I had the same feeling while testing the code. It might be caused by seeing the increasingly coherent output of the different models, makes you feel like it's getting smarter.