Shopping
8 public posts of somebody having an AI agent they text handle shopping, from 7 people, on Muse and Tomo. 1 of them posted a figure, $245 to $245, each the author's own. None reports going wrong; people post successes.
The vendors sell this job and the users post it less than the money-back ones, which is the finding of the piece on getting money back. What they do post is specific: a garage freezer bought to a spec, tracked and registered for warranty; a whiteboard shopping list photographed and ordered from Sam's Club with one round of corrections; a green puffer jacket bought at the cheapest reachable price after the agent found live discount codes; a full wardrobe after a style analysis; a chocolate cake chosen by the agent and by the wife separately, the same one. Tomo's founder published its product's first three purchases, each tied to a goal the owner had set, which is a different shape: the agent buying for a reason it holds. Some of the posts on this page are from people whose profile says Meta, and are marked as the company's.
The figures, as posted
Every dollar figure somebody attached to one of these posts. The figures are theirs, unverified, and never added up here.
Muse
Tomo
- $245spentthree purchases: $81, $55 of whey protein and a $109 alarm clockTomo
TomoThe posts, by agent
Muse
A marketing lead testing both agents has split them by job: Instinct runs her calendar, email triage, the children's schedules and meal planning; Muse runs everything to do with social media, organising, basic editing and analysis for an Instagram account, and her shopping. She says it did a style and colour analysis, found what was missing from her wardrobe using her bookmarks and Instagram search, and helped her buy the pieces, a full refresh in a day.
Why it is notable
A person keeping both, with a reason for each, rather than a switch: the life admin to the one that acts first, the shopping to the one already inside the account where her bookmarks and searches live. The wardrobe is the second outside account of Muse buying clothes end to end; the earlier one was a jacket.
Posted by a user by Kim Currier (@Kimcurrier): x.com, 2026-09-18. Added 2026-09-19.
His wife wanted chocolate cake and did not trust the agent to choose. He had her pick one herself and sent his Muse to find the best; they chose the same cake, and Muse ordered it.
Why it is notable
A trust test run in a kitchen, with the sceptic's choice as the control, and the smallest purchase on the record. It says more about how households adopt these products than a larger order would.
Posted by a user by Bryce Del Rio (@BryceDelRio): x.com, 2026-09-14. Added 2026-09-17.
Muse shopped for a garage freezer to a required spec, found the best price, ordered it, tracked the delivery and submitted the warranty registration. The user watched every step and had to enable a few more permissions along the way.
Why it is notable
The whole life of a purchase, through to the registration card nobody fills in, and a user who says he watched each step before deciding he could trust it. The permissions he had to enable are the approval model, seen from the user's side.
Posted by a user by Infinite Improbability (@v01dp1r4t3): x.com, 2026-09-13. Added 2026-09-17.
One line: Muse completed a Ticketmaster purchase where Instinct had failed.
Why it is notable
Ticketmaster is the checkout with the most bot defence in consumer commerce, so the outcome is a fact about which agent's browser gets through, not about taste. A single line from a user, with no detail, and recorded as exactly that.
Posted by a user by Weida Tan (@weidatan0): x.com, 2026-09-14. Added 2026-09-17.
A user photographed a messy shopping list on the fridge whiteboard and Muse ordered all of it from Sam's Club; one round of corrections fixed the ambiguous items, and it then ordered what Sam's did not have from another store.
Why it is notable
A photo as the whole instruction, and a second store for the leftovers without being asked. The correction round is the honest part: a list written for the person who wrote it needed one pass of feedback.
Posted by a user by Dylan Heagy (@dylan_heagy): x.com, 2026-09-09. Added 2026-09-17.
- WhatsApp
Asked to find a green puffer jacket, Muse searched, found live discount codes, added the jacket to a cart, applied the codes and bought it at the cheapest price it could reach. The person reporting it was showing it to his wife rather than testing it, and describes her reaction rather than the mechanics.
Why it is notable
It is the first public account of an agent in this directory running a complete purchase end to end, including the step nobody demos: hunting live discount codes and applying them at checkout. Meta's own launch material describes Muse as able to use a browser and connect to apps. It does not claim this sequence, and a claim from the company would be worth less than this anyway, because the interesting part is that it worked on a real store the vendor did not choose.
Posted by a user by BuccoCapital Bloke (@buccocapital): x.com, 2026-09-10. Added 2026-09-10.
- WhatsApp
The second example from the same person: Muse found a recipe and loaded the ingredients into a Whole Foods cart, leaving the order to be placed. Reported in passing, as one of two things he showed his wife.
Why it is notable
It is the same shape as the kitchen agent listed here separately, arriving as a side effect of a general assistant rather than as a product. That is the pressure a single-purpose listing is under and it is worth dating: one of these two products was built to do this and the other picked it up on the way past. It also stops one step short of paying, which is the distinction this task page exists to draw.
Posted by a user by BuccoCapital Bloke (@buccocapital): x.com, 2026-09-10. Added 2026-09-10.
Tomo
- iMessage
Tomo's founder published the first three purchases the product ever made, each tied to a goal its owner had set. The first was $81 on a Japanese textbook for a user in Hawaii who had always wanted to learn the language. Then $55 of whey protein for someone getting in shape, and a $109 sunrise alarm clock for someone trying to sleep better.
Why it is notable
Three amounts, three items and a reason for each, which is more specificity than any other product here has published about money it spent. The pattern is the finding: all three are aspirational purchases against a stated goal rather than replenishment, which is the argument the company makes for why people let it spend at all, and it is checkable against whatever the next batch looks like.
Published by the product by Justin Quan (@justoutquan): x.com, 2026-09-10. Added 2026-09-10.