Hmm, would be interesting to see a "race" between this and isearch; from the examples, it looks like this has an advantage only if you have lots of similar strings, so "the one you're looking at" vs. "typing more letters of the word" is an advantage? (externally timed, see The Humane Interface for details on how self-reporting doesn't actually work to measure this kind of thing)