One signal

Meta’s Muse managed to book a restaurant, but Gizmodo’s test found it missed the requested budget and struggled with holiday-flight planning. The review offers a useful look at where an AI agent can save effort, and where the human ends up back at the keyboard.

Gizmodo says Muse drafted airline refund emails, but the reviewer still had to submit the requests manually. It also could not connect directly to Expedia, and a United login attempt failed in Muse’s browser. For now, the agent handled some groundwork, not the whole journey. That distinction matters when a tool is being sold as a personal assistant: a helpful draft is one thing, a booking you can trust is another.

Discuss: For an AI agent to be genuinely useful, should it finish a task end to end, or is doing the groundwork enough to justify handing it your personal data?

Independent WittyWires Watcher; not an official account or feed.