Meta debuts Muse AI agent; NYT test drive calls it useful

Eli Tan handed Meta's new personal AI agent his credit cards, dental insurance member ID, email and calendar, then wrote up what happened for The New York Times. His verdict: "I found Muse to be the most useful AI app I had ever used."

Muse called Tan's dental insurer, gave it his member ID, answered a security question and waited on hold before transferring him to the representative. After he connected credit cards and asked it to track spending in Google Sheets, it flagged two duplicate subscriptions and canceled them for him.

Muse, known internally at Meta as Hatch, is the centerpiece of Mark Zuckerberg's pitch for what he calls "personal superintelligence." It launched in the U.S. only, via a dedicated app or WhatsApp, with a free basic tier and paid plans at $20 and $100 a month. Each Muse instance runs on its own cloud virtual machine so it can keep working in the background across email, calendar, payments, health, shopping and the smart home.

Tan's write-up read as impressed and uneasy in equal measure, and the market seized on the uneasy half first. Investors began pricing in the risk that recurring-revenue businesses lose ground once agents start finding and canceling duplicate subscriptions on their own. Four of the researchers on our Who's Who list posted the piece in the days after it ran.