Let me start by saying that its HARD to make it assist you learn in a good manner. I use AI to assist me learn new things all the time, but it always takes a couple of prompts to mold it into the assistance I need, and even then it gets to overstep ones in a while. I’ve tried adding instructions and whatnot, but nothing is sufficient. Keep in mind I do not use the most expensive models, so that might be part of it.
What I’m currently learning nowadays for example is to build an app using said and said technologies, incorporating certain libraries etc. My tactic is to make AI act sort of as a freeCodeCamp curriculum, splitting things into small tasks and going step by step. Usually I would use it to devise a larger working plan. One big issue AI tends to lean on is to leave a lot of specific guidance, which later it follows silently, yet that guidance might be wrong assumption at that time. So its important to always make it doubt and elaborate when making decisions, or providing solutions, and also not leave too much preemptive guidance, until the point to implement something comes by.
Another major struggle is to make it build at the right pace. It would either spit a large successions of steps, or it would feed you small spoons, so it usually takes a few prompts per session to get in the right pace.
When we talk agents, they always tend to rush to implement things and would take hint for assistance as a nudge to write code, another thing that needs to be overwritten and established per session.
As someone that has some experience, I do have my own opinion and vision how to do things, and I use AI to lay down the options in front of me, for me to question them and come (on my own accord) to the right solution and then implement it. Sometimes I ask it to provide me only the steps, do this and that, and I implement it, but most of the time id also request for code snippets simply because it becomes a redundant task to write the same functionality over and over. I still follow the snippet and write on my own, just to follow the logic and eventually doubt it, or tweak it, in order to understand whats going on.
Now keep in mind this learning strategy is totally adapted to my needs, and might not fit for someone who is learning recursion, or goes through the actual fCC curriculum, or needs to focus more on coming up with complete and thorough solutions on their own. In that latter case, id use AI more for asking more local and specific questions, that commonly come when you doubt something which is not explicitly told in the curriculum lesson.
I could go on with tips and my own learning experience though AI, but this already becomes too long
