Coherence-Based Cognitive Architecture — LessWrong