Critic: Large Language Models Can Self-Correct with Tool-Interactive Critiquing | Dark Hacker News