Fresh-data testing

Build-along · ~10 min at your own keyboard

A model asked to test its own work generates test cases that flatter its own assumptions. The fix: test with data it has never seen, like a real closed listing from last year or a comp set you know is weird. Fresh data finds the failures polite tests never will.

Your starter: County roll refresh

Prompt

Make it yours: Name the roll you pull and where the last one sits, set what counts as a change on your parcels, and keep the list of things it must never do.

Compare the new [county] roll at [path] with the last one at [path].

Report, for the parcels on my watch list at [sheet]:
- Owner change: old spelling, new spelling, both exactly as the roll has them.
- Assessed value change: old, new, from the roll.
- Delinquency change: years, amount.
- Parcel gone from the roll: say so; do not guess why.

Anything you would need another record to explain is [PULL THE RECORD: which record]. Record first, then the number.
Write the report to [folder]/roll-diff-[date].md. Do not change the watch list.

The coach checks that what you paste back is your own work. The starter pasted back unchanged won't pass.

Do this

0/4

Share what happened

Paste what fresh inputs you used, what broke, and the category of failure.