Intelligence in TPT is interesting, and very hard to create, but I think i have a good method for it. Introducing the D-BOX method, in which a decision is made based off of data from the past collected from previous decision. Essentially, you have data storage, a decision box (Hence the name, D-BOX). The data gathered is either good or bad, if a decision is made, it is marked in the storage as bad or good, if a new decision is made, it is reas to see if this is a bad or good decision, then, if it is good, it is used, if it is bad, it is not used. This method should be good for brains due to the fact that it won't make as many bad decisions. However, it may have to have initial neurons that make the first 1 or 2 decisions. 

PS: This is a method made by a noob, so leave changes that I should make to the method.