Yes I plan to dig into the fundamentals and mechanics a bit more than the layman because I may have to train models in the future (i.e. from better training data than Reddit for example).
But I also advocate just jumping into using LLMs and experimenting as well. Based on recent experience there is a good deal of difference between the models commonly available and the free tier is much more prone to "hallucinating" (fabricating information) - although that can be entertaining at times.
