Muse

So Muse hasn't even reached a million users yet, and it's already struggling to serve them. The status monitoring platform SaaSHub currently lists Muse as 'Degraded,' indicating that 'users are reporting problems.' The platform lists 'can't search' as the most common complaint.

According to a tabulation by Similarweb, the DAU metric for Muse is now up 10x over the past 11 days to hit around 700,000. While this is phenomenal growth by any measure, the total scale of the AI agent's proliferation is still quite modest. And yet, a growing number of ancillary evidence points to the conclusion that Meta is already struggling to serve Muse to its growing userbase.

By packaging device cross-sync, full disk permissions, audio streams, and private chat histories into an un-sandboxed, signed agent with a modifiable debug endpoint, Meta effectively gave commodity malware a zero-effort conduit to bypass platform protections without raising operational alerts.

WIRED reports users are automatically opted in to having their interactions used for AI training, and there is no switch to turn memory off. Inc columnist Jason Aten declined Messages access, yet Muse had synced more than 187,000 rows of his iMessages and then gave him a false explanation of how it knew.

Instinct pinged Resy 'hundreds of times every hour of the day' for one investor, CNN reports, and Resy banned his account.

It's been a busy afternoon for Pip, the AI agent I've been using in Meta's Muse app. The agent, which Meta launched on September 8, booked a Friday date night and made a packing schedule for my upcoming move. It also emailed a friend about a trip next month and made a planning document with places we want to visit.

When I asked for a list of places to go for date night, it pulled up two options in my neighborhood that had been closed for years. … In its list of prompt ideas, Muse said it could haggle on Facebook Marketplace … But when I tried this, Muse said it couldn't message sellers directly. It would instead vet the listings and draft a message for me to send.

While Muse was able to browse Target's website, I still had to enter personal details manually because the retailer blocked autonomous clicks at checkout. (Muse uses Stripe's Link digital wallet service for purchases, which it says generates a one-time use card to prevent your credit card number from being shared.)

My first experience with Muse was on Monday. I needed to book a flight for myself and my wife. I found it myself on Google Flights, then, for journalistic reasons, asked Muse to take a crack at the booking. I put my airline login credentials in Muse's vault, but Muse met a terrifying obstacle, a captcha page, and asked me if I should take over or if it should attempt to pass the test asking if it was a human. Authorized to try, Muse failed; I took over and logged in, then passed the sticks back to the agent. It got kicked out of the browser a few minutes later. The process took 30 minutes and achieved nothing. I then booked the flights myself at a cost of five minutes and just over $500.

Amazon is blocking Muse from shopping on its platform, similar to how it was already fighting another A.I. business, Perplexity, over its shopping tool. … There is no more obvious implementation of a tool like Muse than online shopping, and the king of American e-commerce is just saying no.

It is shockingly easy to spend money when a cute little guy is running around the internet with your credit card.

I'd been using it for all of 40 minutes and already given it my full name, address, credit card number, a photo of my backyard, and the credentials for my home security account login. Going great, I'd say! If AI decides to kill us all I just secured my spot at the front of the line.

It made an attempt to access Amazon and was roundly blocked, complete with a spicy message about AI agents violating terms of service. Anyway, it found what I was looking for between OXO's own website and Walmart.

Marley was most useful here — it set off looking through listings for hauling services and highlighted a few in the area. It asked for my address, phone number, and a photo of the tree limb, and used that information to put together requests for quotes from a couple of companies using online form submissions. I got a text back from one company not long after, and confirmed a time the following day for tree limb removal.

One employee who asked Muse to call an internet and cable provider to negotiate his bill said a transcript showed that the contractor had made a racist reference during the call. Meta apologized to the employee and said the human agent involved would not work on any of its projects again.

Internal tests found that handing calls to human agents could increase the success rate to between 95% and 98%, compared with a much lower rate for calls handled entirely by AI.

Strangely, Muse provided a number but then said to use a different number instead. When I asked why the change, it said it simply made up the first phone number out of thin air and didn't confirm it. Muse then told me not to trust unverified numbers from it.

Muse found four different companies in my area and sent out my request for quotes. About two minutes later, my phone was blowing up with calls from two of the movers. One company offered a quote straight up, which Muse provided to me. The other movers sent over emails.

I even used Muse to help grab a new defense for my fantasy football team. I connected the agent to my Yahoo Fantasy Football account, and it searched for the defenses with the highest projected scores for the coming week. Since I could only claim them on waivers, Muse put in a waiver request for the Kansas City Chiefs and dropped Green Bay. The next day, when the claims were processed, Muse slotted Kansas City into my lineup as my active defense.

The app has hit the top spot on both Apple's App Store and Google's Play Store and, according to market intelligence firm Sensor Tower, reached 560,000 daily active users after just 11 days. The app's early success sent Meta stock soaring 11% on Monday.