Model literals, semantic aliases, and preference-aligned routing for LLMs | Not Hacker News!