It baffles me that even now people still use ChatGPT to learn new things. LLMs are the antithesis of learning. They're inaccurate so you're left reasoning from flat out incorrect principles, they save you from encountering neighboring information to the answer you seek which is a central mechanic for developing a mental model, and the further away you are from common thought patterns the less their training applies which makes them fundamentally and unfixably terrible at detecting and correcting misunderstandings.