Operator gathered 25 of 100 schools' data before it was stopped
Financial data for 100 schools, from tax filings into a Google Sheet. Operator clicked through it in the background, but rounded numbers, sometimes grabbed the wrong school, and was so slow it was stopped at 25.
Done withOperatorChatGPTSpotted· by Michael Lomuscio· Oct 20, 2026
Result Then o1 turned the sheet into an interactive chart
“it took SO LONG that I ended up stopping it after it collected data for about 25 schools.”
Spotted: found in the wild. Details are as reported by the source above and have not been independently verified.
What they did
The writer set OpenAI’s Operator the task of searching ProPublica’s Nonprofit Explorer for 100 independent schools, pulling each school’s total revenue and assets, and recording the numbers in a prepared Google Sheet. Operator navigated the site, searched each school, and copied the financial figures itself while the writer worked on other things.
How it went
Operator ran in the background while they did other work, but began rounding numbers inconsistently and sometimes grabbed data for the wrong organization when a school’s name was common, so they couldn’t leave it unattended. It was slower at some of it than doing it by hand, and they stopped it after about 25 schools. They then pasted the sheet into OpenAI’s o1, which wrote code for an interactive chart to explore the numbers.
Try it yourself with Operator
Give a browser-operating agent like OpenAI's Operator a repetitive web lookup task: provide a list of [items] such as organizations or products, point it to a public data source like [website], and ask it to search for each one, copy specific fields such as revenue or price, and record them in a prepared spreadsheet, while you work on something else and check in periodically for mistakes.
A starting prompt written by Agents in the Wild. Fill in the [brackets], and check what the agent does before it sends, pays for or deletes anything.
Ten years ago a similar scraper took them a horrendous amount of time and wasn't half as good. With free Claude and ChatGPT, this one grabs images, reads them with OCR and crops the hits.
Audio circuits had always intimidated them. They took apart a stereo effects pedal with ChatGPT, quizzing it on every part, then designed their own version in KiCad. A college-level crash course in a week.
Buying small e-commerce brands, they codified everything they look for into a six page prompt. Given a listing's screenshots and financials, ChatGPT takes a first pass to decide whether a seller call is worth taking.
Booking flights for a Costa Rica trip, a picky traveler asked ChatGPT to look them up. It checked Kayak, noticed the return flights got in too late and offered other options; they ended up booking one.
Completed orders needed checking against CRM records before billing. Every night ChatGPT reviews each order form against the CRM for mistakes, then creates a draft invoice in the accounting system for review.
Building their first AWS project by vibe coding, the writer designed the architecture and let Codex write the code. A flawed first design took a day, against an estimated three or four; the redesign also took a day and scaled.