Meta is adding a feature called spaces to its Muse AI assistant, allowing users to describe customized, app-like experiences and interact with them inside Muse, according to The Verge’s November 6 report. Examples include a shopping hub, school dashboard, music tracker and trivia game. (The Verge)
The central promise is an interface that does more than display a conversational answer. According to Meta’s description relayed by the report, connected services can supply information that keeps a space updated. Those are advertised capabilities, however, not independently established measures of accuracy or reliability. (Meta)
From a description to an interactive experience
Meta’s proposed creation process starts with a user describing what they want. The resulting experience operates within Muse, rather than being presented as a standalone application. That distinction matters: an app-like interface should not be read as evidence of exportable software or a replacement for an existing application. (Meta)
The examples span organization and entertainment, but they should be understood as illustrations of the product’s intended uses. A shopping hub does not, by itself, establish that a space can complete purchases. Similarly, a school dashboard should not be taken as proof of access to school records. Those capabilities require separate evidence, not assumptions based on a label.
The Verge names Gmail, Google Calendar, OpenTable and Notion as examples of services that connect to Muse. That is not an exhaustive, verified integration list for spaces, nor confirmation that each connection supports every proposed workflow. (The Verge)
Updating information is part of the pitch
According to Meta’s description, outside connections can keep pricing, dates and other information updated. The practical appeal is straightforward: a custom interface would be more useful if users did not have to rebuild its contents whenever the underlying information changed. Whether spaces delivers that benefit consistently remains a testing question. (Meta)
“Continuously” should not be interpreted as a guaranteed real-time refresh schedule. The evidence approved for this story does not establish update intervals, stale-data warnings or what happens when a connection fails. A meaningful evaluation would compare what a space displays with its source information, including after a change or interruption.
For a shopping example, the relevant question would be whether a displayed price remains accurate when someone acts on it. For scheduling, it would be whether a changed date appears correctly. These are proposed evaluation criteria, not results from hands-on testing or reports of failures.
Access and permissions remain verification questions
Detailed country, platform, language, age, account and pricing requirements have not been established in the approved reporting for this article. Readers should not infer eligibility from another Muse feature’s rollout. This draft therefore does not provide a definitive access checklist or claim that spaces is free, paid or universally accessible.
Connected-account controls require the same care. The approved evidence does not establish spaces-specific permission scopes, read-versus-write authority, disconnection procedures or treatment of previously retrieved data. These are gaps in the verified account here, not evidence that controls are absent or that a privacy failure has occurred.
The distinction between displaying information and taking action is particularly important. Evaluating a connected experience means asking separately what information it can retrieve, what it can change and which actions require approval. A successful-looking dashboard alone cannot answer those questions.
A broader shift toward interfaces inside assistants
Other AI companies have also been developing experiences beyond text responses. In June 2026, Anthropic announced artifacts in Claude Code: interactive pages such as dashboards, system explainers and release checklists built from a working session. The company described pages that update as the session progresses. That is a developer-oriented comparison, not evidence of Muse’s performance. (Anthropic)
OpenAI took a different approach in December 2025, opening submissions for developers to publish apps in ChatGPT and introducing an app directory. Its announcement described apps that bring outside context and actions into conversations. Unlike Muse’s reported prompt-created spaces, that model centers on developer-built integrations. (OpenAI)
The same announcement also addressed data access: OpenAI said users would receive information about data shared when connecting an app and could disconnect it to end access. Those stated controls provide a useful comparison point for evaluating connected assistants; they should not be attributed to Meta or treated as independent verification of either product. (OpenAI)
Together, these announcements suggest competition over how assistants present useful, interactive work—not just how they answer questions. For Muse, the unresolved test is whether a generated space remains accurate, understandable and controllable after its initial creation. No independent hands-on evidence establishing those outcomes was retrieved for this article. (Meta)
