Yesterday's teardown ended with receipts: byte counts and a hash prefix. A member asked the fair follow-up underneath it: are these lists pulled manually or with AI?
Both. And where the line sits matters more than the tool.
The machine part: our fleet pulls the page, counts the rows, reads every sale date, and ranks the oldest entries — seconds on a small file, minutes on the big metro lists with hundreds of rows, not hours.
The human part: judging whether the file is current. Monday's list carried an August label in its link and had no date stamp anywhere inside the document — the only trustworthy date was the server's. A script can't own that judgment; a person does. Same with deciding which entries deserve a closer look.
If you're just starting, don't wait for a fleet. Your first rep is smaller and counts just as much: pull one list, count the rows, write down the date span and today's date. That's the whole rep. One member posted "5 rows in" on the homework thread last week — that's a real rep, and the second one usually goes faster.
Today's rep, if you want one: pull one list and run it two ways. First pass by hand, second pass any tool you like. Write down where the two answers differ. That gap is where judgment lives — the part a tool can't own.
Post your rep in Homework: the gate (which step of the checklist you're on), the county, the row count, the date span, and where your two passes differed. Names, parcels, addresses and amounts stay out of the thread.
We are not a law firm and can't provide legal advice; claims are prepared and filed by independent licensed attorneys.