Muse Living Inside WhatsApp Is the Interesting Part, Not the Model

Meta's personal AI agent Muse is available through the web, dedicated apps, and directly inside WhatsApp chats, and the WhatsApp part is doing more work than the model itself. Every other major assistant has launched as a destination, a new app or a new tab you have to remember to open. Muse launches as a contact inside an app most people already have open all day.

That distribution choice changes the kind of task an assistant gets asked to do. A tool you have to open for is a tool people plan to use, usually for something that already feels important enough to justify the extra step. A tool sitting in your existing message list gets asked the small, half formed questions people would never open a separate app for, forward this, remind me later, what does this mean.

People choose exactly what Muse can access, whether it reads mail or can send on someone's behalf, and can revoke that access at any point, which matters given how much trust living inside a messaging app actually requires. The real test here is not whether Muse is a capable model, plenty are. It is whether an assistant with no dedicated screen of its own can still earn the kind of habitual use that a standalone app has to fight much harder for.