The main difference versus IDEs is that everything we show is informed by all the public code we've collected from the web. So e.g. there are a ton of arguments

The main difference versus IDEs is that everything we show is informed by all the public code we've collected from the web. So e.g. there are a ton of arguments to matplotlib.plot and IDEs can show you them all ranked alphabetically, whereas we can show you common patterns of how people actually use matplotlib.plot in practice, which is often far more useful. (We'll show you the docs, too.) Another example is if you type "load('abc.json')" without having imported json: there are hundreds of python packages that define a function called "load", but "json" and "simplejson" are by _far_ the most widely used, so we can suggest that you "from json import load". That's something you can't do unless you have a good model of a lot of real-world code.