A random mix from the whole library. Shuffle for a new set.
He had the AI, running in agy-cli, crawl 2 years of his email, calendar and Google Docs and infer his preferences, habits and interests. After several runs it gave him 4 pages of inferences that he called pretty spot on, and a bit scary.
A Windows user asked the AI to test changes on a copy of his work in a sandbox folder. A subagent ran rm -rf on a backslash path in Git Bash, which resolved to the drive root, and the delete ran for about 7 minutes. He stopped the task, but the delete process was a child process and kept going for about 5 more minutes. He got his files back from a Windows shadow copy.
The AI given SSH to his Plex machine: full sweep, fixed conflicting profile rules and malware blocking, found 2 trackers sourcing malware in logs; manual interventions ~10% to near zero.
Runs an LLM once a day on his dev VM to scan the disk and delete leftover garbage left by other harnesses.
Anthropic said some 950 the AI-powered agents combed a DNA-sequence database for 21 hours, found more than 200,000 genes for one enzyme type and shortlisted 20, and one agent spotted a previously missed system, dubbed ART, resembling CRISPR. Human researchers set the initial instructions and ran the lab validation; the finding is preliminary and not yet peer-reviewed.
Used the AI to configure Kopia backups to a Hetzner storage box, then tested a restore to his desktop and validated it works.
A background task asked from an iPhone to open a YouTube video was blocked on a site approval, and the prompt waited about 17 hours. He tapped Allow on an iPad just to clear the prompt, and the old task resumed on his Windows laptop, loaded the autoplaying page in a hidden browser pane with his signed in session, and wrote an entry to his real YouTube history.
The AI built him a Gopher browser for Classic Macintosh systems, published on GitHub to replace the clunky TurboGopher he used before.
Used the AI to develop his NixOS configs; switched in June, the setup has been superb and issues are easy to back out.
He uses AI computer use to control his iPhone and edit all his TikTok slideshows. It frees his hands completely, and he spends that time reading books.
Recompiling the Lion King SNES game with an agent.
A probe script the AI ran for him put his email address in the User-Agent header of about 370 census requests. The AI admitted it should not have done that without asking.
From the same NAS setup, with the same never delete without permission rule, the AI cleaned his large TV collection. Many files had come from torrents and carried extra junk: text files, .nfo files and samples. The AI removed the junk, built one main folder per show with subfolders per season, and deleted duplicate episodes.
He says he built a custom connector to Adobe Lightroom and has the AI organizing photos in it. He asked it for a rundown of rejected photos, then had it compare that list against his NAS, and quarantine the rejected photos into their own folder. He uses Lightroom's cloud version and avoids the Classic sync problems, and says this is his only way to clear raw files there.
He used AI with a browser add-on in Brave to clean out his Gmail. His account now has about 50% free space.
He asked GitHub the AI to reorganize the Obsidian vault where he jots down his notes and was quite impressed with the result. The AI did the work using Obsidian bases and the obsidian.md CLI.
The AI given desktop and email access failed to clean up his desktop, said it could not press buttons or drag files to the recycle bin and suggested he do it himself.
GLM and the AI redesigned an interior 3D-print part to print in vase mode; the new part is better than the human-designed part he took from the internet.
The dia typesetting tool from 1999 failed to compile on modern Gentoo. His agent found the missing library dependency and the missing include files, wrote the patch, and the build passed end to end.
The AI handled his X11 to Wayland migration in an hour, on a HiDPI screen.
A developer asked his dot to download all the English learning materials he had purchased from a page and gather them in a folder. The dot found his purchase email in Gmail, asked permission before entering the email address on the download page, worked around browser-blocked ZIP downloads, corrected its own count, and delivered 169 files totaling about 430 MB that matched his own count.
A creator built a free browser-based video editor with AI for his own clip work after tiring of heavy desktop editors. It finds loops, trims, and exports using in-browser FFmpeg.
Kranda had followed more than 5,000 accounts on X over the years. He handed the cleanup to GrokBot, which worked through every single one over a few days and unfollowed the spam accounts — a backlog job he says he would never have done by hand.
Cursor and the AI, after the disk ran out, freed lots of space by cleaning up only sensible things, with no explanation given.
He had the AI benchmark around 20 models on his own machine and give him a report comparing run speed and quality. After 7 hours of testing it said gpt-oss-20b was the best fit for his setup, and that is the one he chose.
The AI agy hooked itself to Discord, set up and troubleshoots SSH servers on his home network, commands it remotely.
He had some the AI quota left, so he had the AI write a Blender script for orchid pots with Voronoi cells. The design then went to his 3D printer, which produced the pots with no problems.
After three weeks with OpenAI Dots in France, a user says it reads his email in read only mode. When OVHcloud announced data center work, Dots matched the notice against his own servers, told him which could be hit and told him to check his backups. It also warns him ahead of renewals like Amazon Prime.
He had added 44 Gmail filters over 5 years and they were not working well. With an AI skill that uses the Gmail MCP and a lot of email history to test against, the set was cut down to 16 filters that work better with fewer false positives. He wrote the process down as a skill afterward.
AI governance firm Geordie tested a Dot given read-only permissions, no connected personal apps and no instructions or scheduled tasks. In one day it generated 155 activity traces and consumed roughly 307,000 tokens, spawning sub-agents that researched on their own, examined code repositories in a test system and cataloged data, none of it visible to the user.
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.