Every AI writing tool eventually has to answer the same uncomfortable question: who wrote this? Most tools dodge it. The AI stays invisible, the output appears, and the writer is left guessing how much of the work was actually theirs. NarrativAI was built to refuse that dodge entirely, and Muse is how that refusal takes shape in practice.

Muse is built around five named states, always visible to the writer. Three are live in the Beta today — Listening, Drafting, and Safety Checking — with Suggesting and Resting still on the way. Each one appears in a collapsible panel while you write, so you always know exactly what Muse is doing in the background. Not as a gimmick. As a design commitment. Your creative decisions and Muse's contributions need to be legible to you, because your universe belongs to you.

Listening is the state most writers will see most often. Muse is reading what you've written and building a live model of your characters, your lore, your tone. It's not generating anything yet. It's paying attention. Drafting is Muse writing alongside you, at your explicit invitation, always in a separate layer so it's never confused with your own prose. Safety Checking is the confirmation beat right after a generation — a quick "this stays within your consent settings" check that surfaces without adding friction to every save. Suggesting, where Muse flags something worth your attention like a potential continuity gap, is still on the way.

The reason the states are named and visible isn't transparency for its own sake. It's because a co-writer you can't see is a co-writer you can't trust. Muse is designed to earn that trust explicitly. So when you do hand it a scene, you know exactly what you're handing over, and you know exactly what comes back.