It’s tempting to copy-paste traceback directly into an LLM to solve your problem and get on with your work. In the short-term, this gives you insane productivity boosts but I still try to solve problems by scouring through dozens of stackoverflow threads and documentation pages manually.
When you ask an LLM, you get a solution to your problem right away. It may or may not work, thats not the point of discussion here, but the fact that you do not have to go through the process of searching the solution is key.
Reading documentation without the help of LLMs or a RAG system forces you to go through heaps of information irrelevant but tangential to the systems you are working with. You learn a lot more than just the problem at hand. You might stumble across a solution for a completely different problem, or a new way to implement something. This might seem like a waste of time but it makes you a better engineer.
But what about StackOverflow? StackOverflow lets you skips directly to the answer. Not quite, StackOverflow answers include a variety of solutions, including discussion about best-practices, correctness and deprecation. You might also be interested in reading similar questions, with even more valuable knowledge. LLMs might present you with outdated, incorrect or dangerous solutions. People on the internet love to point out how wrong others are. Use that to your advantage.
In the long-term, its better to solve problems at-hand using your own knowledge, use documentation and forums to increase your surface-area of knowledge.
and my theory is somewhat aligned with some of the research on this topic: