The variety of tasks available in openAIGym makes it easy enough to use even if you have no prior experience with RL, but more importantly are solving real world problems (e-commerce recommendations).
I also appreciate how much effort has been put into making sure things run well under all environments/platforms they support - not just their own Windows environment! It would be great if we could get better documentation about using external libraries such as PyTorch rather than relying so heavily upon Keras models supplied by them instead. Also there's some inconsistencies between what works when training within each task versus transferring across different ones; this can sometimes lead us down paths where our agent isn't actually doing anything useful because at least one step was required before an action became possible again after having paused execution due to lack of progress / reward generation.