Takeaways from a workshop on Chain of Thought monitorability, and why we should reduce our dependence on CoT for AI safety.