Day 100: Find the story hiding in a small spreadsheet.

Hey, Ivy here. Day 100 is about reading a small spreadsheet without rushing past the boring checks that make the story trustworthy. A tidy-looking table can still contain blank cells, duplicated records, mixed formats, or numbers stored as text. Any one of those can distort a total or make a normal row look unusual. Attach a non-sensitive CSV or spreadsheet and paste this prompt: "Inspect this spreadsheet before drawing conclusions; list every column and its apparent type; count missing values by column; identify duplicate rows and the columns used to define a duplicate; flag inconsistent dates, numbers, currencies, categories, and text formats; then report patterns and anomalies with exact row numbers and calculations; use two sections titled Observations and Interpretations; put only verifiable spreadsheet facts in Observations, show the calculation behind each claim, and put possible explanations in Interpretations with uncertainty clearly labeled; do not invent missing context, and ask me questions when a field is ambiguous." Read the response in order. First compare the column list and apparent types with the sheet. Confirm the missing-value counts, then inspect the exact rows labeled as duplicates or format problems. For each reported pattern or anomaly, open the cited rows and redo at least one calculation in the spreadsheet or a calculator. If the AI cannot identify row numbers, give the table a stable row ID and ask it to run the inspection again. The final separation matters. Observations are things the cells and calculations support. Interpretations are possible reasons those facts occurred, and they may be wrong without more context. Keep both sections, because that record lets another person verify the finding without inheriting the AI's guess. Tomorrow, we will question a document before filing it so dates, references, and actions do not disappear.