2026.10.12 | Substack | META | Meta Wants More of Your Life

Forwarded this email? Subscribe here for more

But would you let it?

Gaetano

Oct 12

READ IN APP

Well, today was a big day for Meta.

Check this out.

They just released Muse Wonder and Muse Wonder Full, their newest AI models. A lot of the work behind these has been focused on getting agents to remember what you want and actually follow through on it. And after going through all the updates and the interviews with Zuckerberg and Alexandr Wang (Chief AI), I think we have a much clearer idea of where they want to take this and how they think Meta can make money from it.

This connects directly to what I was discussing in my hyperscaler report. I have spent a lot of time investing in the companies supplying the AI buildout, and I want exposure to the businesses that could benefit as that spending turns into products people use. With Meta, there is already a massive advertising business where AI can improve results. But I also think there is an opportunity for them to become much more involved in what people and businesses do throughout the day.

If you have used agents, you have probably experienced some of the problems they are trying to work through. You give it a task, explain what you want, and then somewhere along the way it forgets something or loses track of the original goal. You end up checking its work and repeating yourself, which takes away a lot of the reason you wanted to use it in the first place. I have had that experience myself and by the time I correct it and approve all the things I need ot, I should’ve just done the task myself.

Meta is putting more work into the model’s memory and its ability to stay focused as it works. They describe the goal as helping it “keep track of what matters to you and work diligently toward what you want to accomplish.” I think that is a useful way to understand what they are trying to improve, because the experience has to hold together across the whole task.

Wang explained in his interview with David Senra that the team identified more than 100 specific behaviors and capabilities needed for the personal agent. He said, “we built evals for all of those. We figured out how to train the model to be really really good at every single one of those things.” They would review where it was struggling and keep working on those areas. That gives us some insight into how much of this comes down to making the experience dependable enough that people actually want to keep using it.

Zuck gave a pretty good example of how this could become part of someone’s routine. He has been using Muse to help organize weekend baking with his daughter. They talk about what they want to make during the week, and it helps arrange the ingredients so everything is ready when they want to start. He said, “I went and got an Instacart account to give Muse, and now I use it like once or twice a week.” Before that, he said he barely used the service.

There are plenty of things people would like to do that involve enough preparation that they just never get around to them. If the agent can take care of some of that work, it becomes easier to follow through. In his case, that turned into recurring purchases through a service he hardly used before. It is one person’s experience, but it gives us a concrete example of how agents could change purchasing behavior.

And if you are the business receiving those orders, you have a reason to make it easier for agents to work with you.

Zuck described how he thinks this develops. Muse can start by operating through existing websites. Direct connections with willing partners can make that faster and more efficient. Over time, he expects businesses to have their own agents that communicate with personal agents, and Meta is building the business side as well. So you could have one agent that understands what the customer wants and another that knows what the business can provide, working through the details before bringing something back for approval.

This is where I think Meta’s existing relationships become valuable. They already help businesses find customers through advertising, and their apps are already used for conversations between customers and businesses. If agents can help move those conversations through to completed purchases, Meta could become involved in more of that process. And on the business side, Zuckerberg also talked about a “business in a box for small businesses,” helping owners handle more of the work involved in running them.

That is very similar to what I was thinking through in the original report. If an agent understands the business well enough to help with customer questions, orders and the economics of its advertising, it could become something the owner uses regularly. The value to the business would come from how much time it saves and whether it helps produce better results.

Zuck was also fairly direct about how Meta could get paid. He discussed the possibility of taking “a small cut of what you are saving or making for people.” There is no announced fee structure behind that statement, but I think it gives us a clearer view of the ambition. Subscriptions can pay for more capable assistance, businesses can pay for useful services, and management is considering ways to earn from the value agents create around commerce.

At the same time, Meta already has somewhere to put AI improvements to work financially. Today, its engineering team also explained how it uses pockets of idle computing capacity to improve ad recommendations, either by evaluating more potential ads or using stronger prediction models. They said this can improve advertiser outcomes “without requiring additional capacity orders.”

This shows another way these investments can contribute. Meta can work on building a new consumer business while also finding ways to get more value from its infrastructure inside the advertising business.

The cost of doing all this is still significant. Meta generated $31.9 billion of operating cash flow in Q2 and spent $31.1 billion on capex leaving $784 million of FCF. So delivering these experiences at a lower cost is an important part of how the investment eventually pays off.

Zuckerberg said he hopes to eventually “have a model that is as good as Wonder, but in the Spark form factor,” referring to Meta’s smaller model. He connected that goal to serving hundreds of millions of people. If they can preserve more of the capability in cheaper models, they can make the everyday assistant more useful to a larger audience while charging for more demanding work.

All this being said, I think we are at very very early stages of agentic adoption. And most people don’t even know of good use cases for it aside from just playing around with it. That’s how general AI chatbots started too. Most people just asked it silly questions and then never used it again. But we are starting to see chatbots penetrate deeper into everyday lives to the point where they are being used in a variety of ways that are very sticky. I myself rarely ever search anything on Google anymore.

And I think agents will have niche applications that work really well, and the most challenging part at this point is creating a consumer product that is so seamless and trustworthy that we get mass adoption.

I can see it playing out where it becomes so advantageous to have your own set of agents, whether you are an individual or a business, that it becomes too hard to ignore. And at the same time, we have gotten to a point where adoption is so seamless. Anything that I can do to be less distracted and have more time doing what I like to be doing, the more likely I am to adopt a product.

One specific example for me is booking appointments or speaking with customer service. Such simple tasks that yet seem to take up way too much time. Right now I have seen a lot of pushback from companies that refuse to speak to agents over the phone. I do not see that sticking around. I think most businesses will adopt their own agents to handle communication with personal agents. And we can just waste less time.

I think Meta is uniquely positioned to create the consumer agent that people will adopt.

I do think that Google has earned people’s trust more. But Zuck is a visionary and he knows how to get people to use a product. Subscribed