Self-Adapting Language Models Is the Way to AGI? | Dark Hacker News