Meta's new AI assistant, Muse, has recently become the top free app on the American App Store, signaling a notable recovery for the company following several setbacks in the artificial intelligence sector. This success has been welcomed by financial markets, with Meta's stock rising more than 35% in September. The performance of Muse has sparked discussions about whether Meta has finally discovered the right approach to consumer-focused AI. Meta's strategy for distributing Muse highlights the distinct nature of Apple's different platforms. On iOS devices, Muse strictly follows sandboxing rules, which are necessary for apps to be listed on the official App Store. However, the Mac version of Muse is not available on the Mac App Store and must be downloaded directly from Meta's website. This discrepancy is due to the Mac version's ability to access deeper system controls, which conflict with the Mac App Store's security policies. John Gruber, a technology analyst, noted that although these restrictions may be frustrating for developers, they offer users a greater sense of security. In Silicon Valley, the emphasis is shifting toward "intelligent agents," which industry leaders believe could make personal assistance more accessible to the general public. However, allowing a digital entity to manage aspects of one's personal device raises concerns about privacy. John Gruber stated that he would not allow a stranger to use his Mac, even with his supervision, and expressed similar hesitation about trusting an AI. These concerns extend to situations where an AI assistant might scan emails, manage calendars, or access bank accounts. While hiring a human assistant involves similar levels of access, trust is typically built over time. Without granting such permissions, tools like Muse might remain more like technical novelties than genuine productivity enhancements. Unlike traditional assistants like Siri, which respond to direct user commands, new AI agents are designed to be proactive, aiming to predict and fulfill user needs before they are explicitly stated. This shift raises broader questions about digital intimacy and the extent to which users are willing to allow an AI to act on their behalf, including managing their Macs.