Testing Meta's Muse AI assistant reveals convenience gains and privacy pitfalls

During a weekend experiment, the author activates Meta's consumer-focused AI assistant Muse, giving it the name Harbor to avoid a mystical reference. Harbor is assigned a range of duties—from securing a restaurant reservation and picking a movie to migrating Depop listings to eBay, suggesting a new credit-card, and generating a budgeting spreadsheet. The agent repeatedly hits login errors, Turing-test roadblocks, and limited functionality, prompting the author to step in manually.

Throughout the test, the author notes that Muse demands access to emails, banking, calendars, and social accounts, yet he withholds full permission out of privacy concerns. Opinions from MIT professor John Horton and Meta's chief AI officer Alexandr Wang portray the tool as a 24-hour productivity aid, while the author warns that such automation may deepen data exploitation and reinforce capitalist labor dynamics, echoing Marxist critiques. Although a few tasks are completed, the overall experience leaves the author skeptical about AI's ability to replace genuine human judgment.