Not arguing with your anecdote, but I expect LLMs to respond better with question type queries. Due to learning with books, I tend to use a keyword based to my

Not arguing with your anecdote, but I expect LLMs to respond better with question type queries. Due to learning with books, I tend to use a keyword based to my search, and while reading I tend to collect technical terms as that how they will be referred most of the times. As for your example, I would type "c++ string literal std::string" as the query. Or I'd probably look for the documentation of std::string and seek the answer there. Why? Because books, articles, and documentation often have related pieces of knowledge next to each other. It's like experiencing a whole journey instead of a short slideshow of pictures and video clips. Maybe that one specific answer is all you needed, but having the complete picture helps me refine my ideas better. DuckDuckGo (What I use) is still useful for me. I either know the resources I'm looking for or I'm hoping someone has written about it. And the keyword based approach works great. It's like the index at the end of the book and it's table of content. You look at them to see if what you're seeking in in this book and where. It's just a short stop to the actual information.