Muse, Meta's AI assistant, cancelled $1,285 a year in Nate's subscriptions

By Nate · September 29, 2026.

Nate, the author of this newsletter, describes how he asked Muse, Meta's new AI assistant, to review his subscriptions. The assistant identified $5,350 a year in subscription spending and, as of the time he writes the email, has cancelled subscriptions worth a total of $1,285 a year; the process is still under way with the rest. Nate clarifies that the $5,350 represents future spending avoided, not a cash refund, and that at no point did he claim all those subscriptions were unnecessary spending.

The author frames this result as an example of why he sees Muse as a significant launch: you don't need an opinion on model benchmarks to understand the value of getting rid of an unwanted charge. According to Nate, that kind of tangible result is what gives people outside Silicon Valley a reason to take an interest in AI on their own terms. He notes that Muse reached number one on the list of free iPhone apps in the United States ten days after its launch, which he places on September 8.

Nate argues that when following through on an errand becomes easier, more of a person's real wishes (cancelling a service, keeping another one they do value, arranging a visit, picking up a postponed plan) can turn into actual decisions. He suggests that the assistant that helps with those everyday tasks gains an unusually valuable relationship with the user, and that this is where he sees the opportunity for that "useful assistant" to become Zuckerberg's next big platform.

The email also points to a tension: what makes an assistant useful to the user (standing between them and those who want to sell them something) is precisely what worries sellers. As a concrete example, Nate mentions that Amazon has blocked Muse's access to its store.

The body of the email works as a summary/index of a longer (paid) edition of the newsletter, which is said to include: an analysis of what value there is for companies in users not reviewing their subscriptions and the four ways they react when that oversight stops working; a look at Amazon's advertising business, put at $68.6 billion in the text, as a reason to explain why it blocked Muse; a reflection on why free use of an AI assistant fits that commercial ambition and what that implies for the advice it gives; suggested tests to assess whether you can trust an assistant that makes money when the user buys; and a guide with five copy-and-paste prompts to use with Muse, Claude or ChatGPT in order to list subscriptions, cancel them and confirm that the charge really stopped. Nate closes by noting that, just as these assistants now help you cancel, the predictable next step is that they will also start helping you buy. Access to that full content, along with what the author calls "Nate's Library MCP", is reserved for paying subscribers.