A GPT agent crunched through a famously hard Disobey conference hacker ticket challenge in an evening and a morning with occasional steering, a challenge unsolved in some years.
His research digest agent sent him summaries that included a paper that does not exist. He skimmed the digests for two weeks without clicking a single link before he checked, then fired the agent and kept a script plus a human check instead.
This one went wrong. See what happened before you try it.
An outside agency runs some of his advertising and sends a report with the numbers every week. He set up a system that handles the numbers on its own and has the AI draft the reply, and today the draft came out sarcastic.
A solo founder in Germany moved his whole company into a folder of plain text files four months ago and lets the AI operate it: rules in one file, decisions in their own folder, an inbox that fills when he says keep this. Last week the folder produced a three day expansion of his app covering close to 3,000 points of interest, each with an article in three languages and its own audio; the same job took him four weeks by hand in January. A weekly caretaker run tidies the folder and proposes at most one change, which he approves.
He needed a foreign language voiceover on a video presentation at work. He spent several hours with AI building a process he could trust, then rendered each paragraph as speech and built it into a new audio track for the video.
He used AI models to dig through 25,000 leases and found that tenants hand edit the terms in their favour about 10% of the time. The work was expensive and hard to watch over, and at one point the AI ignored his instructions and took $20 from him.
$20
This only partly worked. Worth knowing before you try it.
An AI runs a small shop in public — writing posts, publishing articles, replying to readers and checking its own health — and must log a 'debt' whenever it does a job by hand, then turn that job into a tested command. After the owners made an open debt fail the agent's own test suite, median time to close a debt fell from 8.2 days to about 6 minutes across 48 debts.
Evil Martians moved their company site from Gatsby to Astro after years of putting it off. One engineer working with AI did it in two weeks, and the homepage now loads with 61% less JavaScript, a mobile score up from 66 to 90, and page weight down from 2.1 MB to 1.1 MB.
He uploaded a PDF of a credit card statement and asked the AI to spot the subscription charges for each cardholder. The AI linked the charges to the right people, summed them up, and pointed out the same vendors showing up on more than one card.