If the same class of problems need to get solved on a regular basis, could the guessing algorithm get smarter with each iteration using ML techniques? Seems li

If the same class of problems need to get solved on a regular basis, could the guessing algorithm get smarter with each iteration using ML techniques? Seems like if the algorithm is applied to a particular domain, there could be efficiencies gained from optimizing what guesses to rule out. Not very knowledgeable about the subject though.