The final lesson runs the whole finder over a realistic file tree the way you would use it for real - type a query, watch it rank, move the selection, and accept a path. It exercises every piece the project built, end to end.
Drive the finder over a realistic file-listing corpus through a full session, and prove the ranking and the accepted path are exactly what the design predicts.
This is the promise the whole project was built to keep: a working fuzzy finder, driven over a realistic project tree the way you would actually use one. Typing score narrows six paths to the two that contain it; both match the contiguous score run equally, so the shorter score.go leads the score_test.go on the length tie-break; pressing Down and Enter accepts the test file. Every layer is in play at once - the subsequence gate, the boundary- and consecutive-aware score, the dynamic program that finds the best alignment, the ranking and its tie-break, the selection, and accept.
From a one-line subsequence test you have built the honest core of fzf: fuzzy matching with smart-case, a real scoring model, an optimal alignment found by dynamic programming, ranked results with extended query syntax, highlighting, incremental narrowing, a bounded top-K fast path, and an interactive finder over a live corpus. The empty-query and no-match edges behave, the tie-breaks are deterministic, and the same pipeline serves both the scripted session and the -f filter. That is the core every real fuzzy finder - fzf, skim, selecta - is built around.
// Build the finder from the corpus, then run the scripted session:// type "score" -> Down -> Enter// Assert the header count (2/6), the ordering (score.go before// score_test.go by the length tie-break), and the accepted path.// This is your finder, used exactly as a person would use it.