Meta launches Muse, a versatile AI agent for the general public

Meta introduces Muse, a personal agent that combines a detailed activity feed, permission management, goals, a content library, and an animated avatar. Trials report mixed results in web navigation, highlighting a successful family PDF and fine-tuning of the news feed and prompts.
Navigation, Reported Failures, and Highlighted Animated Avatar
Trials indicate variable results when making online purchases through Muse's integrated browser: the purchase of New Balance 9060 was deemed "not great," while buying IMAX tickets for The Odyssey went more smoothly. The same feedback mentions failures that illustrate the current limitations of this category of tools. The interface also features an animated avatar, with a character named Slime, a teal dragon, succeeding Polly. A decision regarding this avatar is presented as an example of high-level AI product design.
What the Agent Actually Does: Activity Feed, Permissions, and Content
Muse showcases its operations in an activity feed that details tool calls and the step-by-step inheritance of tasks. The product manages permissions through an explicit mechanism and offers a customizable news feed. It includes an "Ideas" feature designed as a library of ready-to-use prompts, as well as a library that aggregates documents, websites, images, videos, and podcasts.
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisissez votre rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Tested Use Cases: Calendar, Family PDF, and Personal Goals
The journey begins with a dedicated login and integration. One scenario describes managing a family calendar, including the removal of a football practice and a request to produce a morning newsletter PDF. A tester notes that this PDF was generated all at once by Muse, whereas Claude and Codex would not have accomplished it correctly. Muse also allows users to set personal goals, cited for water intake, shoes, and sleep training, the latter being used to assess the agent's tone. Feedback also mentions a distinct permission model compared to competing agents and an activity feed deemed desirable in Codex and Claude Code. Finally, the tester discusses what they plan to use in the future, without providing specific details.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.