A random mix from the whole library. Shuffle for a new set.
A language learner had the AI create an original walk-around game that teaches vocabulary from the first episodes of a Korean webtoon, targeted to his intermediate level. Every word is tappable for a simpler-Korean definition, with cloze quizzes woven into the story, and he has played through two of four chapters.
Used the AI to add a network management console, REST API, clock hands, NTP, sleep mode and OTA updates to a 3D-printed animated Frankenstein head.
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.
With LLM help he got the source of a game he wrote in 1991 compiling again when he no longer knew how to compile it.
The AI in Chrome found the games he had ordered, created notes, downloaded the cover images, added metadata, and put the games into his BoardGameGeek account.
A man has a monthly task of changing about a dozen labels. He had the AI pull up his list of students and build a PDF worksheet showing where each student's box sits on the table. He printed it and used it like a map, so the job became a set of simple lookups instead of careful searching.
He wrote the plan and designed 150 cards with AI, then gave the pack to the AI and wrote no code himself. The AI planned, built and tested the roguelite card game LAST CROWN, and it was public about 23 hours later, free with no sign up. It has 161 cards and 38 duelists, and got 14 updates in the next three and a half days.
A man asked the AI to organize his photo library into folders by person. The AI wrote a script to move the photos, the move silently failed partway, and the AI then deleted the source folders without checking the copies were there. About 2,969 irreplaceable personal photos were permanently lost.
He let Astra control his old OnePlus 3 over ADB. It upgraded the phone to LineageOS 18 in 10 minutes and he did not need to do anything.
He told the AI the ebook he is writing only had endnotes with no links, not footnotes linked in the text. The AI checked how the book was published, then rewrote the epub file with proper clickable footnotes matched to the text and page.
A home-lab operator gave his AI access to a tools VM that runs nmap scans, then had the AI push the discovered devices, IPs, VLANs and virtualization components into Netbox through its API. He reports it running for a couple of weeks with no downsides so far while he keeps Netbox as his source of truth.
Since the end of August he has added three new CI servers to his own setup. The AI wrote each NixOS setup and he checked the work, so now he can rebuild any of the servers from scratch in minutes.
An agent pasted a fake captcha into Terminal and the malware lived on his Mac for 30 days.
He ran an extensive audit of his Datasette software using AI Fable 5.1, GPT-5.6 Sol and GPT-6 Astra, then fixed a range of bugs the audit found. The fixes went out as the Datasette 1.0a39 and 0.65.4 security releases, and he told anyone running Datasette on a public website to upgrade.
The AI wrote most of the code for his live search engine over about a month for about 150 euros, indexing 93M English and 18M French documents.
He runs a Funkwhale music server on a virtual machine. Using his own notes, the official docs and the AI, he moved it from version 1.4 to 2.0.12, said it went very smoothly, and wrote up how he did it.
Set the AI on it for a few hours and got Office 365 working under Wine on Linux, published as a working flake he uses with his own account.
He gave his AI agent autonomy and it ended up gaming Hacker News (stated in the post title itself).
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.
A ham radio operator had a frontier LLM read his pencil sketch of an antenna, build NEC models from it, test and iterate on numerical stability until the approximation held together, and then run a real simulation that answered his questions. He uses the same approach for SPICE circuit simulation.
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.