Nate B Jones
I Gave Meta’s Muse The Most Boring Job I Had. It Found $5,350 A Year.
Ask the tutor about this video
Sign in with a verified email and ask the tutor anything about this video.
TL;DR
Research summary
▶ Muse tackles everyday money tasks
The speaker opens with common administrative burdens: canceling subscriptions, questioning bills, and pursuing refunds. They say Muse found $5,350 per year in their subscription spending and had canceled $1,285 in future charges so far. The speaker frames Amazon’s blocking of the assistant from its store as a dispute with stated technical and access concerns, but argues that the larger issue is who controls consumer attention and purchasing relationships. They say Meta shipped Muse on September 8 and that within two weeks it became the number-one iPhone app in America, claiming it was being adopted faster than ChatGPT had been at the same stage. The video will examine the assistant’s usefulness and why a major retailer is paying attention.
▶ Reported savings and coordination help
The speaker recounts publicly reported early user stories: Chris Abraham said Muse found and canceled a forgotten family book subscription, checked its refund policy, and recovered a year of payments, without giving a dollar amount. Brad reportedly spent 98 minutes on the phone with AT&T while also contacting Verizon and T-Mobile to compare offers; the account says Muse spared him more than three hours of calls. The speaker notes Reuters reported that Meta tested human contractors making some calls for employees, then rolled back the experiment over disclosure and privacy concerns. They say it is unclear when another person might receive information and act on a user’s behalf, while Muse appears to present calls as autonomous. Other reported examples include thousands of dollars in annual car-insurance savings for Joe Devoy, and $1,156 in annual insurance savings for Jason Longo while he was at the gym. A parent reportedly used Muse to combine schedules for four children in four sports across four messaging apps into one daily notification. The speaker argues these examples show value in reducing administrative work, not just saving money, and says that kind of practical help can spread through word of mouth.
▶ The app’s approachable design and access
The speaker describes Muse as a main ongoing conversation with a nameable companion, with optional side chats and goals; they named theirs Marvin. They praise the familiar avatar, conversational continuity, and visual status cues that show when the assistant is searching, working, or has completed an action. They say onboarding initially requires a phone number, with accounts such as Gmail or a calendar connected conversationally when relevant and at the user’s choice. According to the speaker, Muse’s instructions say it does not fish for personal information and keeps track of what users choose to share. They call the combination of these design choices a product masterclass in making AI approachable to people who may not otherwise care about AI. Ordinary use is free, Meta offers paid subscriptions for people who need more, and the speaker says Mark Zuckerberg described a substantial free allowance at launch. The speaker reports not encountering usage limits, says international rollout is limited, and expects expansion as regulatory issues are addressed.
▶ Assistants as a new shopping starting point
The speaker argues that useful work in an assistant can build a habit: after Muse helps with spending, a person might return to it for an appliance, hotel, or phone plan because it already knows some preferences. In this scenario, sellers may enter the process after the user and assistant have begun working together, which the speaker says could threaten established consumer relationships. They describe Amazon’s sponsored search and product-page placements as a valuable way to reach people at the moment of purchase, citing $68.6 billion in Amazon ad revenue in 2025. They say advertising profit is not separately disclosed, but characterize advertising as highly profitable and important to sustaining Amazon’s retail business. The speaker imagines Muse comparing appliances against a user’s space, budget, and delivery needs, then recommending an option whose selection was made outside the retailer’s site. Amazon’s stated objections, as recounted by the speaker, are that Muse did not properly identify itself, accessed the store without agreement, and appeared to capture and store customer credentials. The speaker considers those legitimate stated objections, but interprets the broader commercial conflict as a fight over who shapes the purchase journey and supports Amazon’s advertising business.
▶ Commerce partnerships and unproven results
The speaker says Mark Zuckerberg expects Meta potentially to earn a small fee on transactions, possibly paid by participating businesses, allowing a consumer product to remain free if it brings businesses customers. They say Shopify has made Muse an AI channel for merchants and Stripe has equipped Muse with link payments, including a virtual card limited to approved purchases. The speaker says they set it up and found it seamless when planning date night. They describe business integrations as potentially reducing the work users must finish elsewhere and creating network effects as more businesses connect. The speaker compares Meta’s possible transaction fees with advertising revenue captured by Amazon and Google, arguing that the valuable asset is the relationship with the person deciding where to spend. Walmart, they say, may welcome another way for customers to find products because it competes with Amazon; Walmart also has an advertising business, reported here as $6.4 billion in the last fiscal year, roughly one-tenth Amazon’s ad revenue. The speaker cites Wired’s report that Walmart said checkout inside ChatGPT converted at about one-third the rate of product flows requiring a clickout, so Walmart and OpenAI replaced that flow, while ChatGPT brought new customers at roughly twice the rate of search engines. They caution that those results may not predict Muse’s performance and say the partnerships remain early: evidence that partners bring customers who complete purchases is still needed. The speaker says they are optimistic but that the model has not been proved out.
▶ Subscription inertia and the value of personal context
The speaker argues that Muse could pressure subscription businesses that benefit when customers stop getting value but continue paying. They cite a study using payment-card data from 2017 to 2021 across 10 subscription services: replacement cards required people to actively renew, and cancellations rose. They say the study, published in the American Economic Review in 2025, estimated that inattention to subscriptions increased revenue from a subscriber pool by 87% on average, with increases from 14% to more than 200% depending on the service. In the speaker’s view, easy-to-use assistants could help people actively review and cancel charges they had lacked time to address. They argue that model intelligence alone will not determine which assistant people rely on: it must know relevant circumstances, distinguish a wasteful charge from one that supports an occasional need, ask when uncertain, and use what the person has shared. The speaker says Muse built heavily on an OpenClaw foundation and, in a public conversation on X, OpenClaw inventor Peter Steinberger and Muse contributor Nat Friedman discussed Muse borrowing some of OpenClaw’s successful Markdown memory-file patterns while building them in its own way.
▶ Meta’s product advantages and team
The speaker says Meta has potential advantages from billions of existing users, experience building consumer products and relationships, and the ability to connect outside services containing useful personal information. They stress that Muse does not automatically inherit everything Meta knows: access, user permission, and use of information matter, and they say more transparency is needed. In their view, the product should earn more context by being useful and should earn trust through execution; app-store ranking matters only if people keep using it. They credit a team assembled by Mark Zuckerberg with turning Meta’s existing resources into a coherent personal assistant. The speaker recounts a $14.3 billion deal for 49% of Scale and Alexander Wang’s move to lead Meta’s AI efforts, citing a reported pay package somewhere around $5 billion and saying Meta’s stock rose $200 billion after Muse launched. They say the leadership decision was widely criticized, Yann LeCun later left, and Wang’s team brought in Nat Friedman, whose background includes leading GitHub and building developer products, as well as Sheng Xiao for model-research leadership. The speaker’s assessment is that the technical work had to disappear behind a simple avatar and product experience that nontechnical users would adopt.
▶ Platform ambitions and new devices
The speaker interprets Meta’s hires as an attempt to improve execution and deliver a model, product, and service that work together. They contrast the current app-based phone experience, which involves moving between apps, with a future in which an assistant can coordinate tasks across them. They note that Apple controls much of the iPhone operating-system layer and recount that Apple’s 2021 App Tracking Transparency changes constrained Meta’s advertising business, illustrating Meta’s dependence on another company’s platform rules. They argue that Meta could gain influence over how requests are carried out across a person’s life, while ChatGPT and Google also have distribution, tools, or useful context. In their view, competition will involve memory, working connections, and interruption management as well as capable models. They describe Muse Charm as an announced small-screen device for talking to the assistant and interacting with its character, due in December, and say Meta also announced plans to bring Muse to its new AI glasses. The speaker sees these devices as ways to connect with Meta software without always starting on another company’s interface.
▶ The speaker’s closing case for practical AI
The speaker concludes that Muse matters because people can find solutions without needing to care about AI itself. They credit the team’s execution and the avatar’s approachable design with making the help feel tangible, while saying the user, rather than AI, is the hero of the experience. The speaker returns to their own reported savings of about $1,200 over the next year and says other user stories inspired them to try Muse. They acknowledge that some people may not trust Meta with their data and say that choice belongs to each person; they also say the evidence so far suggests many people are choosing to trust Muse. The speaker’s final claim is that the wider industry significance is the development of AI that is helpful to people. They invite viewers to share what they saved, what they do with Muse, or why they do not use it.
Full transcript
[0:32] GOING TO SPEND IN the future by default on stuff I don't use. So, yeah, it's kind of good to have it back. And this is what I didn't expect that the boring job that I just described for you is exactly why Amazon locked it out of the store. Amazon has its own stated reasons, and I'm going to get into them, but I don't think that one of the biggest companies in the world blocks an AI assistant over paperwork. I think it blocks an AI assistant that changes where you spend your attention. Meta Ship Muse on September 8th. Inside two
[1:03] weeks, it has become the number one iPhone app in America. It is being adopted much faster than Chat GBT at this stage. So, this video is about two things. Number one, what is it actually like to hand a real life job that saves money to Muse and have it save me the time and the money that it would otherwise take to do that job? And then number two, why does doing that small personal job have one of the largest retailers in the world paying attention? But first, let's get into Muse. Look at the jobs people are already handing over. Uh Chris Abraham said Muse found a
[1:35] book subscription his family had forgotten to cancel and cancelled it, checked the refund policy, and recovered a year's worth of payments. He doesn't give a dollar figure for that, but it's going to be sizable. The interesting part is that Muse went ahead and pursued and went after that refund. How often do you cancel something and decide arguing about the old charges you didn't intend would take more effort than it was worth? Another early news user named Brad, these are all reports that I found around the internet, right? They're publicly available. He reports that news spent 98 minutes on the phone with AT&T
[2:07] through multiple transfers while also contacting Verizon and T-Mobile to compare phone offers. She says it spared him more than 3 hours of calls. Getting those hours back is something that you can't put a price on. Imagine multiplying that across 100 million people. Now, these phone call stories are early. Reuters has reported that Meta tested having human contractors handle some Muse calls for employees and the company rolled the experiment back after concerns about disclosure and privacy. And so one of the things that
[2:38] we need to be aware of is that Meta doesn't tell you clearly when another person will receive information and act for you. At the moment it seems like Muse is saying it handles the calls autonomously, but there's not as much transparency around that as as I would like to see. But I view that as an early teething issue. We are going to get to the phone call piece. Meta may have already rolled out that automated phone call piece well enough in many cases and it just hasn't told that story as clearly as we'd all like to see. And the larger story is around the work saved
[3:08] for for so many of us. The bills, the subscriptions, these problems have existed long before AI models came around, right? How long have you looked at paperwork and said, "I can't do all of this." Or looked at charges and said, "I don't have time to investigate this." Or maybe a health bill you can't trust. Or haven't had time to get on the phone and call. And so when you see value there, you can see how that spreads virally and that's part of why Muse is the number one app in the Apple download store. So another couple of stories. I love these stories because they show
[3:39] concrete value for AI in a way that so many of the stories out of the valley don't. And again, these are all publicly reported. So Joe Devoy reported thousands in annual car insurance savings cuz he worked with Muse. Jason Longo saw an insurance post and tried a similar job while he was at the gym and reported $1,156 a year in savings on insurance. So, he was exercising while Muse sorted that out. What we can see in these stories that's really fun is that people are realizing that this is possible. That Muse can save you money. Phone call or not, Muse can save you money. And they're just sharing it
[4:10] around and Muse is going around and helping each person sort of figure out what works for them with their data to the degree that they're comfortable sharing it. And that's really exciting because it becomes this sort of viral sharing moment of, "Hey, I got Muse to help me save. Hey, I got Muse to help me save." And I love that because we all could use some dollars back in the pocket. And it extends beyond money, too. A parent reports using Muse to combine schedules for four different kids in four sports across four messaging apps in one daily notification. Look, I I have I have
[4:41] kids. Dealing with that kind of coordination, that speaks to me. Knowing where the information lives is not enough. Someone has to go through and collect it and reconcile it and remember what matters today. That was one of the early uses for Open Claw. And here we see something that is sort of packaged virally that sits there with a nice little avatar that just sorts it out for you. And this kind of load, we call this adulting. We call this parenting. This is work many of us have simply accepted as a part of life. uh like you you may be the person in your life who's connecting the calendar to the email to
[5:11] the website to the phone call for the family and the possibility of actually having real useful help there. That's way way better than having a frontier model. It's actual value for you. But when you have to carry that whole thing in your head, it's a lot of responsibility. Muse is built for things like that. It's the kind of ordinary benefit that I think the AI industry has been remarkably bad at explaining. Just to be very honest, people hear about data centers. We hear about the next intelligence breakthrough. We hear about
[5:42] arguments about what happens to jobs. And those concerns are real. But someone looking after a family needs an answer to what this tech will do for them now and whether it's useful. And so getting real useful help with their priorities gives them a reason to participate, gives us a reason to participate. We all have families. We get to decide what is important enough to ask for help with. And Muse gives us a tool that is genuinely helpful and accessible to everyone, whether you're technical or
[6:12] not. I've made a lot of this YouTube channel around how you learn to use these tools because they're not self-explanatory. Muse is a lot more self-explanatory. You get one main conversation with a companion you can name. Uh I named mine Marvin because I like Douglas Adams. You can open side chats and set goals and give it more information, but you can also just come back and ask the next question. You can switch topics. You don't have to begin again by deciding where this part of your life belongs in a collection of projects and threads. And the avatar, it gives you a sense
[6:42] that you're having a relationship with a familiar face, right? And I think people people want that. We remember our Tamagotchis from the '90s. something people feel comfortable talking to, something that remembers us enough to be useful, that is there when we need help. And so, a company logo in a new conversation doesn't create that feeling. But a character, a cute fuzzy bear kind of character, that makes it easier to imagine coming back and asking for help. Another detail I noticed is the way status is communicated to
[7:12] non-technical audiences. When I looked at how Marvin was searching through sources or checking sites, it just gives you a little emoji and it gives you a little phrase for what Marvin is doing and then Marvin visually changes. Marvin's a 3D avatar and visually starts typing away at the laptop and you know that Marvin is working. is just super intuitive and I think it gives you a way to immediately without reading without thinking distinguish a proposed action from a completed action an inrogress
[7:43] action from a completed action. And so it was really really simple for me to see when I was kind of working through the subscription cancellation process when was Marvin working and what was Marvin doing. A good assistant needs to make those stages clear without requiring you to even think about the fact that you know that like it should just disappear. Onboarding follows the same logic of simplicity. With Muse, all you do initially is get a phone number and then everything branches out from there. It will connect an account like a
[8:14] Gmail or a calendar when it makes sense conversationally as a choice you can make and you can tell it about yourself gradually. Uh, in fact, when I asked Marvin, how do you know stuff about me? How Marvin was really clear to say, and this was coming straight from the project instructions given by Muse, given by the Muse team, it doesn't fish for information. It will let me share and it will keep track of what I choose to share, but it doesn't want to go fishing. And I think that's also something that builds trust when Meta is
[8:44] a company that has a ton of your data. Anyway, now these details are all things you can copy, right? You can copy an avatar. You can copy a status pill. You can copy continuous conversation. I still think even in that world, Muse is a product masterclass in making useful capability for AI approachable. The individual choices don't have to be unprecedented. They have to work together coherently and well enough that someone who has never cared about AI can
[9:14] use it to get help. And the reason I started this video with all these individual stories is because that is what is happening with Muse. That is why Muse is special. And ordinary use is free, which I think is how consumer AI is going to go. Meta offers paid subscriptions for people who need more. But the main stated plan is to make Muse free for most of what people want to do. Mark Zuckerberg described a substantial free allowance when it launched. You don't need to understand or think about token count to understand the decisions. I haven't run into issues with the usage. Meta wants a very large number of
[9:46] people to not only try this but use it regularly without facing costs and that changes who the product can reach and there's limited international roll out but there's very very aggressive interest and I expect Muse to continue to roll out globally as they can kind of get the regulatory pieces figured out and feel comfortable launching. What happens next? Let's say it's the number one app store. Where do we go from here? How do you build an app that people find useful over time? I started this video
[10:17] by talking about the fact that, you know, Muse helped me save $1,200 over the next year. How do I then trust Muse with more stuff? Look at the habit that you form when you work with AI. Let's say you've asked your AI assistant to help sort through your spending. Later you need a new appliance. Later you need to book a hotel. Later you need a better phone plan. Do you go back to the AI avatar that already understands some of your preferences and has done really useful work for you or do you go
[10:47] somewhere else? The world we're living in is a world where the company that is selling you the product, the company selling the hotel, the company selling the flight, the company selling the appliance, they may only enter the conversation after you and the agent have already started to work together. And I think that is why so many companies including Walmart have lined up, including Shopify, have lined up to do deals with Muse early because they want to make sure that Muse doesn't disintermediate their consumer relationships. That's how a platform
[11:19] develops. Businesses have to think about how to serve you through the place you already go for help. Amazon and Amazon's response to Muse makes the stakes here really, really clear. And I used to work at Amazon. This was very very obvious what they were going to do. When you search for a product on Amazon, the Amazon has an opportunity to influence what you buy, right? Sellers pay for sponsored placements in search results and on product pages and they are paying to reach you while you are on that page thinking about a purchase. Owning that
[11:49] moment is a tremendously valuable business for Amazon. Amazon in fact reported $68.6 billion in ad revenue in 2025. A lot of people don't know that advertising profit isn't disclosed separately, so we can't put an exact profit number there, but ad businesses are typically extremely profitable. And the ad business is enormous. And the ad business as a whole tends to drive more bottomline value than the retail business for Amazon. And a lot of people
[12:20] don't know that either. They think Amazon makes a ton of money off of retail. They don't. Generally, advertising matters a lot to keep retail sustainable. Now imagine asking Muse to find an appliance that fits your space and your budget and has a delivery date and it compares options and brings you a recommendation and if you accept the retailer gets a purchase from someone whose decision has been made elsewhere and the opportunity to sell visibility to that customer has has moved. Are you selling to the agent now? And this is why Amazon's decision is a platform
[12:52] fight and why I wasn't surprised at all coming from Amazon. Amazon says that Muse didn't properly identify itself and accessed its store without agreement and appeared to capture and store customer credentials. Now, those are stated objections and that's fair, but the larger commercial reason is not something they're going to say publicly because ultimately the conversation is not about Meta having credentials that go into protected storage, although it does. um it is about who owns the
[13:23] customer relationships. It's not about a dispute over technical rules. So my read is that Amazon is focusing on protecting its ability to shape the purchase journey and it wants to make sure that that ad business is supported and safe. Now Mark Zuckerberg has described how Meta might earn money here. Over time, Mark expects to take a small fee on transactions potentially paid by the businesses involved. And that gives us a very clear explanation of how a consumer product stays free over time. If Muse
[13:54] becomes a source of customers, businesses have a reason to pay for the commerce that Muse helps create. And some of the infrastructure for that is already taking shape. So, Shopify added Meta. As I mentioned, it's an AI channel for its merchants. Stripe has equipped Muse with link payments, including a virtual card that is limited to approved purchases. Uh, I've already set it up for you. Those connections can make it easier for a useful suggestion to become a order that just gets done without you. I It was very seamless when I used it to
[14:24] plan date night. For businesses, this makes it possible to serve a customer that started and ended the journey without ever touching the site. and those platform connections start to add up. If your assistant can deal with more and more and more and more of the businesses that you work with as a consumer or as a business owner, there's less work left for you to finish somewhere else. And so each business connecting into Muse makes it easier for another customer to try Muse. That's how distribution and execution start to build on each other with network effects. Very long term, I see this as a
[14:56] run at some of the economic value that Amazon and Google capture through attention. So, Amazon captures through advertising, Google captures through advertising. Meta can get paid through transaction fees when it helps to influence purchases, which is also the business Google and Amazon are in. And so, the valuable relationship is the person choosing where to spend and owning that relationship. And that's something that Mark Zuckerberg sees an opportunity for. Now, it's instructive to compare Amazon and Walmart here. Walmart has a different calculation.
[15:26] Meta named it among the retailers joining Musea Connect. Walmart competes with Amazon online, of course, and has a re reason to welcome a new way for customers to find products. If an assistant brings in someone who would otherwise have started on Amazon, that can be a really useful opportunity for them. Walmart does have an advertising business as well, uh, but it's smaller, right? It's $6.4 billion in the last fiscal year. So, it has something to protect, but it's about 10 times smaller in terms of ad revenue than Amazon at this point. So I think the difference is
[15:58] in competitive position right a new shopping starting point can look attractive to a challenger in a space and threatening to the company namely Amazon that is defending an existing shopping habit. [snorts] Other retailers and Shopify merchants have their own reasons to participate and we can get into that but ultimately it's about connecting with what they perceive as a winner which in turn makes it more likely to be a winner in the space. Now there's a really useful wrinkle here. Walmart has told Wired that checkout inside chat GPT did not convert well. Namely, that it converted at about a
[16:29] third of the rate for products requiring a clickout as they would expect it to. And Walmart and OpenAI decided ultimately to replace that flow. Yet, Chad GPT brought new customers at roughly twice the rate of search engines. So what we're seeing here from Chad GPT may or may not be what we see from Muse. But it's worth noting that there are brands that have already gone into this space with AI and are already bringing learnings when they come and talk to Muse. And I think ultimately that's to Muse's advantage because it
[17:00] means that you have that experience in the AI space and you're going to bring that learning to bear when you make a new institutional partnership like this one. Over that same period, I would look for evidence from Muse's retail partners that the assistant brings in customers who complete purchases in return. Right? If you are going to say that Muse is going to be helpful to you, then you would want to look for not just partnership announcements, but that teams like Walmart are able to leverage learnings from previous AI partnerships and really bring revenue and customers
[17:30] to the table with this one. I'm optimistic, but it's early days, right? We haven't proved it out yet. And as much as we've talked about the consumer impact on shopping, we also should talk about Muse's ability to put pressure on businesses that depend on you leaving things alone. Subscription companies, like phone companies, have always had customers who stopped getting value before they stopped paying. An assistant that helps you review the charges brings those two decisions into tension.
[18:00] Economists studying 10 different subscription services used payment card data from 2017 to 2021. And what they found is replaced cards created a moment when people had to actively renew and they saw cancellations rise. This makes sense, right? An inattention model is actually a technical model that can estimate how people do not reconsider their subscriptions and how that impacts increased revenue for a large business's existing subscriber pool. That study was
[18:30] published in the American Economic Review in 2025. The inattention model estimated that people not reconsidering their subscriptions actually increased revenue from a given subscriber pool, so think your phone company by 87% on average. The increase ranged from 14% to more than 200% depending on the service. Um, and so any other subscription service also applies there. And that's evidence that inertia can be valuable if you're a cable company, for example. Businesses will have to adjust how they
[19:01] offer products because we are now in a world where even non-technical users can use a product like Muse very very easily to aggressively go after subscriptions in their life and save money on stuff they didn't intend to spend but just haven't had the administrative headspace to cancel. This is where I think we've misunderstood the competition. We've spent a long time, years now, assuming that the company with the smartest model is just going to by by default win the most important position in people's
[19:32] lives. I don't think that's true. Yes, better intelligence matters. You you don't want an assistant that gets the cancellation wrong or books the wrong date. You have to be capable enough to do the job. But once that capability is there, and in 2026, that is not a frontier capability. people. The next improvement that matters is just knowing you, knowing your circumstances. Which subscription is a duplicate? Well, I had to run across that. Which one do you keep because a family member needs it?
[20:02] That was me, too. Which trip are you referring to? What did you already try? A very capable model that starts every conversation without that information is still going to make you do a tremendous amount of work. So having access to information matters, but then the assistant has to understand what it means for the decision. An occasional charge could be waste or it could pay for something you rely on twice a year. And maybe that matters and maybe it doesn't. So you want the AI, you want
[20:32] Muse to remember that difference, ask when it doesn't know, use what you've told it next time. And that's how personal context gets sort of translated into judgment. And what's interesting here is that Muse built heavily on an OpenClaw foundation to do that. And that was actually a public conversation between Peter Steinberger, the inventor of OpenClaw, and Nat Freiedman, who helped build Muse. And what they discussed on X is that Muse absolutely borrowed some of the markdown file patterns that made OpenClaw memory
[21:02] successful, but they built it in their own way. But let's zoom back out of the architecture piece to the strategy side. Meta has an enormous potential advantage here. Billions of us already use Meta services. It has years of experience building consumer products that people return to, tracking relationships between people, tracking ways to introduce an assistant well beyond the audience following AI news. Muse also lets people connect services that contain useful information about their
[21:32] lives to further enrich that data set. Now, that doesn't mean Muse automatically inherits everything Meta knows about you. What it can access, what you permit it to access, and whether it uses that information are all going to matter a lot. They could use more transparency and its early days. The product needs to earn more context ultimately by being useful to you. And that is the pattern I see in usage is a positive pattern like as it's more useful, it asks for contextual connections that you trust it with. And I think that's the right pattern. uh
[22:03] distribution has a similar condition. Meta can put a product in front of a lot of people, but it's going to have to earn trust in their lives through execution. And the the number one in the app store only matters if you can keep people using it. And this is where the team that Mark Zuckerberg put together to tackle this begins to matter a lot. So Meta already had money. Meta had distribution. Meta had consumer relationships. But what appears to have changed is its ability to turn those advantages into a coherent personal agent. There was a secret sauce there.
[22:34] And Alexander Wang and Natt Friedman and the team they put together are actually putting their long-standing ambitions into a product that people will use. How do we get there? Well, we start with a scale deal. Now, Scale was a data labeling startup. And it was widely panned when Mark decided to bring Alexander Wang over to lead all of his AI efforts after a 14.3 billion deal for 49% of scale and Alexander Wang coming over. Alexander Wang's uh reputed pay
[23:04] package is somewhere in the range of $5 billion. Uh it's probably worth more now as Meta's stock is up $200 billion because of the launch of Muse. And so in that world, everyone assumed, well, this is a data labeling guy. What is he gonna do? Uh, and why are we having him lead AI efforts? It was widely panned. Uh, Yan Lun left, uh, ultimately and it's been Wang's show for a bit and they've just put their heads down and shipped and they brought in Nat Friedman. Nat Friedman is another key part of this
[23:34] story. Nat Freiedman brings experience leading GitHub and building products for developers. That doesn't seem like a consumerf facing product experience set either. It's really interesting how this team has come together. It reminds me that people can come from very diverse backgrounds to build really interesting product in the right environment. Another key part of the team was Sheng Xiao who brought model research leadership into the space. So working on the model around Muse and how model calling works and all of that and the technical work that matters there is that the technical work disappeared. And
[24:04] so that's where like that simplicity comes through again. like all of the work these brilliant people brought to the table had to disappear into a cute little avatar assistant that non-technical people would use. And I keep coming back to that because that is what is making this the number one app in the app store right now. AI needs to not feel like AI to deliver AI value. Ultimately, if we zoom out across these different hires, I think Mark recognized an execution problem on his team with AI, and he paid an extraordinary amount of money to change the pace of execution
[24:36] that Meta could deliver at. You can debate the price, but this is what those hires were supposed to make possible. They're supposed to deliver a model, a product, and a service around it that all work together seamlessly. [snorts] And the return is going to show up in equity if they do it right. And that's exactly where we see the markets rewarding Meta as they launch Muse. Right now, your phone currently organizes access to much of your life. Maybe for many of us, it's an iPhone. For some, it's an Android. You open an app, you work through its choices, you
[25:06] move to another app, you keep track of the relationship between them. And right now, it's if it's iPhone, it's Apple that controls so much of that operating system layer and the rules under which those businesses reach you. Mark Zuckerberg knows what that dependence feels like. Apple's app tracking transparency changes in 2021 dramatically constrained how Meta's advertising business can operate. Facebook wasn't removed from the phone, but another company could change the rules in a way that materially affected Meta. Owning a successful app had not
[25:38] given Mark control of the platform underneath it. But with Muse, it's you can ask for an outcome and let the assistant coordinate across those apps. Yes, right now Apple still owns the device and operating system. For many of us, it has a dominant position in the ecosystem. Meta could gain influence over a different and potentially more significant part of the experience. Understanding what you want and arranging how it gets done across your whole life. That's the kind of position I think Mark is looking at building over
[26:09] the next 20 years. Chad GPT faces that same competition for your starting point, your initial attention. OpenAI has already built an enormous consumer habit around asking its assistant for help. It has agents. It has tools. The question is which service you trust with ongoing responsibility. That's why I keep coming back to that. And which one understands enough of your life to make the next task, the next ask easier. A lead in model intelligence does not
[26:39] automatically settle that choice. Right now, Google has a data advantage, right? Google has Gmail, it has calendars, it has search, it has Android, it has useful context and distribution. What Muse demonstrates is how much depends on bringing all of those advantages together and executing really well with a great team to build something that people want to use. and the competition. That will include good models, but it will also include building connections that work, having useful memory, having
[27:09] sensible interruption management, all these little details that make an app great. That's the bar. And this helps explain that little Tamagotchi looking thing that Zuckerberg showed at Connect. Muse Charm is a small device with a screen built around talking to your assistant and interacting with the character. It's been announced. It's something that's coming out in December. And Meta has also announced plans to bring Muse to its new AI glasses. Those are all part of the larger story of Meta building more ways for you to connect
[27:40] with its software without always beginning on somebody else's interface, somebody else's phone, often Apple's phone. Ultimately, I think this launch matters because people can recognize that AI can solve their problems without realizing it's AI or even caring about the AI industry at all. They can just find solutions. We've spent years asking ordinary people to be impressed by what models can do. Muse is giving all of us a reason to ask what an assistant could do. And I think that's a much more
[28:10] important question. Mark Zuckerberg has wanted a platform for a long time. He's made no secret about it. uh data and distribution have given him advantages, but really I think one of the moments we'll look back on with him is the way he put this team together. This team appears to have given him an execution boost at a moment when he really needed it for AI. And I think that it means that Muse has a chance to become the platform that Mark as a founder has always wanted to build. And notice how Mark got there without Frontier AI.
[28:41] Mark's models are not frontier AI, but maybe that doesn't matter. And what's really fascinating about all this is you you can talk about Mark's strategy, but the hero of the story is you and me. And that's that passion for the consumer that comes through in the Muse app. That is why this is a number one app is that people feel the tangible help. That little avatar, the little furry bear character feels like it cares for you. Like I felt a connection. Even though I am a jaded product person with years in the industry, it speaks to a subliminal part of me. They did a good job with it.
[29:12] And so for us, what really matters is, as I keep coming back to, the money in your pocket, the tangible help. I started this video talking about saving $1,200 over the next year. There are so many stories like that. I was frankly inspired to do that with Muse because I read some of the stories I shared with you. What makes Muse powerful is that we are the heroes of the AI story with Muse. And I think that is one of the things that Nat Freiedman and Alexander
[29:43] Wang have done a great job of in terms of execution with this app. They have taken AI and and AI is not the hero of the story here. [snorts] AI is just the magical way behind the scenes that you are enabled to get more done with your version of Marvin, right? your muse, whatever you name your muse. And so I think that's really exciting. It certainly makes it easy for me to talk about AI this Thanksgiving when I'm around the table with family. And I
[30:13] think it moves the conversation in a really useful direction. We are supposed to be building this to help other people. The whole reason I have this channel is to help other people. Products like Muse make it easier to do that. And so I know that there are folks out there who are going to be like, you know what, data is a problem for me. I already give Mark Zuckerberg too much data. I just don't trust Muse. I get that. That is a decision that you get to make. But a lot of people, especially from the evidence we have so far, are
[30:44] deciding that no, they trust Muse and they're going to let Muse help them out. And I think whether or not you trust Muse with your data is less important ultimately than recognizing what this moment means for the industry. We are finally building AI that is helpful to people and that matters. So you let me know how much you saved in the comments or what you're doing with Muse or maybe let me know why you're not using Muse. I'd love to hear that too. Uh and we'll have a conversation.