I'm not sure appealing to survivorship bias is useful here, since presumably you are also a survivor and just as subject to it. But my argument was based on my

I'm not sure appealing to survivorship bias is useful here, since presumably you are also a survivor and just as subject to it. But my argument was based on my observations of other people trying to learn. My experience learning was quite different from most that I've observed, and my observation has actually been that the way I learned is the one that works for fewer people. This is an area of interest for me, so I take opportunities to see how people learn as much as I can. If anything, the better argument against what I'm saying is that it's just anecdata, which is true. I'd welcome research to prove this one way or another, though. Tools like this are quite unlike what I had available to me, and I think they are a much bigger step to creating learning environments for real computing than trying to shoehorn natural language into a box it just doesn't fit. We have been down that road so many times and it never works out the way proponents suggest it should.