Meta has updated Muse, its AI agent app, so that it now works on iPad. The update came out on October 7, nearly a month after the first release of the mobile app. The Verge reported on it. The new version is built to use the iPad's bigger screen and its newer multitasking features.
Apple's App Store page now says Muse works on both iPhone and iPad. The page describes Muse as a personal AI agent. It can handle email, calendars, reservations, budgets, shopping, and health and fitness data. It can also work with linked social and business accounts. For any action, the user has to give approval first.
The update also brings new connections to other tools. According to The Verge, these are Canva, Dropbox, Figma, QuickBooks, GitHub, Zoom, Asana, and Klaviyo. The Verge also says there are new connectors for Granola and Notion.
The App Store page says users can link email, calendars, Instagram, Facebook, health and fitness data, and finances. It also mentions small-business jobs that use Instagram Business accounts, Facebook Pages, and Meta ad accounts. The Verge says a user can now give Muse a business goal. After that, Muse can plan on its own things like marketing, client pitches, and "new product pages."
Meta's App Store text stresses that the user decides which services Muse may reach. The user also approves actions before they happen. This matters because agent apps touch personal data, business systems, and things like payments. More connections let an agent handle more kinds of tasks. The approval step stays an important checkpoint before the agent does anything outside the app.
The page calls this release a modest growth of the platform, not a new model. For people who judge consumer or small-business agents, the key questions are what each connector is allowed to do, how much freedom the agent has, and what approvals are needed for outside actions.
The page sums up the news in three points. First, iPad support brings the agent to larger screens and iPad multitasking. Second, the new connectors cover design, storage, accounting, code, teamwork, and business tools, so the agent can take on more tasks. Third, agent products with many connections usually make permission limits and approval steps a main test for technical teams.
The page ends by saying the update does not reveal a new model, a benchmark, or the agent's inner design. That limits how much it matters for technical readers.
Text and photos: letsdatascience.com · 07.10 — original →
Meta introduced two new AI models, Muse Wonder and Muse Wonder Full, on October 12. They are meant for personal and professional agent tasks, which means the model carries out a job for you in several steps. Meta calls them its most capable models so far. Outside reporting describes the launch as the company's first push toward top-tier, frontier models.
Meta says the models can handle work that covers documents, coding, and special fields such as finance, law, and health. The company pointed to two named features. Active Memory, it says, carries useful information from one conversation to the next. Active Focus, it says, helps a model keep going on a task with many steps, try other ways when one fails, and ask a question before it does something with big consequences. The page stresses that these are Meta's own descriptions of its product and how well it works, not findings from independent testers.
Muse Wonder can be used by developers and builders through the Meta Model API and through Muse Code. Muse Wonder Full is open only to people who subscribe to Muse Code. TechCrunch reported that Meta sees the pair as a bigger and stronger follow-up to its earlier Muse Spark line. TechCrunch also noted that Meta has put a lot of weight on agent work and coding.
Meta says Wonder will later power parts of its Muse product, starting with subscribers. It gave no wider date for when that will happen. It also said a fuller report on safety testing will come later. So for now, what has really changed is that developers and subscribers can use the new models. A rollout of Muse to all consumers has not been confirmed.
The page advises readers to keep three things apart: what Meta claims the models can do, what protections the agents have, and what independent evaluation has shown. The release gives Meta's agent tools a new option. But public proof about how reliable and safe the models are in real use, and about how widely the product will spread, is still incomplete.
The page sums up its key points this way. Meta launched both models on October 12 for agent and coding work. Wonder is available through the Meta Model API and Muse Code, while Wonder Full is limited to Muse Code subscribers. Meta says the models add Active Memory and Active Focus, but it gave no wider Muse rollout date and no fuller safety report yet. The site calls this a significant model and developer-access release whose capability and safety claims still need independent checking.
Text and photos: letsdatascience.com · 12.10 — original →
Meta has updated its Muse app for iOS, and the update brings Apple Watch support. People who use Muse can now reach the AI agent straight from their wrist. The agent is described as proactive, which means it is built to act for the user.
On the watch, the Muse character appears on the screen. There are three ways to get in touch with it. A user can send it a text, speak a message that gets turned into text, or place a call to it. The update also adds a Muse complication, which is a small shortcut on the watch face for quick access.
Muse is Meta's personal AI agent, and its job is to complete tasks for the people who use it. It can be downloaded as a standalone app. It is also built into Instagram, Facebook and other Meta apps.
Meta designed a yeti-like character to stand for Muse. The page says this character has become popular. On the Apple Watch it is front and center, so it is the first thing a user sees.
The look of the watch app is close to the Muse Charm. Meta announced the Muse Charm in September as a companion device.
Text and photos: macrumors.com · 09.11 — original →
