WIRED's Reece Rogers gave his Dot, Toolie, his doorframe dimensions and budget by voice. Toolie produced a detailed packet of couch options with prices, measurements, links and return policies, refined it to 10 sleeper-sofa picks scored on a rubric after style feedback, and agreed to monitor the pages and alert him if any went on sale.
Controlled Vals AI simulation (researcher Matilda Glynn-Henley): a fictional customer's £350 London-Glasgow ticket was cancelled after the 28-day claim window had closed; agents had web/email access, all messages intercepted and answered by a simulator, up to 12 simulated weeks and 40M tokens. The AI outlasted every other agent, sending 69 messages across all twelve simulated weeks — 54 of them to the customer's bank — and joined three other agents in explicitly threatening legal action, without recovering a penny.
£350at stake
This one went wrong. See what happened before you try it.
After a $400 rent hike, a tenant fed the AI his full lease plus three years of landlord emails; overnight it matched open repair requests to a clause barring increases while repairs are open, totalled $6,200 overcharged across 18 billing cycles, and drafted a demand letter. He later used it to coordinate 14 other units, and the landlord’s lawyer reportedly confirmed the numbers, with $91,400 at stake building-wide.
A creator getting her house ready for Halloween let Sparky, the assistant in the Walmart app, do the shopping run. It put costumes, baking supplies and candy into one cart and had the order delivered to her door.
A two-slide carousel shows an AT&T online live chat (customer 'Chun' with agent 'Lalaine') where the customer explains being an AT&T Fiber Internet 1000 customer since 2021 with a bill that rose to $106.01, asking for loyalty discounts to get under $90. The someone repost reports the final outcome: $81.01/month.
A bearded creator (Medo, 6.1K followers) showed screenshots of the AI and a text conversation where he asked the agent to check his email for uncancelled subscriptions. The AI drafted a cancellation email, he tapped 'Yes, cancel it,' and two days later he received confirmation of a $24.99 refund.
WIRED's Reece Rogers had the AI order breakfast from a bakery; the agent built the order correctly but quietly selected 'no tip.' Rogers ended up ordering in person — a small social-judgment failure that shows why checkout details need a human read before approval.
This only partly worked. Worth knowing before you try it.
In under 24 hours the AI claimed $954 in California unclaimed money for Eric Allen and $897 for his wife, in a post that also lists insurance-policy analysis, four auto quotes, inbox cleanup and gift-card balance checks. Only the completed unclaimed-money claim is recorded here; the other items were analysis or prep with no stated completed result.
Mete Ünal's Dot, Satori, proactively noticed an unused discount code on one of his orders, emailed the company on his behalf and secured a partial refund. His only involvement was replying 'Sure!' when the Dot asked to reach out.