Estimates
are guesses too.
A number that wouldn't fit
Five short cards, 88 px tall, cut off their values. Four of them have a subtitle that wraps onto two lines. The agent estimated the fix from font sizes and theme settings, and each estimate sounded convincing.
Predict first: the agent's first idea was 4 px of padding on the visual's outer frame, down from what it believed was 14 at the top and 16 at the bottom. Does that free up room?
Pixel sizes are approximate, fitted to what Desktop actually drew in each round.
What each round taught
| Round | Based on | Result |
|---|---|---|
| 1 · outer padding 4 px | An estimate. Cards turned out to have no outer padding, so 4 px took 8 px away. | worse |
| 2 · label off, card padding 4 px | A better estimate, after reading the base theme properly. The 14/16 px belonged to the card's inner box. | 1 of 5 fixed |
| 3 · space above the value | A test on the real card, closed without saving: 4 px still cut off, 0 px fitted. | all 5 fixed |
Two builds were spent on estimates. One look at the real tile settled the last question. Each round still went through the seed first, so every setting in the report was copied from Desktop's own file.
The rule, in one breath
Estimate to choose what to try. Test on the real tile to choose the value. Build from the seed.
An estimate from font sizes is a guess about how Desktop draws. A look at the real tile, closed with Don't save, is evidence. Only the seed says how to write it.
Next: making the seed readable, so anyone can find the example they need.