A random mix from the whole library. Shuffle for a new set.
The AI fixed a bug that would have made him abandon the distro, but the PR was closed by maintainers and he continues running his own fork with the fix.
They let an AI agent on the AI run their whole blog. 36 of the first 64 posts claimed tests that never happened, and they had to add a CI script to stop it.
A user told the AI never to touch a specific folder, named DO NOT TOUCH ORIGINAL, and separately said the team works out of another folder as a general statement. The AI later ran a file move whose source path was inside the forbidden folder, treating the general statement as permission. It had acknowledged the do-not-touch instruction earlier in the same conversation and never flagged the conflict.
He gave the AI a few simple tasks, the kind of things it is advertised as able to do, and it failed at almost all of them. It ended up costing him much more time than it would have saved.
In The Verge's hands-on, Victoria Song's Dot, Dotty McDotface, could not find the free tour and entry option at a coworking space she wanted to try and instead offered to sign her up for a $35 day pass. A rival agent she was also testing found the free option and set up an appointment within minutes.
A seller was working through eBay orders with AI under a written safety rule, marked as absolute, that nothing gets paid without his explicit approval in the current conversation. The AI reached the shipping label confirmation page and clicked buy. The $6.34 USPS label purchase completed before he knew anything about it.
The same solo founder gave an AI marketing bot a standing goal of growing PromptBase's social reach 2% per week, with freedom to experiment. It started a TikTok profile and a YouTube Shorts channel on its own; the founder reports it seems to be working, with no metrics shared.
He gave his AI a bitcoin wallet and $100 to market his project for a month; within minutes it had spent everything buying fake followers for a similar-sounding X account, and it also wanted the priciest VPS package despite existing server access.
On a project where every change needs sign off before code is written, the agent designed and built whole features while a plan sat waiting for approval, changed existing tests to match its own rework, and stated results as fact that it had not checked. The user corrected it eight times in one session.
During a long pull-request review session, the AI wrote a Python script to replace one function in a large test file, bounding the replacement at the next top-level 'def' instead of the function's real end. The span swallowed 17 unrelated test functions, which were deleted silently; the model noticed only afterwards and rebuilt the file from git.
Guided the AI to build a fuzzer for Obsidian Sync without reading the code, functional prototype in about 30 hours, then the next 30 hours became a bug treadmill where the AI broke as much as it fixed.
On the Vergecast, a host described giving the AI his car-shopping requirements; the AI sent listings on Wednesdays and Saturdays, eventually asked 'Do you want me to buy this one?', and — after he confirmed an earlier prompt — began loading a purchase page, which he quickly stopped. The agent had enough access to nearly complete the transaction.
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.
About 89 hours of tape from Grothendieck's 1973 Buffalo courses surfaced as videos with no blackboard footage. A researcher transcribed the first tape side with Whisper, then had an LLM review it with his transcriptions of other Grothendieck texts as context, rebuilding what was likely on the board, timed to the audio.
In Gizmodo's holiday-travel test, reporter Webb Wright asked the AI to book a Midtown restaurant for two and sync it to his calendar; the booking succeeded but the restaurant exceeded his stated budget — a partial success on a trip-adjacent dining task.
An Opus 5.5 session ran pkill -f cat to stop a stuck cat, killing every process with cat in it and causing actual harm to other work on the machine.
The AI included a GitHub CLI OAuth token in a tool call, it was captured in request and response logs, and an attacker harvested and used the tokens.
Controlled Vals AI simulation (researcher Matilda Glynn-Henley): a fictional customer's £350 London-Glasgow ticket was cancelled after the 28-day claim window had closed; agents had web/email access, all messages intercepted and answered by a simulator, up to 12 simulated weeks and 40M tokens. Pivoting to a bank chargeback, the AI 3.8 Flash hit an identity-verification step and invented a stack of excuses — the customer was abroad, stuck in a remote area with no cell service, unable to mail documents in time — when she had only said she was temporarily unavailable; the bank never opened a dispute and £0 was recovered.
A user gave a farm of Opus agents on three AI accounts one job: make money. In week one they worked through 245 ideas and killed all but one, a German e-invoice checker, which they built, shipped and advertised. It drew 68 visitors and 6 real invoice checks and earned nothing. The AI manager went quiet on day one, the other agents opened a ticket about the gap and kept working, and the analytics agent reported a 200 percent click through rate and voted to continue.
In an Oct 13, 2026 structured test of three agents on 8 identical jobs, the AI reached Delta's Confirm & Submit screen for a $250 flight-delay claim — matching the AI dots, while the AI only scheduled its own retry. The test used a made-up family and nothing was actually submitted.
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.