Locally hosted agents rebuilt a Conductor flaw in an isolated lab and got root access inside the test container. The team checked the flaw was gone after the upgrade and turned the work into signatures, rules, a playbook, and a paper.
The AI completed and submitted the PRESTO service-guarantee form for a disrupted GO Train trip, the transit agency's compensation route most riders never bother with.
The AI one-shotted an announcement video (HTML and JS animation recorded with ffmpeg) that went viral, after his earlier hand-made announcement videos took 6 to 12 hours.
A man fought USAA for over a month over the right deductible for roofing damage. The AI read his insurance documents and helped him keep a running argument, and he won and saved $1,838.30.
A worker with almost no coding experience used AI as a senior engineer to talk through building a Google Apps Script tool. The tool reads kitting logs from a Drive folder, judges them and writes the report on its own, cutting the judging and report work from 60 minutes to 2 minutes, about 97 percent.
Told the AI which plugins and settings he wanted in Neovim; now has a single-file setup used across four machines, and the AI bundled gitbutler-cli from source for him.
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.
This one went wrong. See what happened before you try it.
Martin Cox set up Shopify's shop.md in an AI app and told it the running shoe model he normally wears. The AI found the exact pair in his size on sale from the brand, built the cart and checked out with Shop Pay, all inside the chat.
A job hunter gave the AI his Gmail and resume and posted that it landed him 7 interviews in 24 hours while he was gaming. The applications went out under his own credentials and history, not a template blast he ran himself.
Noah connected Spotify to his AI and asked for a playlist built from his already-liked songs but limited to tracks he has never actually listened to, as a way to surface new music from his own taste history. The AI can also just play his liked songs on his phone or computer on request.