[anavat]

What if we loop it to itself? An infinite dialog with itself... An inner voice? And periodically train/fine-tune it on the results of this inner discussion, so that it 'saves' it to long-term memory?