Benchmarking LLMs with Challenging Tasks from Real Users - View it on GitHub