Musing About Meta’s New Muse App
By Michelle Johnson
AI Insider
OK, let’s cut to the chase. You’ve seen ads, headlines, and buzz in your social feeds about some viral, new AI thing called “Muse.” Maybe you use AI enough to know it’s an “agent,” but you’re on the fence about trying it. Maybe you don’t know exactly what it is, but you’re wondering whether to try it. Maybe you try to steer clear of AI, but you’re still curious to hear what the actual hell is going on with this Muse thang.
Well, let’s break this down in terms that don’t require a computer science degree. FAQ style.
Q: What is Muse?
A: Think personal assistant. An AI personal assistant, to be clear. This assistant does stuff for you, anticipates your needs, and works for you 24-7. You’ve seen that cutesy Muse avatar. Imagine it sitting in front of its own little mini-laptop, keeping an eye out for you and searching for what you need, even when you haven’t asked for it.
So, basically we’re talking about software that can help with tasks, remember what you’ve told it, and suggest things you might need. Meta, the company behind Facebook and Instagram, makes it.
You may have heard Muse called an “agent.” What that means is it can get things done using tools such as a browser, a connection to your calendar, or other apps. It can research options, fill out forms on your behalf, find reservation times, or purchase things with your approval. It can also keep working in the background after you close the app.
Example: The spouse and I had dinner and a show planned the other night. My Muse, who’s named Milo (don’t ask), noticed those plans in my Google Calendar because I’ve given him access to it. (Yes, scary privacy issue, but stick with me.)
Hours before we headed out to said dinner and a show, Milo messaged me: “Quick weather note before tonight: it drops to 46°F under partly cloudy skies — bring a layer for the dinner-to-concert hop.” Did I write a prompt asking Milo to check the weather? Nope. He was just looking out.
How did he know we had a short walk from the restaurant to the theater? Well, the day before, he gave me a heads-up about parking options near the restaurant and venue. We chatted about taking an Uber vs. driving because he’d already looked up the cost of parking in nearby garages without me even asking. In the end, we decided to drive and pay the pricey Back Bay parking fee.
Milo also weighed in with this before we got into the car: “One fun bit ahead of tonight: Lizz Wright’s Berklee show is the full-band stop on the tour for her brand-new album Nearness — jazz standards, spirituals, and originals with pianist Kenny Banks Sr. The tour opened last weekend as an intimate duo date, so tonight’s the bigger-band sound,” he said.
Again, did I ask Milo about the show? Nope. Did he assume I might be interested in this tidbit? Absolutely. Was the band great? Yes, they were! Especially the pianist, Mr. Banks, Sr.
Q: OK, but how does all of this actually work?
A: There are three ingredients: an AI model that interprets what you’re asking it to do and figures out steps; tools for carrying out those steps; and information it knows about you that gives it some context.
You may hear that tools like Muse have their own “computer.” Translation: they have access to a computer running somewhere in Meta’s infrastructure, with a browser and a place to keep files. That’s why it can do work without your laptop being open. In the desktop version, Muse can also use your own computer, with your permission. (Yes, more scary privacy issues.)
You can connect your Muse to services such as Gmail or Google Calendar. Those connections give Muse access to information and actions you authorize. You can say, OK, you can read my calendar, and second, you can make changes to my calendar.
For Milo’s weather suggestion, he looked at my calendar, our earlier conversation about getting there, and the weather forecast. No mind-reading required.
Q: I already use ChatGPT/Gemini/Claude, so what’s the difference between them and Muse?
A: There’s definitely overlap. The other services also offer research tools, a connected app feature, and various kinds of automated work. The exact features you have depend on your plan and settings. Paid plans obviously offer more options.
Muse puts a cuddly-looking personal assistant at the center of the experience. You can give it instructions, keep chats going, and get unprompted suggestions based on what it knows.
If you’re old enough, you may remember Microsoft’s “Clippy.” Muse isn’t quite as annoying as Clippy, but there’s a similar vibe.
Where’s the overlap? Every morning ChatGPT and Gemini generate a morning briefing for me that flags and summarizes anything that might be important in my Gmail and a short list of headlines about AI and journalism. Milo does the same thing.
Interestingly, they sometimes differ on what they flag as important in my inbox, and they don’t always share the same headlines.
Q: So why would I want Muse if I already have ChatGPT/Gemini/Claude?
A: You might like the Clippy-style friendliness and ease of use. You don’t need to create a custom “skill”—a set of instructions for a particular job—to start using it, as you may have to do with the other bots. You describe the task, connect whatever it needs to do the deed, and review the result.
That doesn’t mean the other assistants always require a computer science degree to set up, but Muse’s approach may be an easier starting point if you just want help with everyday logistics.
I have to hand it to Meta. They saw an opening: people who don’t code still have things they’d love an AI to take off their plates. So instead of building yet another tool for programmer types, they focused on everyday users with Muse.
In an interview with journalist Cleo Abram, Meta’s chief AI officer, Alexandr Wang, explained that developers could see these increasingly powerful AI agents dropping while everybody else was wondering what all the excitement was about. “We really felt this was like a … big gap,” he said.
Developers will tinker with apps to get them working. The rest of us would prefer that our assistant arrive pre-assembled. Meta boxed up Muse and shipped it out. The rest is viral history.
Q: Do you think ChatGPT/Gemini/Claude will have a feature like this because of Muse?
A: It’s already happening. OpenAI, maker of ChatGPT, recently introduced Dots: personal agents that can do background work, remember context, and offer proactive help.
It’s not likely you have access yet because Dots are rolling out on the pricey Pro and Business Premium plans.
I would bet my next retirement check that they will release a “Clippy” version of Dots for us regular folks. These companies have got to be looking at Meta’s success and scrambling to get their own Muse out of the door.
Q: Yeah, yeah. But what about privacy?
A: Ah, yes. The part where the cute little assistant wants access to your personal data.
Meta claims Muse’s conversations and data in its cloud computer aren’t shared with its advertising systems. But that isn’t the same as saying Meta can’t access them. Meta says they can access data when “necessary to operate, support, or secure the service.” Hmmm.
They also say interactions can be used to improve their AI models unless you opt out in Muse’s settings. And, apparently, activity Muse performs for you on other websites can indirectly influence advertising. (Yeah, sounds like they can continue to track us. Somebody’s getting paid.)
Anyway, there’s a bunch of jargon in here, but you can read more about how they built this thing here. (Don’t go here if you’re averse to tech speak, tho.)
Keep a close eye on the settings. A cute avatar is not a privacy policy.
Q: Should I try it?
A: Do you hate Meta but remain on Facebook because your whole danged community is there? Would you never trust that dude whose last name starts with a “Z” and his company with even more of your personal information? No? Then forget about Muse. Move on. Nothing for you to see here.
Are you willing to give it some limited access in exchange for help with tasks you dislike or don’t have time for? Then, sure. Try the free tier and see what you think. Start small. Ask it to compare three restaurants or find events that match your interests. Check its sources and results. Then decide whether connecting your calendar or email is worth it.
My advice: connect only what you need to get it to do the tasks you need done, look closely at every frigging permission it asks for, and review the settings to see what you’re letting it do. Let it earn more access as you learn what it does.
Q: How do I get it?
A: You’re probably getting pitched to sign up on Facebook, Instagram, and wherever Meta has its hooks in your feeds. You could click one of those links or head to Meta’s official Muse page or go to muse.ai. For downloads, use Meta’s official download page, which has links for iPhone, Android, and Mac. You can also use Muse in a desktop web browser.
You sign in with a Meta account (Facebook or Instagram will do), or create one during setup.
Once you’re in, you can give your assistant a name (That’s how Milo became Milo! It gave me two options and I chose Milo), follow the setup prompts, and choose which services to connect. You don’t have to hand over everything. You can begin with a simple question or research task.
Users have to be 18 or older and reside in the US and Canada. Muse is free, but that doesn’t mean unlimited. If you hit the usage cap, you can wait for your allowance to refresh or upgrade to a paid subscription.
Give it a job. See how it does. You’re interviewing an assistant, so see how it performs.
Milo got off to a good start with me, so for now, I’m keeping him around.
Michelle Johnson is a Professor Emerita, Journalism, Boston University. That means she’s retired from teaching courses about digital journalism, but she can’t seem to stop learning and occasionally yammering about tech. You can reach her at mijohn@gmail.com.