data-error-drill
Can it catch?
Is your AI quietly cutting corners — results that look tailored, logic that slacks off? This page preps an exam for the checking tool you built: plant known errors into a copy of your data — flip a digit, drop a row, shift a date — and see at a glance which ones it misses. Everything runs in this browser; no file is uploaded anywhere.
1Data
Drop a .csv here, or click to browse
first row = header · your original file is never touched — drills run on copies
first row = header · your original file is never touched — drills run on copies
2Errors to plant
load data first
3Drilled copy & answer sheet
The drilled copy (CSV) and its answer sheet (JSON) will appear here.
Feed the copy to your tool — then come back and score its report in step 4.
4Score your tool's report
Paste what your tool reported on the drilled copy (any text — error list, log, summary). Scoring is evidence-based: a fault counts as caught if the report mentions the planted wrong value or the affected row's key.
generate a drilled copy first