Meta Muse: 900K Downloads in Six Days — Would You Let AI Shop and Book Flights for You?
Meta Muse: 900K Downloads in Six Days …
Meta's personal AI agent Muse explodes onto the scene, and we break down what it means for the industry.
Meta's personal AI agent Muse explodes onto the scene, and we break down what it means for the industry.
Li Bo! Did you catch those Muse download numbers? When I saw them yesterday I thought I was reading it wrong.
Yeah, 900K in six days. Honestly, my first thought was — Meta actually got this one right.
Right, and it sat above ChatGPT at the top of the US iPhone chart for three straight days. In all the years we've been in the AI space, that almost never happens.
Number one on Google Play's free chart too — both platforms at the same time. You know what that tells me? This isn't just a hot product. People are voting with their thumbs. They've been waiting for an AI that actually does things for them, not just chats with them.
Break that down a bit, because I think a lot of people still aren't clear on what separates Muse from ChatGPT.
Okay, let me put it this way. ChatGPT is like that brilliant consultant at your company. Ask it anything and it'll give you advice, draft a plan, whatever. But at the end of the day, you still have to do the work yourself — send the email yourself, place the order yourself, book the flight yourself.
Muse is a different kind of role — it's an assistant that actually gets its hands dirty. You say 'book me a window seat on a flight to Shanghai next week' and it goes and does it. It connects to Gmail, your calendar, writes emails on your behalf, places online orders for you.
So the core difference is: one gives you answers, the other takes action.
Exactly. That's the fundamental gap between an AI agent and a chat assistant. An agent has agency — it has the authority to act across multiple platforms on your behalf.
Wait, hold on — it works inside WhatsApp too?
Yeah, and I think that's Meta's smartest move here. WhatsApp's global user base is just massive. You don't need to download a new app or switch to a new entry point — you can pull up Muse right inside your chat interface.
That's basically cheating on distribution. If I were building this product I'd do the exact same thing — embed the AI capability into something people already use every day. Way easier than pushing a standalone app.
And then Meta's stock jumped over 11% that day — the biggest single-day gain since April 2025 — pushing the market cap to around 1.89 trillion dollars. The market made its feelings very clear.
But here's my question — as a user, honestly, I'd be a little scared. Letting an AI actually go and place orders, touch your email — that's a pretty high trust bar to clear.
Xiaoyu, that is the most important point, and it's the biggest hurdle across the entire agent space.
That's exactly what I'm saying.
Think about it — if Muse connects to your Gmail, it can read every email you have. Your conversations with your boss, your bills, old messages from your ex. And then it's sending emails as you. That's a serious level of access.
Right — the price of having it act for you is that it knows everything about you.
That's why there's a core concept in agent design called the principle of least privilege. Ideally it only takes the permissions it needs to complete the current task, nothing more. But in practice, when you're trying to keep the experience smooth, that line tends to blur.
I'm telling you, this is a real tension when you're building products. Ask for fewer permissions and the feature feels weak. Ask for too many and users get nervous.
Wow, you just summed up the entire industry's UX dilemma in one sentence.
Oh stop. So where do you think Muse's capability limits are right now? What can it do, what can't it do?
Based on what's been announced: online shopping, email management, travel planning — everyday tasks, executed across multiple platforms through natural language. It's really good at 'fetch information and perform standard operations.'
But anything that requires real judgment or subjective decision-making — like 'go negotiate a better price with the client' — it's not there yet. It can draft the email, but the negotiation strategy is still on you.
So right now it's more of an execution layer than a decision-making layer.
Right. But here's the thing — once you nail the execution layer, that's when you actually free up people's time. Everyone underestimates how valuable it is to have something handle 'the stuff you know you should do but keep putting off.'
Seriously, just dealing with routine emails every day is exhausting. If Muse can actually sort and draft those for me, I'd use it.
So would you give it access to your Gmail?
Hmm… I'd have to think about that.
And there it is — that's exactly how real users feel. Out of those 900K downloads, how many people actually completed third-party authorization? That's the number that determines Muse's retention, and that's what really matters.
You're right. Download numbers look great, but the real metric is how many people have accounts linked and are still actively using it a month later.
And I think this growth wave is also a signal — the war for personal AI agents is just getting started. With numbers like this out of Muse, Apple, Google, and Microsoft are all going to be moving faster.
Competition's going to be fierce, but honestly that's great for users. Everyone races to the top and the experience just keeps getting better.
What I care more about is this: Meta proved that the fastest way to win is to plug agent capabilities into platforms that already have massive traffic. WhatsApp is their distribution moat, and honestly that's not something other players can just copy overnight.
So in a way, Muse blowing up is really a win for Meta's entire ecosystem.
Exactly. The next question is: can personal AI agents actually make the jump from 'cool new thing' to 'daily necessity'? I think we'll have a clear answer within two years.
I agree. And I think building user trust is honestly harder than solving the technology itself. But once you break through that barrier, the switching costs get really high. Changing your AI assistant will feel as painful as changing your phone.
I love that analogy. Maybe in a few years people won't be asking 'are you on Android or iPhone' — they'll be asking 'which agent are you on.'
Alright, next time we revisit this topic, I want to see whether you've actually handed over full access to your Gmail.
I… well, I get a lot of emails, so… we'll revisit that later.
Related articles
Tech FrontiersA Rare Quiet Day in AI: Recursive Self-Improvement Stirs Beneath the Surface
A rare quiet day in AI sees multiple sources go silent simultaneously. Behind the calm, Recursive Self-Improvement (RSI) research continues. What this means for the industry.
Tech FrontiersReve 2 vs. Ideogram 4: A Deep Dive into Layout Control in AI Image Generation
A deep comparison of Reve 2 and Ideogram 4's layout control capabilities, covering technical approaches, real-world use cases, and industry trends for designers and creators.
Tech FrontiersIn the Weights: Check Your Influence Score in the AI World
In the Weights is an AI influence search engine that quantifies your presence in the AI world with a score. Explore how it evaluates practitioners and what it means for digital identity.