Been working on trying to generate versions of this puzzle I found many years ago.

But, trying to generate different sized grids with some other restrictions that affect difficulty (like how many moves from the start to the end)

www.tylerauer.com/jumping_jess...

I guess, in a way, that is backtracking... :)

It's actually faster to just fail the puzzle generation after a second or so and then try again from scratch (with a new random seed) than to wait for the algorithm to try every possible state.

One puzzle I'm generating is on a grid. Some grid states are obviously invalid early, so we can backtrack.

But, some checks (particularly ones that affect the puzzle's difficulty) aren't possible until the grid is filled. Which means the backtracking doesn't help much.

In hindsight, this is should maybe have been intuitive, but it didn't click for me until I wrote the code myself, and then watched it take...forever.

Nice. This is awesome. Thanks for sharing. I love that your tools also look beautiful.

I'm building some stuff right now and my tools are all CLI. It's kinda fun in it's own way, but I'm excited to get to building out some real UI.