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.
Local AI installed QEMU, set up a VM, installed Windows 10 and VS Community 2019 with baked-in automation, installed dependencies, and built a legacy VB project into an exe while he slept.
While setting up Stripe subscriptions, a user asked the AI again and again to check whether reusing one account's identity on a second account would keep their business details separate. The AI said yes with confidence and never checked. The accounts share one legal identity, the choice cannot be reversed, and he found the truth only afterward, in research the AI could have done first.
Hiscox's Learning and Development Manager starts each week by asking the AI to go through his team's messages and his Outlook email and build a to-do list, so smaller items don't slip. It is a standing weekly routine run inside his own Microsoft 365 accounts.
Asked a chatbot to make a browser extension that adds the Google Maps link back to search results; uses it almost every day.
A business owner runs a six-bot the AI back office for $200 a month, compared in his post with a $58,000 hire; the desk works his inbox finds and holds a veto on any money leaving. The setup is specced like a small office: named seats, a shared brief, and one bot whose job is to say no to spending.
He used AI to jailbreak an old Kindle and set it up as a small wall dashboard. The whole job took about 10 minutes.
Diana gave her dot Sparky one brief for a standalone review lab page, with a ten test grid, the scoring rules, a checklist and a clear note that it must not invent test results. Sparky built the page in an AI app Sites, published it, told her on its own that the page was set to private, and gave her a link she could share.
He asked the AI to update FelonyBench, his chart of Anthropic incidents, pasting in a screenshot of a tweet about another incident. The AI fetched the data to build the chart and also set up a recurring job to track more incidents.
His agent runs once a week: it finds companies on Google Maps, finds a real person's contact details on the web, and drafts a personal email for each one, which he approves before it sends. Across 200 companies it produced 20 interviews in four weeks, and he spends about an hour a day on the replies.
His LLM agent reviews his robot vacuum's images and either sends the vacuum out or suggests moving things that would foul it up; works great.
Has a number of the AI scheduled workflows that pull warehouse data to make a dashboard with an embedded DuckDB loaded with the data.
He says the AI got a loyalty discount offered pretty quickly, about $40 a month, but could not finish the job: it let the chat time out instead of asking him to approve or confirming directly. Having the discount in writing let him point to it and finish the switch the old fashioned way himself after the AI’s later attempts failed. He still counts the $40 a month as saved.
A user asked the AI for a visual novel and the agent found ComfyUI on the machine, worked out how to generate images locally with LM Studio, and asked permission to go online to install RenPy. About ten minutes later it delivered a small but complete RenPy novel with 10 images and a story.
The AI sold the user's car end to end: listing it, handling buyers, collecting CarMax and Carvana quotes, and preparing the licensing paperwork.
Since the end of August he has added three new CI servers to his own setup. The AI wrote each NixOS setup and he checked the work, so now he can rebuild any of the servers from scratch in minutes.
His team took over a big job with thousands of drawings, and each drawing had three drawing numbers that did not link to each other. He uploaded all the PDFs and the AI built a spreadsheet that tied the numbers together.
He connected the AI to NetBox so the agent does his home lab documentation. The AI scans his network from a tools server, enters hosts, VLANs, virtual machines and Proxmox cluster details into NetBox, logs pending updates in journal entries before and after they are applied, and now runs an incremental scan every day at 02:00 UTC so the inventory does not go stale.
An engineer had just had the AI reapply local, uncommitted code changes. Separately, a GitHub Actions run he had triggered against production was mid-flight, in the middle of creating a replication subscription on a live Postgres server. He said undo the work it just started, meaning the local edits. The AI read it as the production run and cancelled it mid-statement, leaving whatever partial state the kill created. The local edits were not touched until he clarified.
While Dan Shipper's Dot displayed research for changing his flight out of San Francisco, it warned the new option might overlap with his head of video's Wednesday morning recording. The recording was not on Shipper's calendar — the Dot had read an unseen Slack discussion, clocked the proposed schedule and alerted him at the moment it mattered; the rest of the column is paywalled.
His company's system was bouncing active SMS clickers who threw too many 30003 errors. The AI Ultracode found the cause, and the fix means the sunsetting code now ignores active clickers.
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.