Ask HN: Are there any objective measurements for AI model coding performance?