Guide

Turn messy notes into a table you can check.

Choose the columns, distinguish zero from missing, and preserve conflicting information instead of guessing.

Updated 2026-10-13GuideIntermediateAI-assisted editorial

Try these prompts in an AI chat app you already use. This page does not run a model. The app’s own account, privacy and pricing rules apply.

Define what belongs in each column

A table makes notes easier to scan, but every filled cell still needs support. In this exercise, each source record becomes one row. The columns are Record, Stall, Signs needed, Ready day, Evidence, and Issue.

Use the fictional notes below. Record IDs let you trace a row back to its source. Keep three states distinct: a stated zero, an absent value, and two incompatible values. None should quietly turn into the other.

Copy this prompt and its fictional notes

Prompt to try
Convert the fictional notes below into a Markdown table with these columns: Record, Stall, Signs needed, Ready day, Evidence, Issue. Make exactly one row for each record N1–N4, in that order. Use only the notes. Put Not stated for absent values; keep an explicit zero as 0. If values conflict, retain both in the cell, label it Conflict, and explain the issue. Do not choose a winner, infer dates, or calculate a total. In Evidence, copy the relevant source wording exactly. Use None in Issue only when the row has no missing or conflicting values. Notes: N1: Cedar stall — 8 paper signs needed; ready Friday. N2: Fern stall — paper signs needed; ready Saturday. N3: Moss stall — 0 paper signs needed; ready Friday. N4: Willow stall — one note says 5 paper signs; another says 7. Neither is marked final; ready day not recorded.

Check the four rows against the source

Editorial checklist: N1 has 8 and Friday, with no issue. N2 has Not stated and Saturday, with a missing-quantity issue. N3 has 0 and Friday, with no issue. N4 retains the conflict between 5 and 7, has Not stated for the ready day, and flags both problems. These checks were written for this lesson; they are not a model transcript.

Check that all four IDs appear exactly once. Compare every evidence quotation with the notes, then confirm that it actually supports the adjacent cells. An accurate quotation beside an unsupported number is still a bad row.

Resolve the gaps with new evidence

Follow-up prompt
List only the questions a person must answer before this table is complete. Refer to each record ID. Do not propose values or silently repair the source notes.

The questions should cover Fern's quantity and Willow's final quantity and ready day. In a real task, ask the person responsible for the records through the usual route. Save their clarification with the source notes, then revise the affected rows. Do not calculate a complete order total while quantities remain unresolved.

Keep structure separate from correctness

Google's prompt guidance supports requesting a table and defining constraints. Anthropic recommends permitting uncertainty and checking claims against evidence. Those techniques help specify the job; they do not guarantee that every cell is right.

For this four-row task, inspect every cell. If you later copy a reviewed table into a spreadsheet, check the imported columns and values again. Keep the original notes alongside the table, and use only information you are permitted to share.

Sources & further reading

Open the source to check the original wording and current availability.

Hey, I’m Patch.

Your little guide to this website.

Ask about an AI word or find your next small adventure.

New here? Try “Help me get started.”

Prepared explanations + site search. No live model. Your question stays in this browser. How Patch works.