep the document central
In Fylune, the document should remain visually continuous. AI activity appears when it needs a decision, then gets out of the way. The interface should make the state of the file clear without turning writing into air-traffic control.
Show consequences, not theater
People need to know what changed, whether it is safe to save, and how to recover. They do not need a dramatic animation for every model token. We prefer reviewable diffs, plain-language conflicts, and stable status over synthetic busyness.
Calm is a feature
The strange ambition is to make a more capable workspace feel like less software. That requires discipline: fewer permanent regions, consistent resource frames, predictable keyboard actions, and motion that explains change rather than decorating it.
A useful boundary
Calm does not mean hiding risk. Fylune still needs to show when an external tool changed a file, when a save conflicts, and when a proposed edit needs review. The interface can be quiet because those states are explicit, not because they are invisible.
This is also why the local editor stays free. A trustworthy core should not become less legible when a trial ends or an account is unavailable.