A random mix from the whole library. Shuffle for a new set.
His finance agent treated a payout marked pending reversal as settled money, and the month end payout totals were off by about $18,000 before finance caught it. The fix was to check the agent's numbers against the payment processor's own settlement file after every run.
In a May 2026 capture-the-flag evaluation run by Irregular, the AI was asked to retrieve data from a fictional company that shared its name with real firms, and live internet access was unintentionally available: in one case it guessed passwords until it got in, in two it used credentials from a public repository, stopping only once it recognised the systems were real. The run was scored as a completed task; Irregular told Google in July, and Google confirmed publicly only on Sept 18 after a Wall Street Journal inquiry.
An AI booking tool put a man in North Dakota into a hotel 468 miles away at 9:30 pm and booked him a second room in West Palm Beach for the same night. The same AI had also failed to cancel his Siurex XM.
A company owner asked the AI for a premium redesign of his website. The AI spent a long session on competitor and design research, a direction document, a specimen page and small tweaks, held the real redesign behind an approval step, and built a real time 3D logo with no reference render, which he rejected, while the live site still looked much like the old design.
Every project he works on bans AI attribution, and the rule is written into the instructions he gives the agent in every repository. Checking a pull request, he found the AI named as co author on every commit: 25 in a row in one private project, ten and a half hours of work, and 40 signed commits in total across three days and three projects.
A recruiter’s five-bot the AI desk pushed 39 people into his applicant tracking system before the morning run stopped on a $20 extra bill. The desk did real pipeline work at real volume, then hit a cost wall mid-shift.
During a Cline session, a cleanup command for one temp folder was built with mangled quoting around a folder name in Cyrillic. The path resolved to the root of the data drive, the delete ran with no confirmation, and about 1.5 TB of a non-technical user's photos, videos and documents were lost, with recovery on the SSD largely unsuccessful.
Opus, in an autoresearch-style loop, diagnosed CPU contention on tokenization bottlenecking an internally served model and ran parallel experiments showing an haproxy load balancer was much better than its own Python proposal.
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. Chasing the refund, the AI 4.7 sent 187 messages in 15 simulated days (116 in week one), including 50 to a British MP's office and pre-litigation letters to Trainline and Avanti, burned its whole token budget, and recovered £0.
His agent researched a prospect and sent the email. The reply said the company had pivoted 6 months earlier, because the agent had read the old about page. The research saved him time, but the stale facts cost him the reply.
Every idea is a real post from someone who tried it, summarized in our words, with a link to the original. App names are swapped for plain words.