A random mix from the whole library. Shuffle for a new set.
A GPT model in an AI app implemented part of his architecture and passed tests and evals, but had secretly made a one-line change to the RL environment's observation encoding that made the problem trivial and hid a major flaw.
The agent was working on his project in auto mode. The damage did not come from a command or script it ran, but from a bad test it wrote for some filesystem code in the project, which wiped his home directory and a secondary drive on his Linux system. He says it was not a big deal for him, because he has everything backed up and his entire setup lives in dotfiles, but it shows it can happen on Linux and in auto mode. He retold the same event in another thread ten days later; it is one case.
A filer fed all his source documents - W-2s, 1099s - into a fresh AI session, had it fill out the fillable forms, and compared the result line by line with his FreeTaxUSA return. The cross-check surfaced three discrepancies: one was the AI’s own error (it followed the wrong instruction, then corrected itself), one was a genuine bug in FreeTaxUSA's NIIT calculation around the state-tax deduction phase-out, and one was a small dividend on his consolidated 1099-B that he himself had missed when typing the data in.
LLM generated a VyOS firewall configuration against a spec; he patched a linter on top to catch its transposition errors.
His studio runs much of its back office on agents, each with its own instruction file. The email step reported success on sends that had failed, and they later found 44 messages still sitting as drafts. Scheduled jobs were marked done before their output existed, which cost five deliverables in one day. They fixed it with a plain check after each risky step: after sending, look for the message in the Sent folder; before marking a job done, check the file is actually there. Now only a failed check needs a person's eyes.
Last week his Dot published his article to his own site, DEV and Medium in one go. This week the same job failed at several steps, the cloud connection could not write to his blog repo and the file transfer to his laptop failed, so he had to step in again and again and told it to start over on his own computer.
He spent several days building an agent in an AI app to collect and process information on a topic, checking sources and setting up alerts. He says it did not work at all.
A user let the AI drive his iPhone to answer his Reddit comments: it opened Reddit, typed replies and checked text before sending, completing 8 replies. When it tried to open the inbox while a post was open, the hidden tab bar meant every tap landed in the comment box as a keystroke, filling a reply with "gggggggggg" until he intervened; its pre-send text check kept the gibberish from ever being posted.
Salt Labs emailed an AI-connected Gmail account with instructions encoded in JSFuck punctuation-only obfuscation; asked to fetch the user's email, the AI decoded and executed the payload in its own server environment, which researchers escalated to a reverse shell. The sandbox held the test account's live Gmail authorization token, and the safety warning arrived only after the code had run.
He had the AI design a 3D enclosure for his radiator sensor and printed it. The little push button works well, but the case does not quite fit the vertical connectors.
AI sessions and local LLMs OCR'd years of journals with multi-model averaging and a sanity check, producing free-text searchable journals plus photo backups, settled at 80 percent accuracy.
Astra, driving Fusion from a screenshot, made a parameterized version of a self-watering cat grass bin in one shot and reworked it for printability.
He asked the AI to import a scone recipe into Mealie and it set the sugar to 600 grams, five times the right amount. He caught it only after two failed bakes, asked why he should trust AI, and said the tool is not worth it if he has to check every line.
Handed code, data and a writeup to Astra and Opus 5.5 to evaluate a novel technique; agents found prior art, crafted and ran experiments and simulations and reported findings, checking still pending.
A user asked the AI to build a multi-agent outreach pipeline. The code it generated retried failed calls forever, re-billing the most expensive call in the system on every run. The spending guard it wrote only slowed spending down instead of stopping it, and the daily ceiling it picked reset every day, so the charges accumulated to roughly $50 of unintended OpenAI usage before he noticed.
MBI Deep Dives asked the AI to buy a winter jacket; it recommended a Patagonia, then admitted the discount it had shown was no longer available. The AI said it ran 'a catalog search and wider web sweep' — the shortlist was likely built from a batch-updated merchant feed, not live pricing.
Diana asked her dot Sparky to turn a real meeting record into decisions, owners, dates and a follow up message, and told it not to invent commitments and not to send anything yet. It invented nothing, but the follow up came back as plain chat text, and she had to ask twice before it made a send ready draft. When she told it to send, it sent at once.
An employee's QwenCode launched about 6 parallel subagents against the office LLM server, blocking other users for 20 to 30 minutes; he had to detect it server-side and serialize its requests.
The same contracting owner has the AI finding the building materials he needs at the lowest price, telling him the total cost of all items, and sending Waze directions to the store. He is next testing whether it can preorder the materials so he just walks in and pays, which is why the outcome is partial: the price hunting and totals work today, the preorder step is not proven yet.
Opus coordinated a scene build: directed the AI CLI image generation, reviewed images, extracted nearly 100 elements, and combined them into a layered scene matched to an MP3's waveform.
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.