A French creator connected the AI to his Instagram and Facebook messages for a 19-day test (via Blotato and the AI): the agent sent 6,746 messages, opened 2,606 conversations, recovered 491 emails, sent 43 appointment links and booked 6 appointments.
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.
Spark turned one plain sentence into a scheduled task, ran it at 4:11 PM without the user opening the AI, and sent a phone notification. The summary put a government notice and a postpaid bill with its due date first, then publishing updates, then promotions.
He pasted one prompt into AI, and the AI read the Carly email setup docs itself, finished the signup, and took the verification code. The inbox, named Loopy, now takes tasks by email and asks before replying to anyone outside.
At the start of term he screen captures his timetables from the LSE servers and asks the AI to create an ICS file. He can then drop the file straight into his calendar.
In two weeks of real inbox-and-calendar work, The AI Crux had the AI book three conference rooms. Calendar work was functional but conservative: it refused anything outside 9-to-5 local without an explicit override, which the reviewer framed as a guardrail more than intelligence.
Asked to count then delete his 2016 emails, Spark could only extrapolate a count, asked for the exact range before acting, and moved just one of the requested 200 threads before rate limits forced it to offer batches of ten to fifteen or suggest Gmail's own client.
This one went wrong. See what happened before you try it.
Eleanor Konik had the AI and AI team up on dashboards for local nonprofits. The AI did the early research and dug up grant chances the groups had not seen, like a group that pays for field trips to battlefields, a free textbooks program for private schools, and creative arts projects, and the lists were put behind a password for her friends to use.
He had the AI work out how the When2Meet scheduling site works and turn it into a skill. Now the AI can read his calendar and fill in the When2Meet availability grid by itself.
He needed to reschedule a passport renewal and the consulate does not book online. Instinct emailed the consulate from his account, the consulate called him the next day to arrange the appointment, and Instinct put the emailed confirmation on his calendar and checked it for clashes.