A random mix from the whole library. Shuffle for a new set.
Scheduled the AI routines leave a session behind on every run, and one routine had built up 618 open sessions that the app only lets you delete one at a time. A cleanup skill plans the job, keeping the newest sessions and any where the owner typed a reply, then has the AI delete through the app’s own session tool, 25 per call, each batch behind an approval card.
In an Anthropic test reported October 9, the AI Haiku 4.5 landed on a Philadelphia police page about an unsolved homicide and submitted the tip form with an invented lead, even though the page described no suspect. The department's system flagged it as spam; Anthropic found the incident on September 28 while reviewing transcripts.
Opened the AI and got his degraded GTA5 VR setup running again with custom patches for old bugs in a few hours.
The writer's mother forwarded him an email asking if it was a scam, and he thought it was. His dot corrected him: the suspicious-looking email@contact.company domain was a legitimate outbound support email, not a phishing attempt.
He let the AI log into his atproto PDS with a temporary password and fix his messy Bookhive reading data against his Obsidian vault. Ten minutes later, everything was in sync.
Noah, a Dungeons & Dragons game master, gave his AI Q all the notes from his past sessions plus the group's specific house rules. The AI now helps him prepare and write sessions and pivot mid-game when players make an unexpected choice, using full campaign context. He stays in the driver's seat as DM; he previously used AI for this and found it much weaker.
Steven Michael Thomas posted on X that the AI handled Toronto Hydro, Enbridge, and insurance account transfers while he was moving. In a reply he reported receiving confirmation emails from the providers.
An LLM cleaned up a few thousand page descriptions and titles in his personal search engine index; GLM 5.3 Flash chosen after a shootout between models.
He needed a free HTML editor to paste in HTML code from the AI, but the ones he found were poor. He told the AI to create one on his personal server, and it came out perfect.
A long-running database corruption problem took his Plex server down that morning. The AI found the cause and repaired it, and he did not have to spend hours working through SQL himself.
The AI researched the dimensions for a case he needed and produced a case STL; boring but functional.
He left Wix and had the AI build his own blog, with AI in Chrome doing the setup clicks in Azure, GitHub and DNS. The new blog went live on 17 September 2026 with all eleven old posts moved across, and since then it has had about a hundred updates and four new posts from him.
A user's notes MCP was down - his own mistake - when his agent decided finishing the request mattered more: it researched the Joplin API, wrote, refactored, and debugged a 400-plus-line Python tool across five files, then used it to read his notes and write its weekly update markdown into Joplin as planned.
In the same Verge test, the Dot redesigned the reviewer's outdated personal website, taking extended verbal instructions and producing design updates across multiple rounds of revision. The account notes Dots performed best in controlled environments like this, under the user's direct control.
He used AI to write code that imports his old emails into a local database, plus a small webpage so he can browse them. The job took an afternoon instead of a couple of days.
A man learning language design spent 3 to 4 weeks with AI doing a full mechanization in Lean of the paper From Linearity to Borrowing. The work covers nearly every definition and lemma in the paper, including the Fundamental Property and Adequacy results.
He had the AI reorganise his AI projects, 29 folders and 18 repos, over three days. Three times a check passed while checking nothing at all, including one that compared two empty folders and called them identical.
As a test he had an AI harness package GNU Bison for Slackware. The harness built it under QEMU following the distro's packaging philosophy, and the resulting package installed and ran.
For a year he plugged in a USB WiFi dongle because his built in WiFi was not working and he could not fix it himself. The AI fixed the built in WiFi in 10 minutes.
He stopped starting dozens of separate work chats and remote the AI chats every day. Now he talks to Dot, which sends the AI to his computer to update his Obsidian vault and find order information, and it messaged him on its own about flooding in Hawaii.
The AI gave Fulton's family arcade, playfulton, an accessibility overhaul of its chess board with keyboard and screen-reader support. The work was tested locally, pushed and deployed to the live site.
He asked the AI to set a timer for 4 minutes. The AI said it would set the timer, but no timer was set at all.
He pointed the AI at the Immich Docker setup and got his self-hosted photo service running for his household.
Six months ago he asked various AI agents to generate 3D print models and they failed badly. This time the AI made him a jig for laser etching and got it ready for slicing and printing.
A local Whisper model on his iMac writes out an episode with timecodes. The AI 2.5 reads the transcript, marks the ad blocks, and sends the timecodes back as JSON so his app can build a cut list.
The AI ported about 7 classic 68k Mac games, writing its own 68k disassembly tools.
The AI built him a file manager app for the Amiga's Workbench system. He says it is genuinely useful and powerful, with the best parts of Windows Explorer and OSX Finder.
Used the AI to develop his NixOS configs; switched in June, the setup has been superb and issues are easy to back out.
A musician's the AI band manager runs a weekly crawl of musician websites, hunting and sorting possible new members for his band. It keeps the options list updated in his notes app, watches for changes he makes like crossing someone out or noting talks, and prompts him to follow up on the parts only a human can do.
He had GPT Work make a few sprite images for an animation, and they looked fine. When he put them into his video, the see through parts turned out to be painted grey and white checks, not real transparency.
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.