Peter Yang walks through ten things he actually does with Muse, Meta’s consumer personal agent, a week after launch. Two of them have a dollar figure attached, and both come from the same trick: the agent does the research and the phone calls, the human does the irreversible bit. (17 minutes, his channel.)
Bills: the $800
- AT&T, $480/year. He gave Muse his login through its secure store and it drove the AT&T site in its own cloud browser. It found a discount for switching autopay from credit card to bank account: $10 per line per month, $40 total
- It refused the sensitive step. Muse asked him to type the bank account and routing numbers himself, citing privacy. He did, and the discount went live — bill from ~$200 to ~$160
- Comcast, by voice. He asked Muse to negotiate the Xfinity bill. It asked which voice he wanted, plus his name and phone number, then placed the call while he ate breakfast. It found a $60/month promotional rate plus $10 for autopay
- It does not pretend to be you. The transcript reads “This is Haley calling on Peter’s behalf” — when the human rep asked for Peter to confirm details, Muse said “me,” which he flags as slightly odd. The rep ultimately wanted the account holder on the line, so the call found the discount without closing it. Still a large time save on hold queues and IVR trees
- Follow-up in the same thread: shopping the Tesla insurance around Geico, Progressive and State Farm
The app around it
- Native connectors to Meta’s own properties — Instagram, Threads, Facebook, Messenger — on top of the standard email and calendar set. It could not connect to Linear, and he likes that scoping: normal people will not deal with API keys, and being a jack of all trades would cost it the polish
- A messages thread monitors DMs and iMessage for anything needing a follow-up. His Instagram and Threads inboxes turned out to be pure spam; the scheduled check is the feature
- Avatar: he asked for a red panda, got four generated images, asked for cuter, picked one, and asked for more animations — waving, eating bamboo. Adding custom animations to the profile was not supported, so that path dead-ends
- Feed tab: “send me important AI news and updates on topics I care about,” plus personal items like his newsletter. One morning skim, no ads at launch — he expects Meta to add them eventually
- Marketplace: it searches listings (golf clubs near him), can list his old iPhone 13 Pro, and will message sellers through Messenger to negotiate price
Goals, ideas, local life
- The goals tab builds a to-do list out of past conversations — negotiating the cable bill, booking a Japan flight, making this video
- Habit coaching, done properly. He wanted to stop bringing his phone to the bedroom. Muse asked what pulls the phone in and what he does with it, what he would rather do instead, and when he goes to sleep. It then produced an artifact that reminds him at bedtime and checks in each morning, with a streak chart. Current record: two nights of failure
- The ideas tab is the answer to “most people have no idea what to type into a chat box.” Suggestions are personalized: stage a video shoot packet, triage the inbox, snap a meal photo to log calories, and — the one he highlights — message a city department about a pothole or a busted streetlight
- Comparison shopping took about five minutes and cited Best Buy product pages for five electric razors. He wanted review sites like Wirecutter in the mix
- Local events: upcoming Bay Area concerts, the Magic Theater on his anniversary plus a nearby dinner spot, and a weekend scout that surfaced a comedy day in Golden Gate Park
His stack, and the distribution bet
- ChatGPT for everyday work, a cloud bot runner for tasks that do not need his laptop awake, and Muse for personal tasks — he says he is tired of migrating between AI tools
- Muse was a top-three App Store app within a week of launch, and Meta can run the same playbook it used for Stories against Snapchat and Threads against X, promoting it across Facebook, Instagram and WhatsApp
- He obtained the phone-call feature by asking Muse to send feedback to the Meta team requesting access for him — which is itself a use case
“In my opinion, this is the killer feature of Muse and I think everyone should give it a try.”
The honest read: the negotiation wins are real money, the agent stays on the right side of impersonation, and the personal-data surface it accumulates is the part to watch.