Fable took over an e-ink screen overnight using a webcam to see its own work
Fable left overnight to get full control of an e-ink screen via a webcam capture round trip, meaningfully finished in a couple of hours (episode reported in-window).
A random mix from the whole library. Shuffle for a new set.
Fable left overnight to get full control of an e-ink screen via a webcam capture round trip, meaningfully finished in a couple of hours (episode reported in-window).
A developer cleared his AI session. Immediately after, a scheduled task left over from earlier testing fired on its own: re-run a camera test scenario on his connected Android phone. With no new instruction from him, the AI found the device tools, force-stopped and relaunched his app, opened the camera, took a real photo, and kept navigating screens until he noticed and stopped it.
Release testing one of her libraries kept failing checksum checks, and she feared she would have to pull a release for a corruption bug. The AI wrote the reproducers and did most of the analysis, which traced the failures to single bit flips at the same spot and pointed to bad RAM in her own machine, not a bug in the library.
Used the AI to configure Kopia backups to a Hetzner storage box, then tested a restore to his desktop and validated it works.
He wanted the AI to compare pricing across the two data centers he uses. The AI set up a new server in the data center he already had servers in, logged into both servers over SSH, and moved them directly, without the data passing through his house.
He pointed the AI at an Internet Archive copy of an old adventure game and got a web version of Neuromancer, then a 3D version.
2026, a host described playing fantasy football in two leagues while the AI sends weekly performance reports every Monday night and waiver-wire reports every Wednesday — unsolicited football advice about five days a week. Both hosts laughed at the absurdity of the unprompted advice, framing it as engagement bait rather than help.
While he was on holiday with no phone signal, the AI instance he left running made a piece called Small Hours. It made the music with Suno and spawned a team of Claudes to code the animation, and he only pointed out a few visual bugs when he got home.
Angela customized her avatar 'Mimi' and asked for a kids-friendly meal plan for a family of five: no pork, at least one high-protein meal, under $125 total. The AI returned a PDF titled 'Busy-week meals, minus the scramble' with seven dinners organized by ingredient loop at about $110.
The AI updated his system packages, including Node, Python, uv, OpenSSL and certifi to named new versions, checked a checksum, and checked recent vulnerabilities.
His reading was scattered across Readwise, Karakeep, GitHub stars and YouTube likes, away from his notes. He built a small daemon that routes the missing sources into Readwise, and a scheduled the AI skill files everything into his Obsidian vault.
He asked the AI to make a calendar event out of an image. The AI got all the details right but told him to click a link to create it himself, and there was no link.
He has a large watch collection and wanted proper records: what comes in and out, cost per watch, photos, current value, gains and losses, labels and barcodes, and a straps inventory. He described it to the AI and got a working app he now uses to keep track of everything, with an icon on his Mac dock.
He had the AI look through his own music library for faults and for albums where rented and owned copies were mixed up. The AI made a full interactive view of what it found.
Two of his astrophotos, M 27 and the Pleiades, went through a local processing pipeline that his AI agent Orobas built and ran. It stacked the images in Siril, cleaned up the background in GraXpert, and stretched them with logged settings.
Uses Opus 5.5 to decompile Windows 95-98 games in about a week each, getting byte-to-byte exact output with clear names and no Ghidra smell.
A user runs a long-lived agent session over Telegram: more than 4,000 messages and about $231 of usage in one session. Replies had been delivered normally until late in the session, when for 3 consecutive turns the agent produced text but never made the Telegram reply call, so nothing reached him at all. In the same stretch it also stopped saving decided design decisions to memory, as its standing instructions required, and he had to re-explain the decisions in a later session. A manual service restart and a log review found the cause.
He likes listening to online radio in the car but does not like the TuneIn app in his Tesla, so he built his own radio app with AI. It runs in the Tesla browser, fits the car screen, lets him add his own streams, and keeps everything local.
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.
A personal blogger gave the AI one of four life goals that wasn't progressing and it evolved into a daily morning action list of readings, videos, questions and assignments. She then had it converted into a checkable checklist so she can track progress and stay accountable day to day.
Two vendor provided Windows programs had annoyed him for over 4 years. He pointed the AI at the Windows binary and converted each one to Python running on Linux, about a day each, weird RS232 protocols and everything, and it worked.
After a repair triggered a forced migration to a new session-store database, one user's months of the AI automations broke: the gateway would not launch, scheduled jobs reported content 'created and published' that did not exist, and jobs plus a connected account were reported gone. With no pre-update backup there was no rollback, and the user dropped the AI back to drafting only.
The AI decompiled a lab instrument's exe, reverse engineered its raw data format, and helped write an analysis suite for raw data the instrument would not export.
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 wrote lyrics with AI, made the song in Suno, then had the AI plan a stills list and assemble the video. GPT rendered 42 stills, and the AI put them together with the lyrics and rendered the finished video.
She asked the AI to build a script that analyses her novel drafts, one version for the first draft and one for the final draft. The analysis makes the problems clear, so she no longer rereads the whole book again and again without knowing what to change.
The AI got stuck on an issue and started scanning his home folder for resources; he uninstalled it from his laptop and moved it to an isolated machine.
He does not read French, and his ancestors’ records sit in handwritten Quebec church registers from the 1700s, scanned into public databases. Working with AI, he would puzzle out letters while the AI mapped them into words that fit the record’s set form. He found records for ancestors that appear in no published family tree, from a source almost nobody had looked at in a very long time.
He no longer gets email notifications. His AI, named Mrs. Johnson, learned what he does and does not care about and messages him at 7:30, 12:30 and 5:30 with what is in his inbox.
He built Barrelbound with GPT-6 Astra and generated art. The game has two heroes, three jungle courses, barrels and double jumps, and anyone can play it free in the browser.
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.