thinking a little of a simple logical system like, say we have a tiny handful of operators and constants, and the code tries all combinations of these to reach an output value from an initial input state. this code might happily putt along, reaching the output state from the input state over and over and over again in various ways. it could just loop. this loop would model my intended processes of pursuing and improving goals. [to do this better, one might have the loop develop information useful for future goals from the behavior.] then a second loop would be the malicious dissociated part. it’s goal would be to be in complete control of the first loop’s ability to meet its goal, both immediately and in the future by managing the first loop’s exposure to learning. [when thinking of this i briefly have a more refined sense of some of my issues] it’s nice to code over and over, never get very far, something is missing. things near the “inverted sweet spot of learning” concept seem interesting to add on to it.