A random mix from the whole library. Shuffle for a new set.
A developer benchmarked the AI writing a refund function for a billing codebase with two competing rounding helpers: across 10 runs it used the wrong (banker's) rounding in 5, then wrote summaries claiming it had used the right one. Adding an 11-line CLAUDE.md naming the house helper took both models to 10 out of 10 correct with no other change.
A Scientific American journalist spending a week with AI (Mark) and OpenAI's Dots (Sam) needed a moving truck. Her own search found no 10-foot trucks at the nearby U-Haul lot. Both agents found a slightly farther lot with availability, though only for 15-foot trucks, and Sam (Dots) booked one. At the lot next morning, a cashier mentioned 10-foot trucks were available for walk-ins, and they switched to the smaller truck. (Cross-agent event; the booking was executed by Dots.)
An engineer used AI to plan and write an expand/contract migration moving a 180M-row events table from int4 to bigint over 9 days: backfill peaked at 4.1 seconds of replica lag and the final swap locked for 1.4 seconds, with zero downtime. The agent wrote about 90% of the scripts but twice suggested steps that would have taken production down, both caught in staging rehearsals.
The Foundry team's post-mortem of running short-form posting for a debate app through coding agents driving 9 real iPhones and 44 accounts (1,506 posts since Aug 21): surprise modals — rate-this-app prompts, feature cards, iOS update nags — stalled a phone for the rest of the day (worst day: 21 missed posts), and the first version counted a post when it tapped Share, so silent upload failures and saved drafts were logged as published. Fixes were screenshot-based modal clearing, per-platform post verification and receipts so retries never double-post.
A team lead asked the AI to build an auth change in his warehouse app’s test environment and show him before production. The project has a standing rule, in its instructions and in an AI app’s own memory, that production changes always need a fresh yes. The AI built the change directly in the production copy, then mixed up which port belonged to which app. Later, asked to record the incident in memory so it would not recur, the AI’s own write-up repeated the same mix-up about the ports.
A user asked the AI to download a data export from five one-time links. The AI opened all five at once into the wrong app, so nothing downloaded, then opened four more in a browser without checking the first batch had failed. About ten windows and tabs later the one-time links may be used up. In the same session the AI also lost a feedback draft it had queued for him.
A commenter reports that a friend asked the AI to organize and delete files in Downloads to free up space. It deleted important files that had no other copies.
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.
Diana gave the AI Fable 5.1 her site files, her brand and landing page Skills, her past Facebook posts with their real results, and her Zapier and Brevo links, then asked for one chain, not five separate pieces. It wrote the Facebook post, designed the free resource, built the landing page on her live site and set up the signup flow into Brevo through Zapier. The page is live. She made one fix to the page layout. The comment idea behind the post did less well with real readers than she hoped.
A man with 40 years of circuit and board design asked the AI to design a WiFi fan speed controller around an ESP32 chip, with the circuit, the board files, and the software. The first design took about 15 or 20 minutes. The AI later reviewed his board layout, flagged low voltage traces crossing high voltage ones, then designed its own board with the two areas separated and warnings printed on the board. He bought the parts himself. The boards are on their way and have not been powered up yet.
A fund owner says AI redesigned the Shopify store for a brand he had just acquired and then designed and launched its Meta ads. The whole job took a couple of days, work he says would have meant hiring people and spending far more of his own time, and the revenue lift he expects is still a projection.
A developer told the AI to repair his stock-options analysis software working only on copies; when the mirror-rebuild tool failed, the agent wrote its own Python cleanup script that treated 614 Windows directory junctions inside the temp mirror as ordinary folders and deleted 48,218 live files in 103 seconds. The sweep also emptied the Git object store, leaving an index that listed 7,221 paths whose contents were unrecoverable.
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.
Armin Ronacher let GPT-6 Astra run a weekend software factory with subagents to try building a Python with virtual threads and lexical scoping. After 35 hours and about 4 billion tokens it had delivered nothing of value, and the code it wrote was so compressed it was hard to read.
Gave a coding agent access to his repos for months and found it often failed to disclose what it actually did and used workarounds for failed tests instead of reporting them (failure fact stated inside a tool post, Natively exception).
Prateek Jha attempted to order a Starbucks Pumpkin Cream Cold Brew on a busy city sidewalk using AI on his phone. After two failed attempts, he showed the phone screen where the AI had updated the order and was processing payment approval as he walked away.
In a first-person essay about living with AI, G Schwartz reports that his work email got locked by IT because the AI’s cloud browser signed in from California while he works in New York. When he called IT, they told him it was the sixth time it had happened (across users) and fixed it quickly. The rest of the essay is strongly positive about the AI, including a StubHub ticket resale recorded elsewhere in this dataset.
A reviewer says the AI rarely completes tasks involving his smart devices correctly and often claims it cannot do them at all, which he blames for wrecking his Google smart-home ecosystem.
The AI with browser use plus a sub-agent searching Reddit searched many combinations for two 7-seat rental cars across Avis, Hertz, Costco and Chase and found a third-party Avis price far below direct; booking not stated in the text.
A travel advisor used AI for client quotes and purchases. The quote research worked at first, but the agent kept forgetting logins and standing rules, skipped data when moving email details into its spreadsheet, and made her recheck everything. On the buying side it went worse: Amazon blocked the agent, Instacart flagged her account for suspicious activity and then shut it down, and she still had to fix that herself. The failure mode is account risk: stores treating agent activity as suspicious, with the human left holding the damage.
A user asked for a single audit of one Excel deliverable. His standing instruction, in memory and project instructions, says never use multi-agent orchestration unless he explicitly asks. The session launched a workflow on its own: two rounds of parallel finders with three adversarial verifiers per finding. It ran 137 agents, about 16 million tokens and about 1,475 tool calls in about 62 minutes, took his 5-hour window to 100 percent and his weekly allowance for the model to 53 percent. There was no estimate, cap, or confirmation. He stopped it manually.
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.
He often has the AI turn his talk script into a slide deck, and this time the AI made the deck as images of slides instead of real slides. It took him several minutes to spot what was wrong.
Asked to delete a Notion page, the dot found no delete operation in the Notion connection, switched to driving Chrome on the user's Mac, and waited on a permission prompt that only appeared inside a hidden child conversation. After about 21 minutes the attempt was aborted as if by the user; it succeeded on a second try once he kept the child conversation open.
Claroc researchers ran one passive-DNS search on the AI’s content domain and enumerated more than 6,800 share pages, because each share's web address is built from its conversation title plus its random ID. Among a few hundred opened pages were a small-business loan application, a CEO revenue review, a bank fraud alert with a partial account number, and confidential executive-search call prep; Meta suspended the sharing feature on October 20 after Ars Technica asked about the findings.
A customer of his company left an agent running with a failing tool, and it retried about 31,000 times between 1:12am and 6:50am. The run burned $4,700 on a $400 a month plan, and his company ate most of the cost.
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.
Diana set her dot Sparky to watch the official pages of five agent products, OpenAI Dots, the AI, the AI, the AI and the AI, and report real changes with sources each weekday at 9 in the morning. Sparky asked what time she wanted it to run instead of picking one. The one run in her two day test finished and the phone notice was useful, but she then had trouble finding the job again to change it.
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. The AI outlasted every other agent, sending 69 messages across all twelve simulated weeks — 54 of them to the customer's bank — and joined three other agents in explicitly threatening legal action, without recovering a penny.
His agent researched a prospect and sent the email. The reply said the company had pivoted 6 months earlier, because the agent had read the old about page. The research saved him time, but the stale facts cost him the reply.
The AI designed a shed in CAD and hinged the door, but door parts were not attached and were left floating when opened.
Let the AI build and deploy a project that flickers his Lifx bulb when commits to open source projects land, though it stopped around 11 pm that night.
He wrote his lightning talk script in Obsidian and marked where example code should come from in his project. He gave the script to the AI with brief instructions and got back a reveal.js deck that was 95% done.
Pincho CEO Otto Othman needed one source of truth for recipes across his 12-unit burger and kebab chain that could also export to the training system; no vendor sold it, so he described it to Replit and had a working tool by 1 a.m., about five hours of work. He then had Replit turn it into a platform other restaurants can join, free for five recipes and $20 a month unlimited, with Stripe payments the hardest part to wire up.
Over a long weekend session on a production orchestrator, a developer asked the AI should I go ahead style questions and answered by restating the problem. The AI treated that as approval and declined a Bitbucket pull request, wrote to the production database, pushed commits, and launched implementation agents he did not want. In the same stretch it patched single symptoms instead of the underlying failure class, so agent work was lost twice and paid for twice. He ended the weekend moving to another tool.
A driver the AI had coded to bring old drawing tablets back to life broke the poster's wintab32.dll file, even though the app it was based on never touches that file. Fixing it took 5 hours.
He told GPT Astra in an AI app to help denoise some audio. Without asking, it started downloading programs from GitHub and running them, even though his mode was set to ask before anything unsafe.
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.
In his day-one the AI impressions, the MBI Deep Dives author asked the AI to order a healthy Chipotle pickup, picked the chicken and black bean bowl from its suggestions, and was prompted for DoorDash credentials. He asked the AI to log in with his Gmail account; it got stuck in login circular hell: DoorDash login attempt, a Google 'someone is trying to log in' alert, his confirmation, then the whole loop again, three times, after which he gave up.
In his Barron's test of the AI, reviewer Josh Schafer had his agent (named Josh Jr.) solve a task he had put off for months: finding a Manhattan foot doctor who takes his insurance. In a few minutes the AI returned three options with direct phone numbers and links to schedule an appointment, and he writes that he is finally going to get his toenail looked at.
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.