🧵 Breaking down the numbers｜Muse surpasses 10 million users
1/ First, a note on the scope: this isn’t a breakdown of a single post. It’s a breakdown of the data from Meta’s Q3 earnings call yesterday (10-21). There hasn’t been a truly viral single post in AI tech in the past 24 hours, so rather than force a breakdown, let’s break down the numbers.
Zuckerberg revealed that Muse, which launched in early September and is currently available in the US and Canada, has over 10 million users. About 25% of those who signed up now use it every day, and more than a quarter of active users use it to shop, with an average order value above $200.
2/ Here comes the third-party cross-check: Sensor Tower’s independent data shows that Muse had 10.9 million global downloads in its first 40 days (10.2 million in the US). Its 7-day retention rate is 30%, and its 14-day rate is 23%—both higher than the averages for ChatGPT/Claude/Grok, at 19%/16%. On 10-16, daily active users reached 2.5 million.
The official figures and independent third-party data line up: these results are real.
3/ Why is shopping the first use case to take off? Let’s break down the three gears:
① Payment rails: Plaid says activity on its connector is “unprecedented”—money has to be able to move for agents to be useful.
② Trust mechanisms: Sensitive actions require human approval, with isolation via Secure VM—the first hurdle is getting people to “trust” AI with their money.
③ High order value: An average above $200 means the agent is handling decisions “worth entrusting to it,” not impulse purchases for $9.99.
The formula: payment rails × trust mechanisms × high-value use cases = the first piece of the puzzle for bringing agents into the real world.
A dose of caution: the stock’s +5% after-hours move reflects market sentiment, not a verdict on the product. Of Muse’s 10 million users, how many are just trying it out, and how many will stick around? Sensor Tower’s 23% two-week retention rate is a more honest yardstick.
Zuckerberg also announced that the company is entering a “new phase”: monetizing through commissions on transactions—the AI assistant may finally have found a way to survive without subscriptions.
The next question: after shopping, what will be the second use case AI agents crack?
1/ First, a note on the scope: this isn’t a breakdown of a single post. It’s a breakdown of the data from Meta’s Q3 earnings call yesterday (10-21). There hasn’t been a truly viral single post in AI tech in the past 24 hours, so rather than force a breakdown, let’s break down the numbers.
Zuckerberg revealed that Muse, which launched in early September and is currently available in the US and Canada, has over 10 million users. About 25% of those who signed up now use it every day, and more than a quarter of active users use it to shop, with an average order value above $200.
2/ Here comes the third-party cross-check: Sensor Tower’s independent data shows that Muse had 10.9 million global downloads in its first 40 days (10.2 million in the US). Its 7-day retention rate is 30%, and its 14-day rate is 23%—both higher than the averages for ChatGPT/Claude/Grok, at 19%/16%. On 10-16, daily active users reached 2.5 million.
The official figures and independent third-party data line up: these results are real.
3/ Why is shopping the first use case to take off? Let’s break down the three gears:
① Payment rails: Plaid says activity on its connector is “unprecedented”—money has to be able to move for agents to be useful.
② Trust mechanisms: Sensitive actions require human approval, with isolation via Secure VM—the first hurdle is getting people to “trust” AI with their money.
③ High order value: An average above $200 means the agent is handling decisions “worth entrusting to it,” not impulse purchases for $9.99.
The formula: payment rails × trust mechanisms × high-value use cases = the first piece of the puzzle for bringing agents into the real world.
A dose of caution: the stock’s +5% after-hours move reflects market sentiment, not a verdict on the product. Of Muse’s 10 million users, how many are just trying it out, and how many will stick around? Sensor Tower’s 23% two-week retention rate is a more honest yardstick.
Zuckerberg also announced that the company is entering a “new phase”: monetizing through commissions on transactions—the AI assistant may finally have found a way to survive without subscriptions.
The next question: after shopping, what will be the second use case AI agents crack?