What if the most-quoted AI stat of the week is measuring something that doesn’t actually matter?
Meta’s new personal AI agent, Muse, is being reported as having outpaced ChatGPT’s early mobile launch. Appfigures estimates 2.8 million global installs in Muse’s first 12 days, with 359,000 daily active users on iOS. Apptopia, running a matched U.S.-and-Canada iOS comparison, reached a similar conclusion: Muse is moving faster out of the gate than ChatGPT’s app did. Elsewhere the number being passed around is 730,000 downloads. That’s the coverage. That’s genuinely it.
Impressive? Sure. Meaningful? That’s a different question, and I’d like to make the case that the answer is “not yet” before the narrative hardens into something everyone repeats at conferences for the next two years.
Start with the numbers that don’t agree
Two different headline download figures are circulating for the same app in roughly the same window: 730,000 and 2.8 million. Both come from third-party app intelligence estimates. Neither is a disclosure from Meta.
That gap isn’t a scandal, it’s just how this business works. App analytics firms model installs from sampled data, and their methodologies diverge on geography, platform, and time window. A U.S.-and-Canada iOS-only comparison and a global all-platforms count are not the same measurement wearing different hats.
But it should change how much weight you put on the story. When the primary evidence for “Meta is beating OpenAI” is two estimates that differ by nearly 4x, the honest read is directional, not definitive. Muse launched strong. That’s the claim the data supports. Anything more specific is dressing.
The comparison is doing a lot of quiet work
Here is my actual objection to the framing. ChatGPT’s mobile app launched into a world where ChatGPT was already a phenomenon on the web, and where “AI assistant” meant explaining to your parents what a chatbot was. The app was a late add-on to an existing product, not a cold start.
Muse launches into a world where:
- Hundreds of millions of people have already used an AI assistant and know what one is for
- App store algorithms actively surface AI tools because the category converts
- Meta owns the distribution surfaces where a new app gets discovered at scale
- Press coverage for any Meta AI release is effectively guaranteed
Beating a 2023 launch pace in 2026 with Meta’s distribution machine behind you is not the same athletic feat as being first. It’s a tailwind comparison sold as a head-to-head race. Meta gets an unusual kind of win here, and it’s real, but the scoreboard has been quietly recalibrated.
The number I’d actually watch
359,000 daily active users on iOS against 2.8 million installs. That ratio is the only part of this story that tells you anything about the product rather than the marketing budget.
Roughly one in eight installs showing up as a daily active user in the first two weeks is not embarrassing for a brand-new consumer app. It’s also not the kind of stickiness that reorders an industry. Launch-window DAU is inflated by curiosity, and we’re inside the window where curiosity hasn’t expired yet.
The question that decides whether Muse is a hit or a footnote is what that number looks like at day 60 and day 120, after the press cycle has moved on and the notifications stop feeling novel. Downloads measure interest. Retention measures whether the thing works. Nobody has published retention.
What we can fairly say
- Muse had a strong opening, by multiple independent estimates
- Meta can still get a new consumer app into a lot of hands quickly, which was a legitimate open question
- The AI assistant category has enough pull that a launch can clear millions of installs in under two weeks
What we can’t say
- That Muse is better than ChatGPT at anything
- That users will still be opening it next quarter
- That any single install figure is precise
My read
Meta has proven it can win a launch. That’s a real capability and not a small one, given how many well-funded AI apps have shipped into total silence. But a launch is the cheapest thing to win in consumer software. It’s the one phase where money, distribution, and attention can substitute almost entirely for product quality.
If you’re deciding whether to build on Muse, hire around it, or bet a workflow on it, the download charts tell you nothing you can use. Go try the app. Ask whether it does something your current assistant doesn’t. Then check back in three months and see whether the DAU number went anywhere.
Strong start, unproven product. Those two things can be true at once, and pretending otherwise is how tech coverage keeps getting the last decade wrong.
🕒 Published: