Yes, I realize that Levenshtein, the scoring algorithm used by default, can actually be worse than a simple `startsWith` if there are no misspellings. In the case of United States, if typed without misspellings, with Levenshtein, it only shows as the first option after typing "united s", or " us" (because ISO 3166-1 codes are configured as synonyms).

The default scoring algorithm can be replaced via the `scoreFn` param.

Probably a variant of Levenshtein, or something else that gives more weight to matches at the beginning should be used as the default, in order to get the best results in all cases.

I just pushed a new version that uses the Jaro–Winkler algorithm instead of Levenshtein, which gets much better completions with very few keystrokes, while retaining misspelling tolerance. Check it out!