Evaluating Long-Context Question and Answer Systems | Dark Hacker News