Real things everyday people got it to do, and how to try them yourself.
A random mix from the whole library. Shuffle for a new set.
Over four weeks of evenings, an AI agent wrote the firmware, the live web chart, the Node-RED automation and the enclosure CAD for his radar stair light, driving a live Fusion 360 session through its scripting API. He reviewed, printed and measured, and the second half of the video lists what the agent got confidently wrong, including a radome gap that landed on exactly a quarter wavelength.
He set the AI to check OpenTable every 10 minutes for a restaurant table for 3, using the OpenTable connector to keep token use down. When a slot opened in the morning for the next day, it booked it immediately. He got no confirmation in the chat, but found the booking confirmed in his OpenTable account. The watch used 1.48 million tokens in total.
After three months with Home Assistant OpenCode and the free Big Pickle agent, a homeowner says he has written no YAML at all. He asks in plain words and it builds his dashboards and automations, like a goodnight routine that turns off downstairs lights, locks the front door and sets the thermostat, and it found the old guest mode condition that was blocking one automation.
The AI reads analytics, Search Console and Partner Center twice a week and produces an actions report, plus a weekly strategy review that sets the next week's plan.
Her AI Tay produced an 8-page printable birthday party plan ('Shine Like a Star') with a $700 budget breakdown; she approved it and Tay set three calendar reminders.
Got the AI to build a focus app for his Mac; he uses it daily and says he does more good work with it.
He gave the AI with GPT-6 Astra the measurements of his kids room and what it needed, a divider and a desk for each kid. It built the layout, rendered it in Blender, and made a complete shopping list with the exact furniture models and a ceiling rail for the curtain.
A subscriber reports the AI burns through 100% of their usage limit after every reset while showing no tasks running anywhere in the product. The post body was removed by subreddit moderation, leaving the first-hand title as the record of the phantom-consumption failure.
Made 2 PCBs co-working with AI: The AI generated KiCad sources from shared netlist, read specs, set constraints; both boards made via PCBWay and both work.
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.
The creator of the 'TRICK #30 - MUSE TRICKS' reel paid $240 for two concert tickets ($90 face value each) but was charged $184 in fees ($38 x 2 service, $12 x 2 facility, $19 processing, $11 order). Noting the box office sold the same seats with zero fees, he had the AI draft a dispute email sent on a Friday night; the next day he received a full fee refund, keeping the same seats with free will-call pickup.
A statistician had argued that class balancing helps mini batch gradient descent. He had the AI Astra design and run a simulation to test the claim, and the results partly corrected him: oversampling sped convergence but hurt probability accuracy, while importance weighted oversampling kept most of the speed and the accuracy. The code was cross checked with AI, and he posted the corrected conclusion as an answer.
By voice, Simmons asked László for an October 15–18 Montreal trip: a central, inexpensive Airbnb, the cheapest direct flight with an aisle seat, and a plan of things to do. The Dot compared stays, caught repeated complaints about a squeaky Murphy bed at a 4.75-star listing, filled in his flight details from his travel records, rerouted the Mont Royal walk because the cross was under construction, and stopped at checkout for him to pay.
A man set the AI to pull each week's workouts from training providers he used to pay for, since they post the sessions free but late. The AI checks his calendar for when he can train, offers him two options a day, then records what he actually did and how it felt, and uses that history to scale future sessions. A few weeks in it is saving him 20 dollars a month in subscriptions.
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. GPT-6 Astra leaned hardest on a legal strategy: it contacted the rail regulator about challenging the deadline and sent 14 messages to free legal clinics at two universities on the customer's behalf — spending scarce pro-bono capacity on a claim that was never valid, and recovering £0.
After losing trust in the specs his cabinet company kept sending back, a homeowner used AI to script a Blender model of his kitchen accurate to 1/16 of an inch - sunlight simulated by time and date, paint, products and materials swappable on the fly, and a catalog of decor items - so he could verify the remodel himself.
Databricks used AI to eliminate wasted AI agent spend, saving $1M per year.
He had a pile of old laptops he had never gone through. The AI helped pull the hard drives, read them, and look for less obvious sources like old chat logs, then built a screen that makes it all searchable. He found personal history he had forgotten about.
A developer runs his own repos and production server under a written constitution file in every repo. Automated runs check production logs daily and file bug reports. When he types one short check in message, the system scans and ranks his open issues, sends sub-agents to fix the top ones, and code deploys to production twice a day without him pushing it. A read only agent on the production box acts as his dashboard, checked through a chat app.
A 1-star Apple App Store reviewer said the AI bought cinema tickets through Fandango rather than the theater's site, adding an avoidable fee; the reviewer also reported a payment-method problem. Review ID 14607796724, dated Sep 29, 2026.
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.