result

Summary

The estimate, broken into its parts, edited in place.

Raw wireframe

• .context/designs/mobile/scan/result.html

Why it is drawn this way

Components, never a single number. Cal AI shows one confident figure and is wrong a lot. Showing the parts is what makes the follow-up question feel reasonable instead of nagging, and the low-confidence rows are the argument for tapping Improve. See confidence.

One file, where there were four. scan/editing.html drew this plate with working steppers, scan/same-photo.html drew it with a cache banner, and corrected draws it after a rename. The app has one result screen, so three of the four were the set disagreeing with itself about what the user is looking at. The first two are deleted and their content is here; corrected survives because the "You said" state genuinely reads differently.

Three things came off, and each was a caption for a control that was already visible. The "same photo as last time" banner said what the quota meter says, in the place a user who cares about being charged is already looking. The "two questions, much better number" card was the amber Improve button with more words, directly above the amber Improve button. "Tap any item to change the portion" was explaining steppers the user can see.

When came on. The slot, day and time sit under the title, the same pair meal carries. Both controls are sheets over this screen, so neither is a route — see backdate.

Remove and Add came on together, because they are one complaint with two signs: the estimator invented a coleslaw, or it missed the egg. Neither can be answered by nudging a quantity, and a row nudged to zero is a weightless line in the log for ever.

Renaming is much the rarer of the two complaints and now sits one tap further in, inside the editor. See portion.

Rows are two lines, not three. The stepper shares the second line with the confidence chip rather than owning a row of its own, which is what fits four components and the footer on one screen without scrolling.

Green is the primary act and sits on the right, under the thumb. Amber keeps the ask — amber is reserved for the app asking, see the-ask — but takes the narrow half: most plates are right enough, and the screen should not argue with someone who agrees with it.

Related

• confidence
• the-ask
• correct-item
• correct-dish
• fat

Sources

• .context/designs/mobile/scan/result.html