Drop a CSV in, get a self-contained exploration UI out. The prompt below produces a page that infers column types, picks sensible defaults, and lets you switch dimensions.
I use it weekly for one-off data questions. The type-inference instructions matter more than the charting instructions.
I will paste a CSV. Build a single-file HTML page that visualizes it with d3 (CDN). The page must: 1. Parse the CSV from an embedded <script type="text/csv"> block (embed my data verbatim) 2. Infer each column's type: numeric, categorical (<=20 uniques), date (try ISO then common formats), or text. Show the inferred schema in a collapsible panel. 3. Default view: if there's a date column, line chart of the first numeric over time; else bar chart of first numeric grouped by first categorical 4. Controls: dropdowns for X, Y, group-by, and aggregation (sum/mean/median/count) that redraw instantly 5. Hover tooltips with exact values, axis labels with units guessed from the header names 6. A "table" tab showing the raw rows, sortable by clicking headers Handle up to 50k rows without freezing (aggregate before drawing).