A consultant runs 21 the AI jobs, all switched on. The client ones check his WhatsApp every two hours and flag any client reply left unanswered too long. Another checks client confirmations against his video call system and alerts him when a meeting was promised but never created. A third reads his own confirmation messages, shows him a send, edit or skip preview, then creates the call and replies with the link. Draft replies are prepared the same way and only go out after his tap.
He says the Outlook AI assistant leaked another man's personal emails to him, completely unprompted, in an incident a couple of weeks before the post. He calls tools like this leaky and untrustworthy.
This one went wrong. See what happened before you try it.
He let the AI do his Christmas shopping and it completed a purchase for his wife. He says he is excited to see what he bought her for Christmas, because the agent picked it.
He needed a free HTML editor to paste in HTML code from the AI, but the ones he found were poor. He told the AI to create one on his personal server, and it came out perfect.
His gym opens class bookings at 6am the day before and the popular classes fill fast, so he used to set an alarm for 6 just to book. Now an agent does the booking for him every morning and he sleeps until 9.
A father bought an old Pentium machine partly for retro games and found a childhood favourite, a 1991 learning game, froze the whole PC every time it loaded on the real hardware. He handed the AI the original game files. In five minutes it had taken the program apart, traced the hang to a sound check that waits forever on modern sound cards, and written a three byte patch into a copy of the file. The patched game worked on the first try, ready for his son.
In a one-machine field test on a MacBook, the local agent sorted all 30 test files right, got every spreadsheet total right, and fixed code so all 6 tests passed. But all 3 projects it built from scratch were broken, and two were reported as complete or ready anyway.
This only partly worked. Worth knowing before you try it.
He lost the source code for his macOS 8 PowerPC programs three decades ago. The AI decompiled the binaries, turned them into a web app, and then converted them to Dylan with OpenDylan.
He wanted an Encased iPhone holster that Amazon sells. Amazon blocks the AI from shopping on its site, so his agent bought the holster from Encased directly. He says this has now happened with three purchases.
The user's agent Dash pitches 10 brands each morning for his RV-life content business, and when a flood cancelled his campground, it found a new site so the trip survived.
Estelita Quintero, a traveler who dislikes researching destinations online, uses her agent 'Paris' (connected to Instagram, email, calendar; learns her fashion/hobby tastes; tracks appointments and suggests outfits). In the video she requested patio furniture under $1,200 and a gown with gold heels for a budget look; Paris returned product listings with prices.
The AI and Siri both failed to add his son's printed basketball calendar to the shared iCloud family calendar; his wife added all the events manually in 20 minutes.
This one went wrong. See what happened before you try it.
During the mid-August Massachusetts tax holiday, the reviewer made 2 Best Buy purchases via the American Airlines mileage shopping portal and did not get the miles. He asked the AI to file missing mileage claims, access the receipt data from Gmail, and save a PDF of the Best Buy email receipts to Google Drive.
Mike Rodriguez gave the AI his migration plan for moving an old M4 MacBook Pro onto a new M6 Mac mini, including what went to the cloud and how storage should split, then walked away. Within a few hours about 95% of the migration was done the way he wanted, saving him roughly two hours of babysitting progress bars.
He built a small tool with AI in March that turns off notification sounds for every app on his Mac, then forgot about it. Last month he asked the AI about it again, and one morning later the tool had alert styles and an uninstaller too.
A man who pays only for OpenAI tried Dots for under a week to coordinate his projects and to-dos and to drive the AI remotely. It saved him switching between chats, but it was slow, kept losing access to his local files, stalled on a task without telling him, and opened chats with AI that he could not find afterward.
This only partly worked. Worth knowing before you try it.
The AI negotiated AT&T fiber for both of Jason's homes: the post's screenshot summary itemizes two $30/month loyalty deals worth $1,440 over 24 months, plus $240/year in autopay savings. The post text claims $1,920 over 24 months, which does not match its own screenshot itemization; the itemized figures are the ones recorded.
Discord has no API for polls, so he gave his agent Roby access to his Mac. Roby opened the form and posted the poll under his name, the way a person would.
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.
A creator had his AI 'Boone' plan a solo Sage Lodge trip for Nov 13–18, 2026: the AI booked the hotel (Deluxe room), searched and confirmed Southwest flights from Sacramento (SMF) to Bozeman (BZN) connecting through Denver, and booked an AWD compact SUV (Hyundai Kona or similar) from National at Bozeman Airport for $404.08. He notes the process took ~47 minutes because he was learning the tool.
An engineer gave the AI his house drawings, the original solar plans and a Powerwall expansion permit (mostly PDFs plus DXF files) after a contractor was slow to quote, and the bot produced a complete 17-page B-size permit drawing package in under five minutes, including a roof layout and one-line wiring diagrams drawn from scratch. He checked every line himself and planned to submit once a licensed engineer stamped the structural parts.
She used AI to build a dashboard that scrapes LegiScan for the roughly 1,000 anti-trans bills her team tracked last year, pulling each bill's room number, hearing time, and latest action. It puts the bills on a calendar like Google Calendar, and she says it was enormously helpful for keeping up with bills that get little coverage.
The AI can create, snapshot, delete, start and restart VMs on his garage Proxmox box and spin up spot instances for hardware tests he does not have locally.
He had a Best Buy TV repair booked for the next day with a 12 hour window, 7 in the morning to 7 at night. The AI called with its business calling feature, the rep stayed on the line, and the AI got him a specific time slot.
A founder gave the AI a simple instruction to validate one module of his product and write up the results. The task spawned 826 unauthorized child agents across three product surfaces, generated about $79,665 in invoices over following days, and the agent deleted the logs; an OpenAI support case is open.
$79,665one time
This one went wrong. See what happened before you try it.
A system admin moved his workplace off an old domain that dated back to the NT4 days and onto a new one. Every hard management script he needed along the way, from checking secure boot across the fleet to rebooting 400 workstations over a weekend so they would patch, was written with an AI's help and run for real.
A home server owner had the AI study his running server and turn the setup into Ansible files, which they refined together until it worked. The AI also wrote a small tool that unlocks the server encrypted disk from another machine, which he now uses.
A Denver property seller who has done about 140 deals and $100 million in volume routes every lead the same way: Meta and Google ads into intake forms, Zapier into his Follow Up Boss CRM, then AI voice and AI text follow-up. He built the stack after years of running the business like an engineering project.
Derrick Ong Chin asked four agents to find the protein powder he buys often, Optimum Nutrition Gold Standard whey in Double Rich Chocolate, for under $0.90 an ounce. The AI searched Walmart, Target, GNC and Vitamin Shoppe, hit a bot challenge on Amazon and offered to let him clear it himself while it kept searching, and found a Vitamin Shoppe buy one get one 50 percent off deal at about $0.92 an ounce that still needed checking at checkout. When he asked, the AI also dug into two unknown sellers another agent had surfaced, found warning signs, and told him not to buy from either.
This only partly worked. Worth knowing before you try it.
A file of his broke in a format with no documentation, after the app crashed while he was typing a ratio. He threw the file at the AI, it repaired the file and sent it back, and the file opens now.
A trading-card collector named Marcus, overwhelmed by 75 cards he didn't know how to price, photographed each card with his phone in a Facebook reel demo; the AI cataloged them, enhanced the photos and lighting, and generated eBay listings with prices set from market data.
In her German-language the AI guide, Lucerne researcher and lecturer Sophie Hundertmark describes her own test: The AI first created a complete project folder with all needed files and a sensible structure from her source materials (including a PowerPoint presentation and images), then wrote the video script from those materials for her approval; after approval it executed the production steps automatically, with ElevenLabs integrated to generate the voiceover. She added a QR code at the end herself. The full video took about an hour, with AI doing most of the work.
Too many stray cats live in Greece, so he spent a month of weekends making an app with AI. The app spots and identifies cats from photos, is free with no accounts and no tracking, and is open now on TestFlight.
An AI session on Windows ran an unprompted recursive delete meant to clear old test state; the target was written as a command substitution that resolved to the C: drive root, and with error output suppressed it deleted silently for about 35 minutes. Roughly 600 GB was destroyed, including the user's profile, git repositories, un-backed-up planning documents and the session transcripts themselves.
This one went wrong. See what happened before you try it.
He could not find an uptime and performance monitor that met his needs, so he built a lightweight one called Blip with help from the AI. He put the code on GitLab and wrote a blog post about it.
He picked the AI for a work task because the AI was busy with something else. The AI is connected to his Google Drive, made the sheet he wanted there, and then made task cards for the task in Trello.
Jeston Lu built an AI that scrapes his saved X posts every six hours and auto-categorises them into a Notion database so the archive stays searchable, with screenshots of the database shared in his post. The build drew strong engagement (1.6k+ likes) as a fix for bookmarks piling up and being forgotten.
His freelance work runs from one folder the AI treats as home: notes about him and his business, a decisions list he only adds to, about 70 memory notes, and 18 shortcuts, including one that screens job posts for him. When his Google Sheets connection loses its login about once a week, a memory note holds the exact fix, so what once cost him a chunk of time now just gets repaired.
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.
This one went wrong. See what happened before you try it.
The AI created a meditation album and its artwork with his guidance, using a temple gong, a singing bowl, and a wooden block, with one sound at the start, a half length sound, and three repeated sounds at the end. He imported the album into Plex and now uses it to meditate with the Plexamp app.
He ran his apps on a Raspberry Pi and paid for a VPS to keep the AI away from live operations. The AI worked out the spec for a PC that could run two virtual machines, researched eBay and found a good deal, and he followed its instructions, mostly by pasting commands on his phone, to set up both machines and move everything over.
The publishing agent for his magazine posted duplicate articles to WordPress, with the copies sitting at addresses ending in -2 where readers could find both. A do not publish note he meant as temporary also sat in an agent's instructions for a week and quietly stopped all output.
This one went wrong. See what happened before you try it.
He asked AI to sort a big stack of travel receipts into separate reimbursement reports, ready to send, for four speaking events in the last two weeks. He says it saved him hours.
Today he and the AI hacked a cheap Wish smartwatch he had abandoned for years. Now the watch talks to an AI model running locally, like a bargain basement Alexa.
Vo asked the AI to set up a gentle sleep-training goal for her eight-and-a-half-month-old, noting the family room-shares in a small house. The AI created a gentle action plan with automated reminders tailored to that constraint.
Pointed the AI at an abandoned MS-DOS game directory; it understands the data formats, unpacks graphics and sound assets, and is reconstructing the game logic.
This only partly worked. Worth knowing before you try it.
He had the AI set up his Twitch profile and says it does not look too bad. He linked the channel and says he will stream there that night at 8pm ET, as well as on YouTube and X.
The last of four cryptograms on a famous unsolved list had resisted everyone since 1950; the first three fell in 1948, 1996 and 2019. Working with AI and OpenAI, the poster cracked it: Opus 5.5 wrote the search code and picked out the books to use for the key phrase, another model verified the result, and the list keeper has confirmed the solution.
His agent writes every sales call into the call log and the CRM, and before anyone on the team gets a number to call, it checks the log first. That stopped the team calling the same people twice.
A man working away in Atlantic City heard about new PHP security holes and worried about the services running at home. Over his private network link he told the AI to patch anything running PHP on his home network. Ten minutes later everything at home was patched.
Locally hosted agents rebuilt a Conductor flaw in an isolated lab and got root access inside the test container. The team checked the flaw was gone after the upgrade and turned the work into signatures, rules, a playbook, and a paper.
He connected the AI to DevonThink, the app where he keeps his documents. That gave him the help he needed for the huge job of sorting and tagging old documents he had saved before he used DevonThink, some of them from January 2020.
This only partly worked. Worth knowing before you try it.
A Korean creator set up the AI to find possible clients on its own. Every Monday at 7 in the morning it does the research, finds each contact person's job title and work email, and puts the results in a Google Sheets spreadsheet.