I think it could use some adjustments.

If you search for "alias", it appears to search for modules whose documentation matches /\balias/i before it searches for modules whose titles match /alias/i. That is, it lists several modules that use words like "aliases" in their documentation before it lists the module Devel::LexAlias.

For many searches it returns things for which I could find no sensible relation to the search term. At first I thought it was eventually going so far as to take the first letter of the search term and search for things matching /\b$letter/i (that is, things that contain "words" that start with the same letter). But more testing shows that it certainly doesn't always do that...

Perhaps it is stripping letters from the end of the word being searched for (we'll refer to the resulting shortened "word" as $part) and searching for /\b\Q$part\E/i and it will only strip so many letters? I think it should also refuse to strip a word down to a single letter and probably also stop stripping letters if it has already found quite a few matches.

But overall I think it will be very useful, especially if it stays so fast! (: