Fixing Claude with Claude: Anthropic reports on AI site reliability engineering | Hacker News Reader