Here is a 'KBBNP vs krr' four-move chess puzzle created by the program, Chesthetica, using the 'Digital Synaptic Neural Substrate' computational creativity approach which does not use any kind of deep learning. There is also no proven limit to the quantity or type of legal compositions that can be automatically generated. The position below contains 8 pieces which means it simply could not have been derived even from an existing endgame tablebase which is presently limited to 7 pieces.
If you notice any version of Chesthetica 'skipped' from one problem to the next, that simply means additional (minor) changes were made to the program before it was set to run again. The material value is about equal. Try to solve this puzzle. Do try some of the others in the series as well before you go. Solving chess puzzles like this can also help improve your game.
