11 nanoseconds per word: Ridiculously Over-Optimizing Anagrams | Wiredream - Dave Andersen's blog

David G. Andersen released a Rust-based anagram solver that indexes all 178,690 entries in the Scrabble dictionary in 31 milliseconds, or 173 nanoseconds per word, after building a collision-free prime-product identifier that fits every dictionary word into a single 64-bit integer. The solver targets a new Word Ladder variant in which players mutate words by single-letter changes, additions, deletions, or anagramming, and the anagram index is a stepping stone to a full ladder solver promised in a follow-up post. Andersen cut the baseline cost by eliminating the per-word Vec allocations that dominated the original hash-table build, which produced roughly 161,019 distinct anagram keys. The work matters as a case study in how allocation overhead, not algorithmic complexity, drives cost in dictionary-scale string processing, and the prime-product trick offers a compact alternative to letter-sorted or letter-count keys. Watch for the next post, where the anagram index feeds the complete word-ladder search.