Meta on November 6, 2026 added a feature called spaces to Muse, its personal AI agent app: custom tools that a person describes in plain language and the agent then builds, from holiday gift trackers and trivia games to order trackers for small businesses. Spaces can be pinned to the app's sidebar or main tabs, scheduled to appear at set times and, once linked to other apps, refreshed in the background. In the same post, Meta said it is rolling out specialized versions of its models to Muse, trained with learnings from Muse Wonder.
In Short
Meta's Muse app can now build you small personal tools, such as a gift list, a quiz or a wedding planner, from nothing more than a description you type. It matters because these tools can pull in Instagram saves, Facebook Marketplace messages and emails and then keep updating on their own, which puts Meta's AI closer to the moment people shop and plan. Meta has not said what the feature costs, whether anything inside a space is used for ads, or when spaces can be shared with other people.
What Meta published on November 6
The announcement appeared on the Meta Newsroom under the title "Design the Muse App, Just for You," dated November 6, 2026. It runs to fewer than 600 words, carries a short introductory video and is written for consumers rather than developers or advertisers. "Today, we're introducing spaces, an interactive way to build a more personalized Muse," according to Meta.
The definition is deliberately wide. "A space is a custom experience you create inside Muse by describing your idea, like a holiday gift tracker, a daily word game, or a planning board for your next DIY project," the company wrote. A space, it added, "can be as practical or entertaining as you want," and every one created "becomes part of Muse."
The post arrived a day after another Muse announcement. On November 5, Meta published "Give Your Muse a Voice," which, according to that post, brings voice conversations and expressive avatars to the Muse app, starting in the United States. Two consumer features in two days show how quickly Meta is still filling out a product that is barely two months old.
Muse went live on September 8, 2026. PPC Land reported a US-only rollout at the time, across iOS, Android, muse.ai and a WhatsApp chat surface, with free use for most tasks and subscriptions for heavier use. By Meta's third-quarter earnings call on October 21, Mark Zuckerberg told analysts that more than 10 million people in the United States and Canada had set up Muse agents, and that about a quarter of them used the product on any given day.
Spaces follow the agent wherever it already runs. "Starting today, spaces are available everywhere Muse is available," according to the post. No countries are named, no devices are excluded and no subscription tier is mentioned. Whether building or refreshing a space counts against the usage limits that separate free and paid Muse accounts is not addressed.
Describe it, and Muse builds it
The mechanics are conversational. "Tell Muse about it, and it builds it for you," the post says. A first version is not meant to be final: "Your initial idea is just the starting point." According to Meta, a space can be adjusted through Muse over time, whether to add what is missing, change its look or alter how it works. Items can be checked off, notes added and edits made inside the space itself.
This is a consumer version of vibe coding, in which software is produced from a natural-language description and the person asking never sees the code. Meta's post does not use that phrase. Nor does it say whether a space is generated code, a configured template or some mix of the two, and it gives no indication that a space can be exported, inspected or run outside Muse.
Placement is the second half of the feature. Spaces can be added to the sidebar, moved into the main tabs, hidden when not needed and scheduled to appear at particular times. Meta's own example is to "keep your weekly planner up front during the workweek, then swap it for your weekend plans on Saturday." The consequence is easy to miss. Two Muse users may no longer look at the same app at all, because the navigation, the tabs and the tools on them become a product of each person's prompts.
Google has worked on related ideas for almost a year. With Gemini 3 on November 18, 2025, it introduced generative UI, in which the model produces custom pages, tools, games and simulations for each prompt instead of filling fixed templates. In April 2026, the Gemini app began building interactive simulations from a single prompt. Those outputs mostly answered one request at a time. Meta's design is built to persist: a space stays in the app, keeps its data and, in the post's words, "everything will be there when you return."
Nine examples, three of them commercial
Meta lists nine ideas. Six are domestic or personal. A school dashboard finds events relevant to a family's children in school emails and adds them to a calendar. A music tracker follows new releases and upcoming shows, while a wedding planner holds an RSVP list, vendor invoices and to-dos. The rest of that group comprises a trivia game about favourite films, music or sports teams, a podcast studio for episode ideas, research and scripts, and a trip planner with flights, hotel, reservations and a day-by-day itinerary.
The other three involve money changing hands. First on Meta's list is "A shopping hub that brings holiday and birthday gift ideas, wishlists from Instagram saves, and your Marketplace listings and buyer messages into one view." Further down sits "An order tracker for your small business, with new orders, customer messages and what needs to ship today." The third is a client hub for freelance work, with projects, deadlines and invoices.
The shopping hub stands out for its inputs. Instagram saves are the posts people bookmark inside the app; Marketplace listings and buyer messages are sales in progress. Meta already holds both, so a space built on them draws on the company's own first-party data rather than anything a user has to import. Marketplace itself has been taking on AI features for at least a year: in November 2025, Meta added collections, a collaborative-buying test and AI insights for vehicle listings, alongside partner inventory from eBay and Poshmark.
The order tracker builds on the small business product Meta released on September 29, 2026, which linked Muse to 15 named third-party tools, among them Shopify, Stripe, Intuit QuickBooks and Klaviyo, as well as Facebook Pages, Instagram professional account analytics and Meta ad accounts. Meta said then that nothing would publish, send or spend without the owner's approval. The spaces post does not say whether a space can act on what it shows, for instance by answering a customer message, or whether it only displays information.
Meta's closing pitch widens the field further. "Got a half-finished idea sitting in your notes app, a spreadsheet you keep meaning to clean up, or a hobby you want to dive into? Chances are you can turn it into a space."
A space that updates itself
The paragraph with the largest technical implications is three sentences long. "Connect the apps you use, and your spaces keep themselves up to date in the background, so new concert dates, price changes, or savings progress show up without you having to ask," according to Meta.
That describes an agent working while the user is elsewhere. How often a space refreshes is not stated. Neither is which apps can be linked, or whether updates arrive only through the connectors Meta listed for small businesses in September or also from Muse browsing the open web on the user's behalf.
Price changes are the example retailers will notice. If price tracking relies on merchant websites rather than a feed, background refreshes would add automated visits that no person triggered at that moment. Bot-management firms have been measuring Muse traffic since launch. DataDome counted Muse at 33% of agentic browser requests on its network in September after only nine days of measurement. Profound's tests found that Muse drew four of its six typical shopping picks from two merchant sites it browsed live. And at the end of October, DataDome's Jérôme Segura reported that Meta had begun signing some Muse requests with Web Bot Auth without announcing the change. None of those findings covered background refreshes, which did not exist as a named feature before November 6.
Sharing is still being explored
Spaces are private for now. "We're also exploring ways for you to share your spaces with others to build a more collaborative experience in Muse," according to the post. No date comes with that line.
The caution has a recent history. On October 20, Meta paused Muse's conversation-sharing feature after security firm Claroc found more than 6,800 share hostnames in passive DNS records, because Muse placed each conversation's title in the address of the shared page. A minority of the pages Claroc opened held loan details and internal revenue figures. A shared wedding planner with vendor invoices, or an order tracker holding customer messages, would carry material of the same sensitivity. The spaces post does not refer to the October pause or describe how sharing would be secured when it arrives.
Models trained with Muse Wonder
The final paragraph of the post turns to the models underneath. "We're also rolling out specialized versions of our models to Muse, trained with learnings from Muse Wonder, to make your overall experience even better," according to Meta. Muse Wonder is the model Zuckerberg called Meta's first frontier model when the company reported third-quarter results on October 21. On the earnings call it was described as available in Muse's paid tier, while Muse Spark remained the default for free users.
A research post linked from the announcement, published the same day under the byline Meta Superintelligence Labs, supplies figures. According to that post, the current Muse model scores 47% on Meta's Personal Superintelligence Index, against 38% for the previous Muse model. On an Active Memory evaluation, the current model correctly uses critical context in 90% of tasks, compared with 76%. Meta also reports a 2.4x reduction in forgetting critical information and a 67% improvement in completing everyday tasks where unexpected issues arise.
How much do those numbers say? Less than they appear to. The research post does not define the Personal Superintelligence Index, does not name the model versions it compares and does not state a baseline for the 2.4x and 67% figures. It does not explain how Muse Wonder was used in training, and it cites no outside benchmark. Personal superintelligence is the phrase Zuckerberg set out in his July 30, 2025 vision statement; without a published method, Meta's index cannot be compared with any other company's model.
There is also a small difference between the two documents. The research post says the specialized models are rolling out to people who use Muse and Meta AI, while the Newsroom post mentions only Muse. Neither says which subscription tiers receive them or on what schedule.
What the post leaves out
Several questions that matter to advertisers go unanswered. The post contains no reference to advertising, sponsored content or pricing. It does not say whether data flowing into a space - Instagram saves, Marketplace messages, school emails, invoices - is used to personalize ads elsewhere in Meta's apps.
That question has precedent. Meta said in October 2025 that, from December 16, 2025, it would use people's interactions with Meta AI to personalize ads, with conversations on sensitive topics such as health, religion and political views excluded. Muse arrived nine months later. On the October 21 call, Zuckerberg described an optional link: people who want Muse to help improve the personalization of their content and ads across Meta would be able to connect it to those experiences. An analysis by the newsletter MBI Deep Dives, covered by PPC Land on October 4, argued that Muse could shape Instagram advertising without ever carrying ads, and pointed to Meta's stated commitment to keep Muse conversation data out of its ad systems. Where a space, which mixes Muse output with data pulled from Instagram and Marketplace, falls under those commitments is not stated.
The business model is unsettled too. On the same call, Meta said it expected most people to use Muse for free, with businesses paying a fee on transactions Muse helps complete and heavier users subscribing. Under that model, a shopping hub that keeps gift ideas and price changes in view is a place where transactions begin. Meta has published no fee level, start date or merchant terms.
Why this matters for marketers
Spaces change what a Meta surface looks like. Feeds, Stories and Reels are designed by Meta and sold by Meta, with placements an advertiser can plan around. A space is designed by its user, and its contents are whatever that person asked for. No ad placement inside spaces has been announced. The data still sits inside Meta's walled garden, but the interface now varies from one person to the next, which leaves no standard screen for a brand to plan against.
Meta has put money behind the app. Sensor Tower estimated on October 20 that Muse passed 10 million US downloads in 40 days and that Meta directed 39% of its own October advertising spend to promoting it. The same estimates put 14-day retention among September's US installers at 23%. Whether personal spaces lift that figure is the most measurable test of the feature, though Meta has given no target.
The three commercial examples are where the effects will show first. A shopping hub fed by Instagram saves and Marketplace messages, an order tracker for small sellers and a client hub for freelancers all sit where Meta's ad business, its commerce tools and its agent overlap. Merchants and advertisers get no reporting from inside a space, and nothing in the November 6 post suggests any is planned.
Timeline
- July 30, 2025 - Zuckerberg sets out his personal superintelligence vision
- October 1, 2025 - Meta announces it will use AI chat interactions to personalize ads from December 16
- November 13, 2025 - Facebook Marketplace adds collections, a collaborative-buying test and AI insights for vehicle listings
- November 18, 2025 - Google introduces generative UI with Gemini 3
- April 9, 2026 - The Gemini app begins building interactive simulations from a single prompt
- September 8, 2026 - Meta releases Muse in the United States
- September 29, 2026 - Muse for Small Business connects to Meta ad accounts and 15 third-party tools
- October 1, 2026 - MBI Deep Dives argues Muse can shape Instagram ads without carrying ads
- October 2, 2026 - Profound finds Muse draws four of six shopping picks from two websites
- October 15, 2026 - DataDome reports Muse took 33% of agentic browser requests in September
- October 20, 2026 - Meta pauses Muse share links after 6,800 appear in passive DNS records
- October 20, 2026 - Sensor Tower estimates 10 million US Muse downloads in 40 days
- October 21, 2026 - Meta reports third-quarter results; Zuckerberg calls Muse Wonder its first frontier model
- October 21, 2026 - On the earnings call, Meta says more than 10 million people have set up Muse and outlines transaction fees for businesses
- October 30, 2026 - DataDome's Jérôme Segura reports Meta signing some Muse requests with Web Bot Auth
- November 5, 2026 - Meta publishes "Give Your Muse a Voice," bringing voice and expressive avatars to Muse
- November 6, 2026 - Meta introduces spaces in Muse and rolls out specialized models trained with learnings from Muse Wonder
Related PPC Land coverage
- Meta puts an AI agent that buys things inside WhatsApp, US only - The September 8 launch of Muse across iOS, Android, the web and WhatsApp.
- Meta's Muse agent gains access to ad accounts and 15 business tools - The small business layer that linked Muse to Shopify, QuickBooks, Klaviyo and Meta ad accounts.
- Meta's Muse can skip ads yet still shape Instagram ads, MBI argues - An analysis of how Muse could feed Meta's ad business without showing ads itself.
- Meta's Muse draws 4 of 6 shopping picks from 2 websites, Profound finds - Testing of how the agent chooses products and which sites it browses.
- Meta's Muse gains a 33% share of AI browser traffic in 9 days, DataDome finds - Bot-management data on the volume and origin of Muse requests.
- Meta halts Muse share links after 6,800 turn up in public DNS records - The privacy problem that led Meta to pause sharing in October.
- Meta's Muse gains 10 million US downloads in 40 days, Sensor Tower finds - Download, daily user and retention estimates, plus Meta's own ad spend behind the app.
- Meta's ads with creators gain 97% to a $15bn annual run rate - The earnings call where Meta gave Muse user numbers and its transaction fee plan.
- Meta's Muse agent gains a signed ID in a quiet test, DataDome finds - Meta's unannounced use of Web Bot Auth for some Muse traffic.
- Meta plans to use AI chat data for ad targeting starting December - The 2025 policy that put Meta AI interactions into ad personalization.
- Google launches Gemini 3 with generative UI for dynamic search experiences - Google's earlier approach to model-built interfaces.
Summary
Who: Meta, through its Muse personal AI agent app, with a related research post from Meta Superintelligence Labs. The feature reaches Muse users, and its commercial examples touch Instagram users, Facebook Marketplace buyers and sellers, small businesses and freelancers.
What: Meta introduced spaces, custom tools built inside Muse from a plain-language description, which can be placed in the sidebar or main tabs, scheduled, edited over time and refreshed in the background through connected apps. Meta also began rolling out specialized models to Muse trained with learnings from Muse Wonder, and its research post reports internal scores of 47% versus 38% on a Personal Superintelligence Index and 90% versus 76% on an Active Memory evaluation.
When: November 6, 2026, a day after Meta added voice and avatars to Muse, and two months after the app's September 8 launch.
Where: Everywhere Muse is available, according to Meta. The post names no countries; at the October 21 earnings call, Meta described Muse users in the United States and Canada.
Why: Spaces turn Muse from a chat agent into an app each person designs, drawing on Meta's own data from Instagram and Marketplace and on connected apps. For marketers, the feature raises questions the post does not answer: whether space data informs ads, how background price tracking reaches merchant sites, how sharing will be secured after the October pause, and how transaction fees will apply.
Discussion