LLMs are very good at text reading. LLMs read tokenized text, while human use eyes to view words. Another scenario is that ChatGPT is good at analyzing cpp template error messages, which are usually long and hard to understand for human.