This blog post echoes what is said in the book MindSet (http://mindsetonline.com/). Never praise children because they accomplished a task. Always praise the effort done to do so.
Agreed. The use of ML is highly dependent on the data. Having a something like the Prediction api is fine, but seems like the use-cases would be rigid.
I certainly believe this to be a trend. I have been interviewing with different startups and they all have seemed to respond more to my github page than my resume. They all want to hear about personal projects and your thought process through them. Maybe I'm biased towards startups though.