new hobby: asking the AI agent to rate its performance when it doesn't do the right thing > > Poor Communication: Instead of admitting I was on the wrong track, I kept suggesting parameter tweaks and adding complexity that didn't address the fundamental issue.