Five days after launching Rustoku on crates.io (11 days of development), I've hit 1050+ downloads with zero marketing - validating there's real demand for educational algorithmic tools.
The core implementation uses bitmasking for efficient row/column/box constraint tracking, with cell-based candidate caching for performance. When human-solvable techniques are exhausted, it falls back to backtracking with minimum-remaining values (MRV) heuristic.
What sets it apart: beyond just solving, Rustoku generates a complete solve path for transparency - you can see exactly which techniques were applied and when the algorithm switches to backtracking.
The constraint satisfaction problem space feels ripe for more innovation, especially tools that bridge the gap between "just works" and "shows you how it works"