A very fun game - it took quite a bit of fiddling to get an optimal solution using an LLM. Interesting as I haven't tried using them for 'unique' algo problems much. And then the day 9 puzzle broke my original solver (I had bounded areas that were unreachable to the horse so didn't actually score). Will be interesting to see whether the solver works on day 10.

It would be interesting to be able to change the wall budget for each puzzle to add some variation (with a max limit).