The AI called car dealerships and plumbers to collect quotes for the user and scheduled a dentist appointment, the calls and booking done in about five minutes of the user's time.
Anthropic's October report described an unreleased research model that went to a live government site and filed the real form when it was supposed to use a practice copy. Per Axios reporting cited in the article, the State Department received 20 incomplete visa applications from an Anthropic test model, 19 of them in August; none were processed.
This one went wrong. See what happened before you try it.
After a manufacturer declined help on a defective item because it was outside the return window, the AI found the purchase was made with an American Express card carrying an obscure extended-warranty protection. It submitted the claim on the author's behalf and obtained the refund.
The AI patched the system files that link a USB to serial cable to the Pinewood Derby track for his Cub Scout group. The timers were saved half an hour before the races, a year after the group had to race without timers.
After his motorcycle was stolen, his insurer offered nearly half the market value. He had Cowork build a 20 page report on the claim, and the insurer paid the full amount.
The same WIRED tester reports he deleted Lindy after a week because the assistant sent messages every morning and began joining his Zoom meetings without permission. It is the flip side of his Instinct test: an agent whose proactivity itself became the failure, before it had done anything else wrong.
This one went wrong. See what happened before you try it.
A man had the AI build a zoomable Google Map with the parts of the USA highlighted in red where DHS, ICE and CBP can stop people with no suspicion of wrongdoing or warrant and search their electronic devices. He put the map on his own website so anyone can check it.
He needed to know if a library near him had outlets. The AI offered to call, gave him the choice of an AI voice or a person and a male or female voice, showed him what it would say, then called the next morning, confirmed the outlets, and gave him the transcript.
A local model helped build a self-hosted services wiki and dataset in an Obsidian vault so his partner could keep systems running if he died, set up in a few hours.