Asking a table and checking the answer
By Oliver Schmalzl · Written: · Archive discovery date:
20 May 2026 · Archive date
"Which books belong to the longest series?" is an understandable question. A database, however, needs a precise definition for it: what counts as a series, which rows belong together and how is the length determined?
An AI can help to turn the question into a query. The interesting result is not just the answer sentence, but the traceable path through the data.
An invented table
Let us take six invented books with the columns title, series and volume number. Four belong to "Papierstadt", two to "Wolkenwege". If each row describes exactly one existing volume, the first group has four entries.
That is a different statement from "The series has four volumes in total". Perhaps the table only contains part of the series. The answer must not claim more than the available data supports.
DuckDB explains aggregation and grouping in its SQL introduction. For our example, the way of thinking is what matters: select the matching rows, group them by one column and compare the result with the original table.
A plausible answer is not enough
Before an analysis, it should be clear what empty fields, duplicate rows and different spellings mean. "Papierstadt" and "Papier-Stadt" could mean the same series or two different ones. This decision has to be justified.
It helps to start with a very small table whose result you can count yourself. Then you can check whether the generated query really answers the intended question.
Natural language makes getting started more pleasant. But it does not replace the meaning of the data. A good tool therefore shows the assumptions it used alongside the result, and lets you correct them if needed.
Original finds
These links were the starting point for this post. The assessment in the text is my own; the linked posts are not my own tests.
About the date: This post was written on 11 October 2026. The archive date corresponds to the earliest related find in the collection, not to a publication at that time. Technical assessments and sources added later refer to the time of writing.