The problem

Yusuf types iph in the search box of an electronics shop. Before he types the next letter, a list drops down: "iPhone 15 Pro", "iPhone 15 case", "iPhone charger". He taps the first one.

The shop’s first version ran WHERE name ILIKE '%iph%' on the products table at each keystroke. At 1 million rows, every keystroke read the whole table, and the list showed up after the user had finished typing.

You design the suggestion service: 10 suggestions for what the user has typed so far, back before the next keystroke.